Home Electronics

Anything interesting to note down about the various A/V or computer gear is here.


In general we have a lot of older electronics, mostly not worth too much but functional. Either I bought it new and have kept it a long time, or I bought it used.

Log (belated)

Wednesday, September 8, 2010

Picked up the new Panasonic 50" plasma today, and got it installed on top of the new entertainment center. Nice! About $1100 total. They cheaped out on the built-in speakers though.

Thursday, September 23, 2010

The new plasma TV (Panasonic TC-P50G25) is great, except for the nasty little built-in speakers. I could see them through the back grille: small unbaffled ovals, and facing down. I bought some 1/8" mono phone plugs today and opened up the TV, many screws worth, and tapped into the speaker wires, one on each side. When the plugs are in the built-in speakers cut off. I have some old Radio Shack Minimus 7's that ought to work well.

I put it all in the TV, and unfortunately the bodies of the connectors are not insulated from the connections. The TV noticed this and made a horrible screech upon power-up, and shut itself down. I removed the back of the TV again and removed the connectors from it, whereupon it worked just fine. I'll need to insulate them, I guess. For now I've left the back off. Proper speakers help a lot.

Friday, October 8, 2010

Long past time to finish the TV! I drilled the mounting holes out larger, then used a razor blade to slice thin bands of heat-shrink tubing to slip over the barrels of the connectors. I used a gasket punch set to make four pasteboard washers (out of a soda carton) to complete the insulation job, then I installed the connectors into the back. The multimeter showed that the barrels were now isolated from the metal back, so I reassembled the TV. It worked this time. We can now put the TV stand back in its proper place.

Wednesday, April 20, 2011

Tore apart the wretched Canon MF4270 copier/printer. (A Goodwill special that's never worked right, but as I'd invested in a 'new' toner cartridge that was outrageously priced I don't want to just pitch it.) It never wants to print, always with a false paper jam, but you can usually make it cough out a page or two by faking a paper clearing to go with the fake jam. (Remove toner cartridge, blow on things, put it all back in.) Couldn't find anything, but sometimes the laying on of hands does some good. We'll see. You can get all the plastic skins off and run it that way, it's not too bad to fiddle with. [No, this didn't fix anything.]

Wednesday, September 7, 2011

Recently I acquired the materials to try to make a wall mount for the basement's 42" (Panasonic TC-42PX24, $400 at a liquidation outlet last November) plasma TV: 1/8"×1"×3' angle iron, ×4, two heavy bolts with a bit of shoulder for hinge pins, six flat washers and four nuts, four M8×1.25×20 metric bolts sized for the TV's screw sockets, with lockwashers, and four lagscrews. The intent is that one rail is screwed to the wall, horizontal face upwards, and the other is screwed to the TV, same orientation, so that the TV rail rests on top of the wall rail, thus making what looks like channel iron with the opening facing down. (All this ×2, of course.) The rails are then hinged together with heavy bolts at one end so that the TV can swing out from the wall approximately 90°, so that Jill can watch it while treading, or for access to the connector panel. (The treadmill is on the same wall as the TV, and is unlikely to move from there anytime soon.) Today I finally started, it'd be nice to get it up and in use before the TV is more than a year old. I clamped two rails together in their final orientation and then drilled a tight-fitting axle hole through them both, using oil, and trial-fit the bolt. Seemed OK. I then used the grinder to relieve the wall rail's corner so that the TV rail could swing out freely. I then tightened down the bolt and used a second nut to lock the first. I then repeated the procedure for the other assembly. Daniel was a second set of hands, which helped a lot. That was all the time I had today. Looks good so far. (If this doesn't work out I can always get one of the more expensive ready-made mounts, but as I did already pay about $45 full retail for all this hardware I'm motivated to make it work.)

Thursday, September 8, 2011

I placed the TV face-down with the to-be hinged side about an inch from the wall, then opened up the rails and placed them against the wall, thus modeling the opened in-use position. I marked the centerlines for the holes on the TV's two rails and took them to the drill press. (Daniel's hands were again of some use here.) I had to open one hole up a bit because of inaccurate marking (or drilling), but otherwise the rails bolted nicely to the TV.

Saturday, September 10, 2011

I made a TV-sized rectangle out of newspaper and taped it up on the wall in what looked like a good position. After considering it for awhile I then measured where the rails were on the TV and marked the wall where the ends of the rails would go. I drew level lines on the wall and found where they intersected the studs. I drilled holes in the top rail and mounted it to the wall with big lagscrews. Resting the TV on the first rail (hinge pin removed) I fine-tuned the position of the second rail relative to the first, and marked and drilled its holes to match. (Daniel was helping again, which was nice.) I had to relieve a portion of the bottom TV rail so that the antenna jack was better exposed, that was a fun session at the bench grinder. (The two HDMI jacks are pretty much completely covered, but at the moment we don't need to use them anyway.) I put the hinge pin in the top rail, swung the TV away from the wall while holding the TV up, then with our three remaining hands mounted the bottom rail assembly to the TV and then to the wall. This was a fiddly operation but it went smoothly. It would have been very difficult to do alone. I tightened the hinge bolts down, I may apply Locktite later. I then loosened the lag bolts and adjusted the rail positions a bit so that they both take weight when the TV is swung against the wall, then tightened everything to final specifications. There is some flexing of the 1/8" steel when the TV is pivoted out from the wall, but I don't think this will matter much. It seems very secure, and looks and works well. Success!

I then got out the Pioneer DVD recorder, which is (as before) going to be used to convert the S-video from the home theater system to something the TV can eat. I hooked it all up and then Daniel and I watched a movie. Sweet. I also put a DVD in the recorder and played it there, thus using the TV's own downwards-firing speakers for sound instead of the big boys. It worked better than the TV upstairs did, probably because there is more space beneath the TV at this time for sound to escape. (This may change once there is a couch under there.) The speaker design of this series of Panasonic TV's is very poor, I don't think they expect you to actually use them.

As a final step I got out the big roll of garage-sale RG-6 and ran a long wire to the computer room's CATV distribution amplifier. (I re-used the hole in the wall where the Ethernet cable to the old computer cluster used to come through. In fact the cable was still there and routed around the room and back through the door of the computer room, so I re-routed this through the wall between the two wiring closets, along with the new RG-6, much neatening things up.) The new used die-crimper for RG connectors worked very much better to put the ends on the cable than the plier type I was using before. This new antenna wire was routed with the rest of the home theater wiring (three front channel speakers and the S-video cable) underneath the throw rug to prevent tripping. With the antenna now hooked up I ran the TV's channel training cycle and found all the channels we get upstairs. After removing the duplicates and dross we now have two fully-functional flat-panel plasma TV's in the house. (This new installation is a bit smaller than upstairs, but we also sit closer. The one-generation-older screen technology isn't as bright and is more reflective, but the room is generally darker. Vertical resolution is only 720, not 1080, but it's a rare bit of source material that can exploit this. [We have exactly one over-the-air channel in the higher resolution.] I'd say they're effectively equivalent, except that the downstairs TV doesn't have the internet connection and so can't do Netflix and YouTube. There is no doubt, however, that this new installation represents a better value, and any randomly-chosen computer could tie it to the Internet.) To celebrate I watched a bit of NASCAR in the new location, it was working well. The TV's own speakers are a bit on the puny side, but we're stuck with that unless I want to get a long optical cable and tie the TV's sound back into the home theater (HT) system. Not today!

Friday, September 16, 2011

OK, today was the final step. I'd ordered a 50' TOSLINK optical cable ($15, shipped) and it came today. I hooked it up between the TV and the HT system, and it works! I was pretty sure that 25' wasn't enough, I wanted some slack and so bought the longer length even though some equipment supposedly finds that to be 'too long' due to signal attenuation. Now the broadcast channel sound can go through the HT system, if we want it to.

Thursday, September 22, 2011

My wife broke her LG phone's car charger yesterday, for the third time. Twice she's managed to snap off/yank out the cord where it comes out of the cigarette lighter plug; I was able to repair it both times using heat shrink tubing and glue to bolster the weak joint. This time she managed to bend the connector where it goes into the phone! Surprisingly, that connector comes apart with careful lifting of some tiny snap hooks. I found that the metal spring contacts inside were fractured, as was the plastic substrate; It's dead, Jim.

Taking her non-functional house charger, at her suggestion, I cut the plug off of it and grafted it onto the car charger, using heat-shrink tubing to secure the joint. It seems to work, I put her phone on charge since she had depleted it. Oddly enough, the 'non-functional' charger supplied power on its wires, and the connector also had continuity. I suspect the problem may have been that the phone draws about 280 mA (from the 12 V supply) while charging, yet the house charger was rated at only 180 mA at 5 V! The voltage may have been insufficient to actually charge the phone if it was overloaded, it looks like the charger might have been the wrong one. (Perhaps she's inadvertently swapped it with one of the myriad of other rechargeable phone-type devices she's gone through over the years? [Yes, she must have. This adapter was branded Plantronics, which is a phone accessories outfit. Oops, let's hope it's not for her current gear!])

I also used a dot of Shoe Goo to secure a white paper chad over the wretched laser-bright blue power LED on the charger, said LED having been a sore point between us on several car trips. (I hate blue LED's, and overbright ones in particular. The blue receptors in our eyes are by far the weakest, and blue point sources always appear to be fuzzy and out of focus.)

As a joke I wrapped a Band-aid around the repair. I don't know how many more times I can resurrect this thing...

Thursday, October 6, 2011

OK, maybe today was actually the final step for the basement TV. I screwed the 80" Sony projection screen to the ceiling. One of its crosspiece table-stand legs aligned with a floor joist when it was in position, that was lucky. I just drilled a hole in the center of each 'foot' and ran drywall screws (with washers) up into the joist. I used the screen rod as a prop against the fire extinguisher on the side of the kitchen piping concealment post to hold it up so that I could work on the other side, since I was working alone. On the other side the feet were not on a joist, but there was a joist very near the end of the unit. I took a heavy piece of wire (part of a salvaged metal bail from a dead plastic bucket) and made a hook to catch in the end of the cassette, and then ran it upwards behind the unit to the joist; I bent a screw eye in the other end of the wire, then used a drywall screw to tie it into the joist. I also put a screw through the foot into the plaster just to give it some more lateral stability, and to make it look like the other side. It's solidly mounted, and the screen works well in that position. I may have gotten it an inch or so too close to the wall-mount TV, when the support rod swings down (and I can reach it without having to stand on anything, which is nice) it wants to scrape on the TV unless you're very careful, and the rod ends up right in front of the power switch. It's easy to activate the TV when setting up or taking down the screen. I can always move the screen out from the wall later. The popcorn ceiling will hide many sins, the extra holes will be nothing.

The original plan was to put the screen on a small behind-the-couch table, or perhaps a shelf equivalent, so that it was completely out of sight when not in use, but I just don't see that happening anytime soon. Today's approach has the major advantage of being done, and functioning well enough. Just not quite as aesthetic. Maybe. It's certainly a lot more secure and out of the way like this, though not as stealthy, which was one of the big advantages of the projector as we had originally envisioned it.

I spent some time re-setting-up the projector, getting the lens adjustments just right. It's not quite as impressive as it was when we got it, I think there are a couple of factors at work:

  1. In its new position the screen's actually a foot or so further away from the primary seating than when it was propped up on the old console TV. As we found out with the 50" TV upstairs, that kind of move can make a surprising difference in subjective size. I guess there's nothing preventing me from moving it out from the wall, which would increase its subjective size. That is the advantage of a ceiling mount, though I can't do this too much or it'll impair the view from the secondary seating positions.
  2. The 50" upstairs TV, which we got after the projector, has perhaps acclimated us to larger sizes.
Still, I watched part of a movie and it was very nice. It is bigger than the upstairs TV, and it does have a reasonable sound system. I have no idea what we're going to do about the center speaker once there is a couch on that wall instead of a bookcase. We'll deal with that when we get there, I guess.

Sunday, November 13, 2011

The new batteries (4) for the Uniden 5.8 GHz TCX-800 phone cluster came yesterday, I replaced the two that were definitely bad. $3 each, quite tolerable. BT-446 (3.6 V 800 mAH Ni-MH batteries, labeled BG0004), I ordered four. We have six handsets scattered around the house, plus the three wired phones and the original AT&T cordless. Plenty of phones on the ol' landline.

Tuesday, November 29, 2011

For a family Christmas present I bought a Panasonic BD35 Blu-ray player at the liquidation outlet. $60, with a 30-day warrantee and complete with an HDMI cable. Though discontinued, it has a good reputation. The choice of this particular unit was swayed by the fact that it would integrate with the Panasonic TV, sharing the remote. Blu-ray is not a particularly attractive format due to the continued high media cost and the invasive and ever-changing attempts at copy-protection, but there are some programs that are worth having in high definition. The BBC's nature documentaries, for example. Once the price of a player had dropped enough I finally jumped on one. I anticipate that we will have very few Blu-ray discs, DVD's work very well for most things.

Wednesday, November 30, 2011

I ordered a boxed set of the BBC's Natural History collection from Amazon. $50, used. The whole purpose of this Blu-ray excursion, really.

Thursday, December 1, 2011

The BD35's firmware was old, v1.4, but the player is still being supported by Panasonic. I was unable to get its network firmware update to go these past few days so I downloaded the latest (v2.4, October) firmware self-extracting-but-only-for-a-PC .EXE to the nasty little $20 Compaq Presario 1200 laptop that we keep around for just such purposes and unpacked it there, then FTP'd it to the G3 and used Toast to burn it onto an ISO 9660 disc. I fed this to the BD35 and it went through its firmware update cycle and seemed happy about it.

The BBC collection came today, so I had something to test with besides regular DVD's. It was able to play them, but the picture quality I was seeing on my tiny little yellow-plug test monitor was so bad I needed to prove it out on the big TV. When nobody was around I took the BD35 upstairs and plugged it in to the big plasma. I must say that the control integration worked well, putting in a disc and hitting play switches the TV to the Blu-ray player HDMI input, etc. I got the settings all dialed in and watched some of the program, it looked great. I expect no problems Christmas morning, so I put it all away 'til then.

Jill pointed out that her Pismo's screen hinges seem to be broken, the screen now flops down. I ordered some used but good ones on eBay, $10, but I suspect that we're going to be upgrading this machine soon anyway. (The natives are restless...)

Saturday, December 10, 2011

For the last couple of weeks Jill's Pismo (Mac Powerbook G3) screen has been flopping down, the hinge clutches are broken. I ordered a new used set on eBay about a week ago, $10. (New ones were $60: I don't think so!) Today I installed them. It's a PITA, imagine that, made harder by my not knowing exactly what to do. Here's the short story: Remove keyboard and battery, CPU, Airport card. Pop off the hinge cover, the thing with the power button in it. Remove the inverter board. Remove the four hinge screws behind the folding I/O panel. Unplug and lift off the screen, making sure all wires are free to come out.

Now that the screen is free, pry out/off the four rubber bumpers and two screw covers, all on the face. Remove the screws thereby revealed. Carefully pry the back cover away from the front cover, there are a lot of snap clips. With that off you can see the exposed hinges. Remove the four screws holding the hinges to the front cover. Lift up the screen slightly and remove the four screws that hold the hinge brackets to the screen. Snap off the black plastic wire retaining cap on the one hinge, the one that is easy to replace. Replace that hinge and reverse its de-installation instructions. The other hinge is similar, but more involved. You have to remove the two screws on the screen bracket to release the Airport antenna, and you have to peel loose the tape that's holding the wrapped flexible circuit (that connects to the screen) to the hinge. Get that all loose so that the hinge bracket is released, then remove the bad hinge and replace it with the good one. Do this hinge last so that you best remember how to re-install the flex circuit and the wiring. Pay attention to where all the wires route so that nothing gets pinched or in the wrong place during reassembly.

Now reverse the above instructions to reassemble the computer.

In my case I had a lot of false starts, and ended up taking out many more screws than were necessary, but it worked out anyhow. The computer screen is now nice and tight. I wish Jill would quit slamming it shut! Of the four Pismos (and one Wallstreet) in our fleet this is the only one that has (so far) broken hinges. It's also the only one that gets slammed shut. Coincidence? I think not!

Sunday, December 11, 2011

Tore into the wretched Canon MF4270 copier/printer, again. Since I had decided last time that there was nothing overtly wrong with it I had wondered if its paper passage flag wasn't sufficiently opaque to properly actuate its optical sensor. (Its movement at the proper times is what syncs printing to the paper travel, and determines whether there's a paper jam, yet it seemed fine, and false paper jams are the problem.) I've seen opacity problems like this before, though they're rare, and it was certainly preferable to an actual electronics problem. So I took off the side and back panels (which was a pain) and wrapped aluminum foil around the flag, then put it back together. Preliminary indications are good, I was able to make several test copies and copy some sheet music out of a book, all without incident. Fixed? We shall see. I'm going to call it a non-romantic anniversary present for Jill.

Wednesday, January 4, 2012

I bought a used Apple 30" Cinema Display. $500 on CL. It had been offered originally a few weeks ago at $800, but I guess nobody was biting. The PO had two of them, but didn't really need that kind of capacity anymore. My intent (at the moment) is to replace Jill's creaking 800 MHz G4 iMac, a hand-me-down from my Dad, with a monster G5 system that will be better suited to doing her graphics work, yet still run the software we already have. Such a system shouldn't cost too much to put together, the monitor (which will work on newer machines as well) is the expensive part. Daniel's games work fairly well on the G4, so it will still get used.

Jill says that she'd rather just have a new(-ish) laptop, but I think that the lure of a big screen might be hard to resist for her work. If she won't use it, I will!

Monday, January 30, 2012

A G5 finally turned up on CL, $200. Only a 1.8 GHz dual, with 2 GB of RAM and a 320 GB disk, but it should do for now. When I went to see it the thing wouldn't boot, he'd been using it to download and it'd gotten stuck on a USB stick and he'd shut it down improperly, and the B-Tree Catalog had gotten corrupted. A session with the boot CD put it right, though. He says it was an ex-SCC unit, and as such had been (by law!) required to have no hard disk in it, which was a real pain for a refurbisher like himself. Hard disks are in relatively short supply right now due to a flood in Asia, it doesn't help that there's a bunch of State employees with hammers out there 'decommissioning' older computers' hard drives.

This unit has the slower RAM bus of the G5 line's two, my plan is to use this for now while searching for an affordable faster quad, then keep this as backup.

Tuesday, January 31, 2012

The G5 works, and drives the 30" Cinema display. It also drives the orphan ADC-connector Studio display we've had kicking around for awhile, making for a nice dual-monitor setup. The Studio display has a sleep/power button on the front that's handy. The G5's video card is insufficient to drive the Cinema at full resolution, I need to find a better one but it'll do for now. I want an inexpensive one that has a dual-channel DVI-I plug and an ADC plug; gaming performance is irrelevant.

Wednesday, February 8, 2012

After much researching and hair-pulling I came to the conclusion that the cheapest/easiest way to get a video card that'll drive the 30" display properly, and upgrade the memory and hard disk at the same time, was to just get the right G5 quad! I bought one on eBay: $350, shipped. It has 4GB RAM, 500GB of disk, an upgrade (over the stock 6600) 7800 video card, but not its special power cord. Assuming it really works as advertised, and isn't water-damaged, it'll become the primary and the first G5 will be the backup, as the quads are markedly faster than the older dual. (Faster busses and RAM, and of course 2× the CPU's.) The Studio display won't go on the thing without an adapter, but we don't really need two monitors if the big one is working to spec., and we can live without the snooze button on the monitor; I can save the Studio for the other G5. I just need to be very careful to keep an eye on the water-cooling situation. I'm thinking that if the thing is on its side any leaks will fall away from the CPU's, motherboard, and power supply. No harm, no foul? Maybe they won't cool right in that position, though, and you couldn't use the Superdrive. At the very least some kind of drip shield ought to be possible, or maybe just some absorbent material wadded into the right place.

Wednesday, February 15, 2012

It came today. Didn't really look like it was (or had been) all leaky so I powered it once it had warmed up. Works, but is definitely louder than the other one, and uses more power. It needs a special 20 A power cord, one I was able to borrow from work. This cord has a sideways blade, though, which none of the outlets in the house has. I'd bought an outlet to mate to it, and used a dead switching power supply from work to supply the cord and a plastic housing for an interim adapter cable. The G5 quad as currently configured draws more power, yes, but only about 350 W, nothing that a regular cord can't supply. (And that's with the 30" monitor.) It draws about 500 W when crunching images with all four cores.

It had nothing on the hard disk at all, so I did a fresh install of 10.4, using the firewire cable to upload everything (automatically) from the G5 dual. It took a few hours, but in the end we had two identically-configured computers. The quad is driving the 30" monitor perfectly, and the upgrade 7800GT video card will support Apple's Aperture program, should we wish to use it.

This thing doesn't have Bluetooth either, and its adapter is combined with the Airport card, which was only used in a couple of models and is relatively expensive for gear of this vintage. It's Apple's MA252G/A, selling (if you can find it) for $94.95 or so with all the necessary parts. (A.k.a. MA252FE/A? That might be the upgrade kit.) Another option might be a D-link DBT-120 dongle. (Which I already have, but have misplaced.) Bluetooth for the G5 dual is a lot less expensive, but you need both the card and an antenna, the quad has the antennae built into the case.

Tuesday, June 26, 2012

Jill was asking me if the downstairs TV could do streaming Netflix, probably for treadmill duty, and I said that the answer was a definite maybe, and that any of the home computers ought to be able to do it. Well, they can't. Not new enough, not any of 'em, though it's purely an artificial cutoff. (Don't get me started!) There was no way I was going to replace that TV with a newer one, not so soon and certainly not for that reason, but there are any number of devices (including a new[er] computer) that could do the job. Computers are slow to start up, and an incessant maintenance headache, so that's my last choice if one we already have won't do the job, nor do they have nice handy remote controls. Roku pisses me off, since they require a credit card number in order to enable your device, so that's out even though it's probably the most natural solution. The gaming systems are all somewhat unattractive to me, as I dislike games in general and I do not wish to encourage my son's gaming tendencies, which leaves Blu-ray players. I did a bit of research, and Panasonic has some that will do the job, and would integrate well with the TV. I checked online and it looked like their DMP-BD65 (or newer) should do the job. After work I stopped by our local overpriced electronics thrift shop and found that they had a DMP-BD655, which had an ethernet jack and claimed to do Netflix, reasonably priced at $35 and with a 30-day warrantee, so I bought it even though it was rather dented and scratched. (Plus $10 for the remote. Grr!) Now assuming that all this works and it's no more trouble to install than the (older) BD35 upstairs was then this should work out great. Upstairs the BD35 doesn't do Netflix, but the TV does. Downstairs the TV doesn't do it, but the BD655 should. Both systems integrate well enough that a single stock remote is all you need, normally. The only real problem is that the downstairs TV's HDMI jacks are completely obscured by my homemade TV wall mount. I may, however, be able to take it partly down and fit a right-angle connector in there. I'll test it all out first! I'll also need to run an ethernet cable over there, but it can follow the home theater cable routes already there.

Assuming this all works as I envision it the Blu-ray's remote will be all Jill needs, at the TV side of the room and using the TV's own puny speakers. For louder sound the home theater system would have to be fired up, and set to source sound from the TV. The down side of this configuration is that Blu-ray or Netflix would not be usable on the projector, but we can forego that for now. It is likely that this player will only be used for streaming (or for regular DVD's) while Jill's treading. In a pinch we could move it across the room temporarily for streaming and/or Blu-rays on the projector. These things are potentially cheap enough that I could just get another one.

In the evening I took the new BD655 upstairs and connected it in place of the BD35 player, and moved the network connection from the TV to the BD655. It fired up OK and hit the network for a firmware update, which went from version 1.23 to 1.60! Unlike the BD35 the update went smoothly in one attempt (no need to make an updater disc), and after the upgrade everything seemed to work. We were able to register the device with Netflix, and we watched a show on it. No problems at all, though I couldn't find a way to turn off the BD655's annoying key ping. I restored the normal configuration and we watched another show back on the TV's Netflix service, just to make sure nothing had gotten confused by the process. The BD655 is cleared for installation downstairs.

Wednesday, June 27, 2012

Yesterday when I was at the thrift store I noticed another Canon multi-function machine, so I checked it out online today. The MF4350d is the successor to our wretched MF4270, it was generally well-reviewed. It uses the same toner cartridge (#104), and offers duplex printing. It also has a superior paper storage drawer. I bought it, $40 with a 30-day warrantee, and it came with the disk, power cable, and manual. There was no toner cartridge, but we already had one of those. I brought it home, set it up, and it seemed to work. The document feeder paper support wing was broken off, but the MF4270's fit nicely. Let's see if this copier holds up better than the other one did. (I think our MF4270 was just a lemon, it was well-reviewed and on the days when it worked we liked it just fine.) If the MF4350d does work well I will consider the situation salvaged, since we'd have a functioning copier, duplex even, and I thus wouldn't have wasted my money on that overpriced toner cartridge I bought for it. (Which was more [!] than this replacement copier. The original lemon was the cheapest part of this, I believe it was $25 at Goodwill. Aftermarket toner cartridges are as cheap as $12 each online, which is what I should have done in the first place.)

I also bought another piece of angle iron at the hardware store, I think the best long-term solution to the basement TV's HDMI access problem is to modify the mounting bracket to expose the jacks. I should just need to weld a dogleg bridge over the connector bay, then cut away the original bracket in that area.

I also ordered another BR50 battery for my RAZR phone from an online vendor, the phone's battery life has been degrading noticeably lately, I'm sure it's time for a replacement.

Thursday, June 28, 2012

I ordered a 10' section of rubber cable channel on Amazon (for the free shipping). About $37 for Electriduct's D-2, in black.

While I was on Amazon I checked again for the D-link DBT-120 Bluetooth USB dongle. (One of which I already have, but have misplaced. Accept no substitutes, this is the only one that works without driver games on our older PPC Macs.) They had a used one for about $20 (shipped). I ordered it, since it can be left semi-permanently in the G5 quad, besides using it on my Pismo to transfer photos etc. from my RAZR phone. (Which I need to do before I can ditch the pink phone for the gray one I picked up awhile ago.) I had been resisting buying another DBT-120, but this has been blocking me long enough. The dongle is a lot cheaper than the official stuff for the G5, when you can find one.

Monday, July 2, 2012

The DBT-120 came, and I used it to offload the photos from the pink RAZR phone and put the ring tones onto the replacement gray RAZR. Finally. Call for Super Chicken! Buk-awwwk!

Monday, July 9, 2012

The rubber cable channel came today and I installed it in the home theater room. I ran the L/R speaker wires down the corresponding unslit channels, they pushed in easily at first but got rather difficult by the end. Eventually I prevailed, it was obvious to me that I was going to need all three channels in order to get all the cables under cover. The main channel took the rest uneventfully, I may still have room for a network cable. It looks much better than it did, and the cables are now protected. The 10' length was generous, and easily bridges between the two couches.

Thursday, July 12, 2012

I made the bridge piece for the downstairs TV's wall-mount bracket. I cut and bent the angle iron so that it had 45° angles in a /===\ shape, long enough to span the connector panel on the TV, and welded the cuts closed. I ground the beads off the side that is to contact the TV itself, but the other side I left pretty rough. I made a few mistakes that meant that I essentially started over on one end, fortunately the exact length of the final part was not critical. This all took long enough that I did not go further in this session, but it should be ready to weld to the bracket.

Friday, July 13, 2012

I welded the bridge onto the bracket and cut away a section of bracket to expose the two HDMI connectors. It worked out pretty well. I ground off a lot of the ugliness and wire-brushed the rest. Getting the bracket on and off the TV was a pain, but there was no mishap. The one HDMI cable I have fit nicely into the sockets now, but protrudes too far and runs into the wall-mount side of the bracket. I'll still need a right-angle HDMI cable, I think.

When it was connected, though, the two remotes were able to control each others' devices, so that is good. A couple of cables and I guess we're done.

Saturday, July 14, 2012

I snaked a length of network cable from the network switch, through the wiring closet, and to the downstairs TV. I ran it through the rubber channel, then I crimped ends on it. (The crimping tool is a POS and gave me a lot of trouble, I ended up having to push the IDC fangs in with a tiny screwdriver. I need a better tool, this consumer-grade stuff sucks.) It works, we watched a show on it.

Tuesday, August 7, 2012

The G5 quad's nVidia 7800GT video card has a very noisy little fan, which has been quite an irritant. Small sizes do not for quiet make, and as this system doesn't have anything else in the PCI card cage there's lots of room for a bigger, quieter fan. I raided the junkbox and found some old PC fans, DC brushless, 12 V, and tried a couple out. All were noisy on 12 V, but one of them was nice and quiet on about 5 V, and would reliably start spinning at that voltage. It moved plenty of air, and drew about 125 mA at that voltage, and had a plastic shroud that held the fan slightly away from the business end. I removed the heat sink/fan assembly from the nVidia, and removed the fan and plastic shroud from the heat sink proper. I blew out the dust bunnies, which were substantial, and put back the heat sink. I couldn't find a good local source of 5 V, so I got a 68 Ω power resistor from the junkbox to use in series with 12 V; I cut the old fan's connector off and soldered together the assembly and taped up the bare wires. The assembly draws about 100 mA from the 12 V supply. When I reassembled the computer I laid the new fan over the heat sink, angled so that it tries to blow through the heat sink's channels. It's only held in place by gravity but that should be no problem. it's not like these things get moved around in operation. I tucked the old fan and shroud into the case for safekeeping. The computer came back up OK, and was much less noisy; the shrill little whine was gone, we will see if the new fan keeps the nVidia happy. The operation is reversible, if need be.

Friday, August 10, 2012

Yesterday the two HDMI right-angle adapters I'd ordered came, and I installed them in the downstairs TV. One of them interferes with the antenna cable, so it probably couldn't be used, but a shell-less connector without the adapter might work in that slot if we ever were to need it. (The adapters were cheap-cheap-cheap, no loss there.) The other adapter fits fine, and the Blu-ray player plugged right in and worked. As of now the Blu-ray/Netflix project for downstairs has reached completion, so far as the initial plan went anyway. (And the Blu-ray's annoying key ping can be turned off down in one of the Viera cast screens, it's another network 'application'.)

Tuesday, December 18, 2012

Not electronics, per se, but I set the G5 dual (G5D) up more permanently upstairs since we have plans to use it there, for financial record keeping and for convenient speed-surfing. (At least until we get some faster laptops.) It's been on a table in front of the window for months now, which was less than attractive. Recently I procured an old $10 thrift-shop desk, very small, and solid pine, though it needed a bit of gluing. It was just the right size to hold the flat-screen monitor, keyboard, and mouse, plus three drawers for storage. As there is no wired network in that part of the house, and the G5D doesn't have wireless, I'm using one of the Pismos as a wireless gateway; it works just fine. Anyway, none of the chairs we had would fit into the kneehole, so I hit the thrift shops today and eventually found a $5 rickety one that would fit. (It had to be narrower than 17" in width, which turned out to be harder than I would have thought.) I started gluing the legs back securely.

Wednesday, December 19, 2012

Finished gluing the chair. It's sturdy enough, now, but still creaky. That just makes it more interesting!

Sunday, January 20, 2013

The Panasonic TC-P50G25 plasma TV made a nasty loud 'POP' today and shut down completely. No resurrection, no blinking error code, etc. I took the back off and probed around, and found one dead 2A fuse on the power supply (806ASH/NPX806MS1), and a shorted diode in one of the bridges, one that appeared to run the always-on supply. Nothing else obvious. I gambled and replaced these parts, whereupon it went 'POP' again and blew the fuse. Oh well, it was worth a try. A replacement board seems to be $80 or up. Not that bad, if it fixes it.

Monday, January 21, 2013

I ordered, via ShopJimmy, a replacement NPX806MS1Y, Item #sj-ETX2MM806ASH, w/180-day warrantee. $87.16, free shipping. Order #1399003.

Wednesday, January 23, 2013

The power supply board came today, it's V7 versus the original V3, so perhaps it's a more durable design. I installed it, and it worked perfectly, so I reassembled the TV and put back the furniture. Success!

Tuesday, January 29, 2013

The replacement BT-446 batteries for the cordless 'phones have not been holding up. Just over a year and two of them seem to have failed already. After the second one I decided to look into it. The symptom was that it said it had charge, but could not pick up after a ring or turn on the speakerphone, the 'phone would blank out and seem to reset. Oddly, you could place calls from it just fine, and talk for quite awhile.


I peeled open one of the packs. The 'charged' battery had a good voltage, but probably had high internal resistance. I rigged a 10-ohm resistor across one of the cells to discharge it. It took more than 12 hours to drop to a half a volt or so. All cells exhibited this behavior. Plenty of capacity, but no peak handling ability. The cells would bounce back substantially when left alone. I then charged the pack some on the bench, to get it out of the hole before putting it back in the 'phone for testing and formal charging. At this point there was no improvement in behavior. As an experiment I soldered a 22µF tantalum capacitor from the junkbox across the pack. (This being small enough to fit into the battery compartment with the battery, and a type known to have low internal resistance.) Success! The 'phone could now act normally, with the battery providing the average current and the capacitor fielding the peaks. We shall see how long this helps.

These replacements are obviously crap batteries, ones that have a high (and rapidly worsening) internal resistance, which is a known failure mode of NiMh batteries. Not too sure what to do about it, price is no sure indication of quality when dealing with with the current Asian market. For now, capacitors.

Friday, March 29, 2013

We'd been looking for awhile, sporadically, for a newer laptop for Jill, and today it finally happened. A 2007-ish Macbook Pro, 15", Intel Core 2 Duo, 2.2GHz, 2GB RAM. Complete with its 10.5 OS install discs, carrying case, etc., for $400. Looks good. It was a bit recalcitrant when it came to copying information over from her Pismo, but eventually I prevailed. (I think it was the Pismo's fault.) It seems to be able to run all of our existing software, including Adobe CS3, but might not if Jill upgrades the OS.

The battery was totally comatose, I ordered a new one for $40. The computer had sat in storage for a couple of years before the PO's heirs sold it. The signal strength on the wifi is lower than the Pismo, but I understand that is normal due the metal case.

The lowered asking price was due to a non-functioning DVD drive, but I opened up the computer and extracted a stuck (and gummy) teach-yourself-Spanish CD. After that it worked fine, at least while all opened up. Later, case fully reassembled, I couldn't get it to keep a disc in.

Saturday, June 29, 2013

I picked up the CLD-D703 laserdisc player at the shop. It had been languishing there for several years, he said it was fine and I said it still had an intermittent problem. Time to pull the plug on the issue, an intermittent player is better than no player.

Saturday, December 21, 2013

We'd been wanting a non-inkjet color printer for some time, and finally did something about it. I found a Xerox Phaser 8560MFP multifunction printer on Craigslist, I got it for $240. It was owned by a church in Kennewick (they got it from a church in Hermiston), but it was surplus to their needs, and the pastor was making a trip to Spokane so it all worked out. I plugged it in and a few clicks of the mouse later it was connected, and working well. It's fast! The "spray-on crayon" print technology actually looks pretty good. Takes quite awhile to warm up at first, which is not too surprising, and I've been told not to turn it on and off a lot, which wastes ink in a cleaning/purge cycle. So it's in a room that we keep heated with electric heat, so its waste heat should just make the heater cycling on a bit less.

Saturday, January 11, 2014

A power outage let the Xerox 8560 come to full cool, so I took the opportunity to dump the waste wax tray. Not that much in it, maybe a little over 1/8" or so. All black, no sign of color segregation but that may have been due to mixing while hot. Now I know that it's empty we can keep an eye on the wastage from here on out. The tray was latched in place, I had to hook a finger in through the maintenance tray's hole to release it.

Thursday, February 13, 2014

I bought some used ink for the Xerox 8560 off of Craigslist. $40 for 2 cyan, 3 yellow, 2 magenta, and 2 black blocks. These look like they were dug back out of the hopper of their defunct printer, they're all loose and not in their original packaging, and there is some surface cross-contamination of color and some deformation. I'll have to scrape them off a bit before using them. Not a very attractive buy, really, but these little blocks are around $40 each, new retail.

Friday, February 14, 2014

Yesterday I salvaged some RAM sticks from some Dell PC's they're scrapping at work, and I put them in the G5 Quad. 4 1GB sticks filled out the machine's memory slots, and doubled the RAM to 8GB. (Two in the topmost two slots, and two in the bottommost two slots.) One of the sticks is one I had removed from my work machine some time ago when it got unreliable, which seemed to help, but I think the fault was the machine's and not the memory's. (The machine eventually died the rest of the way, hence its scrapping.) We'll see if the G5Q has any problems with this stuff. It seemed to boot faster with twice the memory.

The same scrap raid netted an HP 2100TN printer, with extra paper tray. (Both full.) If nothing else we just got some toner and paper for the other HP printer.

Tuesday, March 18, 2014

I salvaged two 2G RAM sticks from some Dell PC's they're scrapping at work, and I put them in the G5 Quad. That ups the RAM to 10GB. They report a mix of speeds, the new ones all say -288, the original 4GB all report as -444. (I'm told that sometimes these lie.)

Tuesday, March 25, 2014

Bought a full G5 quad system off of Craigslist, $250. Complete with 8GB RAM, 2 250GB hard disks, 10.5 installed and with a retail install DVD, iLife '06, and a buncha other software. With 20" Cinema display, keyboard, and mouse. To serve as a backup system for Jill, besides being usable by me. (Its 6600 video card probably will not drive the 30" Cinema display at full resolution.)

Saturday, April 26, 2014

They were auctioning off some of the older equipment at work (proceeds to benefit company parties, etc.), and I picked up an HP 8150DN printer for $25—with paper, cart, extra toner cartridge, cables... I'd always wanted an 11×17" printer, but was never prepared to pay much for one. When I got it home I found that it wouldn't connect on the network, and it turns out that many JetDirect cards are quite prone to sudden death like that. (This has a 610N.) Taking a cue from the internet I put it in the oven at 350°F for 8 minutes, and that fixed it right up. Unfortunately it completely melted the plastic handle too! That was a mess, but I'll take a mess that works over something nice-looking that doesn't, any day. (Next time use a torch, or remove that handle before baking.) For initial testing I did use the 610 card from the HP 2100, with that I was able to prove that the printer worked and was worth messing with.

Once communications were established I blew out and wiped out the insides as best I could, there was a lot of dust. (The printer has nearly 300,000 pages through it.) This included washing off the selenium drum in the toner cartridge with alcohol and a soft cloth. That really helped clean things up, before that the print quality was dreadful: smudgy and poor.

Wednesday, June 4, 2014

We were given a first-generation G4 years ago, and never had a use for it. (This is a 400 MHz AGP "speed-dumped" M7824LL/A, codename: Sawtooth. It was made in late 1999, and retailed for $2500.) I loaded it up as a Mac demo unit for some curious PC-head neighbors, but they never got around to even powering it up. I reclaimed it when they moved, because it is actually fairly low in power consumption. It draws 36W when quiescent, one disk drive. (42W with both drives powered on. In contrast, our borrowed Buffalo 500GB backup server draws about 15W when quiescent.) If I can get it rigged with two big SATA mirrored drives and 10.5, then it could be our Time Machine backup destination, something that the Buffalo just can't do. Some sources indicate that 10.5 can work on these systems, but a different video card is required. Apparently what Leopard needs is "Core Graphics", whatever that is? The System Profiler says that the stock Rage128 card does not have it. One suggestion is an ATI Radeon 7000 or 7500 AGP.

After much thrashing around, with target disk mode attempts (both ways!) and various boot DVD's, etc., the light finally went on. This stock Sawtooth G4 can install Leopard, but only if the G4's firmware is fully up to date! To do that you have to install MACOS 9.1+ and download Apple's 4.2.8 firmware updater. I did this, on the empty drive, and the firmware was indeed out of date. Once I upgraded that I was, after using LeopardAssist to lie about the CPU's clock rate, able to boot the install DVD and proceed normally, putting it onto the no-longer-empty drive. The new firmware also fixes up the target disk mode and lets it boot from Firewire.

Also of note, Apple's built-in Screen Sharing application works fine to control the G4 from the G5's. (I didn't need to scare up a compatible Vine server and fight Chicken to gain access, like I did with the G3, just turn on Screen Sharing on the G4.) I gave the G4 a fixed IP address, suitable for its duty as a household file server, and made a Dock-able script so that it can be administered after only one click on a client machine. I pieced together the following shell commands from Google-mining to create this 'application':

mkdir -m 0755 -p $P
cat >$P/$NAME <<-"EOF"
	open $0.vncloc
	exit 0
cat >$P/$NAME.vncloc <<-EOF
	<plist version=1.0>
chmod +x $P/$NAME $P/$NAME.vncloc
(Replace "" above with the IP address of the server.) After executing this you can click on the resulting G4Server application, and/or drag it to the Dock. Instant window to the server.

Saturday, June 7, 2014

The G4Server spent most of its day upgrading itself over the network. It's amazing how many trips you have to take through Software Update before it all is finished and it has nothing more to suggest. 10.5 is a dog on this relatively slow machine, but that will be immaterial for its intended use; as a file server it should be just fine. (Upgrading from 768MB to 1GB really helped, supposedly the machine can take up to 2GB, and to good effect, but I won't spend any money to get there.)

Next up is to get a big SATA RAID hooked up. Ideally I could force a cheap PC PCI/SATA controller to work rather than buying one of the few remaining Mac-compatible cards, and which tend to be a bit on the expensive side. Possible links of interest:

It strikes me that yesterday's script (above) is overly complex, given that the open command can do URL's directly. We shouldn't need the plist file at all. Take 2:
mkdir -m 0755 -p $P
cat >$P/$NAME <<-EOF
	open vnc://$ADDR
	exit 0
chmod +x $P/$NAME
This seems to work equally well.

Monday, June 9, 2014

They were scrapping some laptops at work, and I nabbed a few 2GB SODIMM DDR2 sticks of RAM. I put a couple in the Macbook Pro; it seems to like them, it's reporting 4GB of memory and seems snappier.

Saturday, June 14, 2014

One of the ideas for a replacement web service is to host it myself on the G4Server, but to do that you need a back-link from outside. Hurdle #1 is the fact that standard residential ISP products don't give you a fixed IP address. There is a product/protocol called Dynamic DNS that can be used for this, but it is predicated upon the address you are given not changing too often. A coworker has a script on his site that can tell you what your address (of the moment) is. I thought I'd start collecting some data to see just how stable is my transient address.

Today's lucky winner:

I don't need a fancy domain name to call my own, nor do I need 24×7 availability or world-wide connectivity or high bandwidth. All I want is static site contents, reasonably available at modest bandwidth, for me and a few fellow enthusiasts that may wish to consult this information. Oh, and I do want standard search engines to have access to and to be able to index the site. (Else the content will never be found, as there is no custom domain name.)

One strategy to do this is to set up a link on any friendly site, one that is already indexed and crawl-able, that points back down to G4Server. If my transient IP address is in fact fairly stable, this link could be maintained by hand. (Or rather, by a friendly-site CGI script triggered periodically from G4Server, either automatically or by hand, that would check and rewrite the link as necessary.) With my firewall opened up to incoming HTTP connections, redirected to G4Server, that ought to do it.

So, the steps in this zero-cost strategy are:

  1. Identify a friendly site, one that is willing to host the simple link-(re)writing script and that doesn't mind being a redirecting target for my crap. Site (and redirection) needs to be indexed and crawl-able by standard search engines.
  2. Write the link-writing CGI script. Needs to check for some kind of query key intended to ensure that the update request is from my site, and not from some would-be pirate.
  3. Induce G4Server to make the query periodically. Daily cron task? If the transient IP address is in fact fairly stable, the probe time would be the maximum amount of time the site would be unavailable in the event that the transient address did change.
  4. Put a redirection rule on our router, one that will take incoming HTTP requests and vector them to G4Server. In the worst case this will require a new router, especially since I can't get in to the Mikrotik anymore.
  5. Ensure that G4Server's Apache webserver cannot upload at all. Other general jail-ish settings, etc.
  6. Implement a bandwidth limiter on G4Server's Apache webserver. This mostly to prevent normal use, or even potential abuse, from interfering with our normal internet service use, and to discourage such abuse attempts in the first place because slow-ish sites just aren't very interesting to crackers and vandals.

Sunday, June 15, 2014

The Web server on G4Server is easy to turn on in Leopard, it's just a click away in the control panel. At that point, though, it's only giving you access to links that are in the various 'home' directories, their Sites folders. The secret to getting those indexed is to change the /Library/Webserver/Documents/index.html file, give that a new top-level index page that points to the various home-directory sites you want to expose.

Wednesday, June 18, 2014

Our ISP came up with the default password they had placed on our Mikrotik router when they configured it for us in April of 2012. So I'm in, and have poked a hole in the firewall for gaining access to our web server from outside.

Thursday, June 19, 2014

Requested Google to index our site, by address, now that it's sort-of available again. They did!

Saturday, June 21, 2014

Yesterday I grabbed a 20GB hard drive and a likely-looking video card from a scrapped PC at work that was of a similar vintage to the G4. The HD turned out to be exactly the same model that was in there already, but the video card (NVIDIA MX4000 based) turned out to be too old for the Mac. While testing it, though, I did find that the system will come up without any video card at all, which has interesting implications for a low-power installation. You can still get in and operate it via the Screen Sharing application, so having no video card at all is no real hardship except that Time Machine's GUI can't be used, and that's no change given the (in-)abilities of its stock Rage 128 video card. I reconfigured the machine for only one hard disk and removed the video card, keyboard, and mouse, and after it settled down and got quiet I measured the power draw at a svelte 30W. Even less, presumably, if/when the hard disk spins down. This will go up, of course, once the big RAID is installed. (In contrast, in the idle laptop department a G3 Pismo draws about 15W, and the Intel C2D MBP about 20W, so the G4 isn't doing too badly here.)

Here are the GPU's that are supposed to exhibit full functionality with Leopard:

Of course, few of these are probably suitable for this machine. I'm thinking that the NVIDIA 6200 is one of the more likely candidates.

Tuesday, June 24, 2014

I have created a free account on staticcling.org, formicapeak.staticcling.org, to handle our Dynamic DNS needs for now. With this our site is now findable through a domain name, rather than just via a Google content search.

The holder of our preferred domain name, formicapeak.com, who created it for us as part of the sale price of Jill's SL in 2007, but for which no content ever showed up, consented to release the domain name to us. (He was still holding it, probably inadvertently.) We now own it, but will of course have to pay now to keep it hereafter. It's not hooked up to anything yet, though.

Wednesday, June 25, 2014

The cheap SATA card arrived today, it was the last piece to arrive. I installed it and the drives in the G4, and removed the modem card while I was there to perhaps save a little power. The drives don't want to fit easily where I had planned, on the floor of the case next to the two IDE drives, because the SATA cables protrude a bit more than IDE cables do. There was interference when both drives were stacked, but side-by-side things were able to clear and the case can be closed up.

The card had a sticker "SATA RAID" over the flash chip. I peeled it off and found it to be an AMD 29F010-120. (1 Megabit, aka 128 KB.)

Unfortunately the 3112 flashing utility does not recognize the card. I knew there was a pretty good risk of this when I started. Do I need to edit the AppleVIAATA.kext file? Is there even such a file on a G4? The PCI Device ID is 0x1095:0x3112/0x1095:0x6112.

Here is a very interesting discussion: https://68kmla.org/forums/index.php?/topic/13557-sil3112-flashing-sata It tells me to retrieve Wiebetech's utility from a web archive, now that the parent company has killed that kind of support. The wayback link is: http://web.archive.org/web/20081205192504/http://www.wiebetech.com/download.php/id=120 Also interesting from this thread is the fact that the firmware file has a few bytes at the end, which the 3112 chip reads into its PCI registers.

Anyway, when I ran the utility it rejected the card, it's the 0x6112 (vs another 0x3112) in the PCI ID that seems to be the problem.

I made a copy of the Wiebetech .kext driver, and used emacs to change everything 3112 in it to 6112. The kextload utility liked it, and the System Profiler says that the card then had a driver, but nothing else of note happened, and there certainly weren't any more drives available.

Friday, June 27, 2014

Just in case I brought home a crappy PC from the scrap pile at work. It's of a vintage to have a PCI bus, USB peripherals, a VGA monitor plug, and a floppy drive. (We'll just ignore the game port and the serial peripherals.) Just what the doctor ordered, for using a DOS-based flash utility. It's got a 700 MHz Celeron processor in it, not snappy, but probably adequate. (No hard disk at the moment, but it can boot from the CD using its Sony CDU-5211 CD-ROM drive.) The BIOS says Intel UNDI PXE-2.0, and the on-screen setup utility is quite primitive.

Saturday, June 28, 2014

With the RAID card installed, at boot time the PC said "Silicon Image SATA RAID 4.2.47" at one point. That's a reasonably good sign!

I burned a copy of Wary Puppy 5.5 Linux onto a CD and booted the nasty PC with it. It wasn't too happy with only the 128 MB of RAM in it, so I put in another stick to bring it up to 256 MB, that seemed to help a lot.

The best/simplest info I have on converting this to a Mac card is here: http://forums.macrumors.com/showthread.php?t=1690231. Those are the instructions I followed, and it just worked. (The logistics of actually doing this, given what I have to work with, was a mess. One of the most tricky bits was actually finding a blank floppy disk that still would work.) The short answer was that I used Virtual PC to build a floppy disk image that contained both DOS and the flash updater, and I then used the Puppy to dd this image from a USB stick onto a real floppy. Once I had that it was a simple matter of booting this floppy and flashing the card, which was itself pretty uneventful.

I put the flashed card back into the G4, and it worked, and found the two new 2T drives, which I then put into a Mirrored RAID.

With the video card removed, the unit seems to settle down to 39W quiescent power draw. Not too bad for three hard disks, plus the computer. The two extra disks and the controller card only added 9W of power draw. (The internal modem was also removed.)

I think now, though, that I was very wrong when I thought this machine would be a good file server. I discovered the problem when I set the first G5 to try to back up to it. It's slow. Deadly slow, and the G4 goes almost comatose while it's taking the data. It stops serving web pages and you can't even administer it. The Time Machine throughput is only around 100 MB/hour. If you put the sender to sleep the G4 eventually recovers and acts normally, but at this rate the initial backup, of just one of the four intended clients, will take about a month. (It's around 88 GB.) The network traffic pattern is bursty, comb teeth in the Activity Monitor, with big chunks of idle time about as long as the toothy burst time. I believe the G4 is CPU-bound, because of what its Activity Monitor shows. (It's hard to be sure, the machine is so unresponsive. You get a bit of screen update every minute or so.) There's no way that this strategy is going to work.

Plan B? Yank the big terabyte drives and put one in each of the two main G5's and unleash their Time Machines to back up the entire system? (That's not what we're doing at the moment, their second drives are too small.) There's no backup redundancy that way, though. I think I'm going to have to conduct some experiments, perhaps put one of the drives in the third G5 and see how well it functions as a backup server. I wonder how low I could get the power draw on one of those. Maybe start with an air-cooled dual and remove one of the CPU's? (Were there singles?) There's a lot less room in a G5 case than this G4, and the power supply situation is less friendly. To get the necessary minimum of three drives in one would require a SATA controller card, and I don't think the one I just worked up will be suitable. Sigh...

Sunday, June 29, 2014

A new day, a new start... I stopped the never-would-finish backup. I tried some experiments with dd to make some random-contents files, and the old ATA drives absolutely kicked the cookies out of the SATA RAID. That's weird. I took the two 2 TG disks out of RAID and started the G5's backup again, and it started backing up vastly faster than before, at about 4–5 MB/second sustained, and the G4 remained usable during this. CPU utilization, though high, is not pegged. The machine is acting more like I expected, just not with mirroring. Potential causes:
  1. This SATA card sucks;
  2. 10.5's Mirroring driver sucks;
  3. 10.5's Mirroring driver is designed for a faster CPU, and trips over its feet on such a slow machine;
  4. 10.5's Mirroring driver reacts badly with this SATA card when trying to use both drives 'at once';
  5. And, for all of the above, substitute in the SIL3112 driver that is burned into the flash on this SATA card;

I could try narrowing this down a bit by purchasing a second SATA card to see if putting one drive on each card helped, and also seeing if I could find a different on-card SIL3112 driver. That'd narrow down some of the options, as would finding and trying a faster G4 machine. Not too attractive, any of these.

Still, the purpose of the exercise was to get redundant Time Machine backups of all of our other machines. Mirroring was merely the most obvious way to gain redundancy. Nearly as good would be to back up everything using TM to one of the 2T drives, and use the G4's own TM backup software to back itself (including the first 2T drive) up to the other 2T drive. A little more confusing and potentially amenable to disruption, and with the potential loss of any backup set that hadn't yet been relayed to the second drive, but probably pretty acceptable in the end.

With the RAID broken apart I tried the G5 backup again, and it seems to be able to sustain about 7 MB/second over the long haul, which if not stellar is at least reasonable.

I experimented a bit, and if you're trying to use both SATA drives at once the machine chokes. (Copying files from one to the other, etc.) I'm guessing it's not the mirroring function in the driver, anyway. Also, using the G4's TM to copy between the two 2T drives is a pipe dream, because on a networked drive, TM doesn't keep things as separate files, but rather in a sparsebundle container. Looks like the dream of RAID on the Time Machine backup store is not to be at this time. I removed one of the 2 2T drives and put it aside for now. With only the two disk drives in it the unit seems to settle down to 32W quiescent power draw.

Sunday, July 13, 2014

The new SATA card came and I flashed it with the same firmware as the other, and installed it, one 2TB drive on each card. Using the Finder to copy, though the system was indeed sluggish I was able to copy big files from one drive to the other. I rebooted with both drives on one card, and it was again essentially unusable as it copied. (30+ hours estimated to copy a 47GB backup file, vs 40 minutes when using both driver cards.)

I tried the 733 MHz CPU card I bought cheap, but it doesn't fit at all. The card is larger and runs into the second IDE connector, and the existing heat sink interferes with some power supply toroids on the new card. Not a chance, at least it was cheap! (It was for a Quicksilver, not a Sawtooth, and apparently that makes a difference.) The original card bears an MPC7400 CPU, the proposed replacement has a 7450. After a bunch more digging, it appears that you can put a Quicksilver CPU into a Sawtooth, but you have to remove the second IDE connector from the motherboard, and run a +12V line to the fourth mounting screw hole of the CPU card. Messy! (The extra power line is no problem, but I don't really want to lose the optical drive.) Apparently only Sawtooth and Gigabit Ethernet CPU's (and, of course, the appropriate aftermarket upgrades) fit that space without modification. Also, the more detailed reading suggests that the 733 MHz CPU will only be running at 551 MHz (733×100/133) because of the slower memory bus of the Sawtooth. (Presumably it could be up-clocked back to near normality.) During the operation it became clear that the existing heat sink would not be of any use on a dual CPU card (assuming that one would even work in this particular Sawtooth), it only has one heat riser boss. Before I get too carried away I must remind myself that the purpose of a faster CPU was to make mirroring usable, and if a second SATA card is going to do the job then I don't need to pursue this anyway.

I fooled around with OSX's "diskutil" program for awhile, based on some online information, and finally got it to turn the two 2T drives into a RAID1 without reformatting. The GUI utility was able to trigger the necessary rebuild, estimated at 27 hours to complete. That's slower than I'd like, of course, but for the moment acceptable. (The offsite backup strategy is to periodically swap out the mirror, so this rebuild time will be a factor in that policy. It is clear that the mirroring is done at the driver level, rather than in the filesystem as in DNIX, because it appears to be copying the entire 2TB between drives rather than only the [at the moment] relatively minuscule amount that is actively in use.) The updating process appears to be CPU-bound, so the faster CPU may yet enter the picture.

Monday, July 14, 2014

I found a copy of the UniNorthASICChecker utility that reports the version of the CPU support ASIC, and this G4 Sawtooth is not capable of supporting a dual CPU. (Version 3, not version 7.) Oh well.

I looked at a couple of PPC data sheets, and the power draw goes up pretty rapidly with clock speed when the CPU is fully busy. I'll want to think carefully about whether or not to move up in the CPU department on this one.

PPC Power

Of course, the significant value here would be the quiescent power draw, not the peak load, and that's not so easy to determine from a data sheet.

Yesterday I took apart my G5's "Mighty Mouse" to try to clean the roller ball on top so that it would work again. (It wouldn't scroll up, but was OK in all other directions.) I had quite a time of it, and ended up breaking off the lower trim ring permanently. Eventually I got the roller disassembled, and removed an amazing amount of gunk. During reassembly I lost one of the magnetic rollers for awhile, but eventually found it in the carpet. After all that it was a lot cleaner, but didn't really work any better. I did some surfing, and it was recommended that you spray cleaner into the ball and then vigorously roll the ball against a sheet of paper on the desk. I did this today, using Deoxit Fader spray, and it worked!

I also have decided that I really don't like the flat chiclet keyboard that came with my G5, and I tried a spare keyboard that I bought some time ago at the thrift shop. It was horribly dirty, and the keys were all intermittent. The lucite case was gross, and full of choads. I took it apart and put it in the dishwasher, and it really cleaned up nicely. I blew dry what little circuitry it had, then disassembled the keyswitch array to dry it out too. It has three printed-circuit mylar sheets underneath the silicone rubber dome array that trapped a lot of moisture between them. I spread it all apart and mopped it, and left it to dry overnight.

Wednesday, July 16, 2014

Now that the G4's RAID is up and running I moved its Time Machine backup destination to the RAID, joining all the other backups. Unfortunately it appears there's no good way to move the two months of existing backups from the external 1T drive to the RAID, but I hope that this isn't going to be significant. (I had tried a Finder copy, but it got stuck for two days and wasn't making any progress. Apparently directory hard links confound even it! Supposedly 10.6 can handle doing this, but that OS can't run on this machine. I could have done it if there was nothing already on the RAID, via a Disk Utility volume copy/restore, but that is not the case here.) The only thing I've lost access to is the system's pre-webserver state, and that is in fact still available on the external drive, so long as I don't delete the old backup data set.

I reassembled the G5's new (?) keyboard. It mostly seemed to work, but eight keys are inoperative: space bar, caps lock, escape, backtick, right control, left option, tab, right arrow. Not good! I put it away for now.

Thursday, July 17, 2014

There are some network switching products we've discontinued at work, and the prototype development units are being scrapped. (Ciena 3911, née WWP LE-58H.) I brought one home, as it's got 8 10/100/1000 copper Ethernet ports, 2 copper/fiber ports, and doesn't use huge amounts of power (only about 10W) or a fan (unlike just about everything else we've ever made, or plan to make in the future). Unfortunately they do not wake up as unmanaged switches, you must configure them before use. The customary way is via its console serial port, and I had a Radio Shack 26-183 USB serial dongle of my own that I had used at work for that very purpose. It was only supported by XP, which we don't use there anymore, but it appears that it uses a chipset (PL2303) for which there is a Mac driver. Unfortunately I was unable to find a good driver for our G4/G5 machines, but the actual guts of the one open-source driver have not really changed over time. I grabbed the sources for the most recent version, for 10.9, and opened up in 10.5 Xcode. It wasn't entirely happy opening a 'new' project in an 'old' builder, but after several hours of twisting knobs on the thing I finally got it to compile, and eventually to load without error. (No code changes were required, it was all infrastructure settings.) I installed it on the G4, and used the "screen" program to drive it. Success!
screen /dev/tty.PL2303-091
I was able to log into the 3911 without incident. Now I need to look up how to configure it to be an unmanaged switch.

Friday, July 18, 2014

I spent this morning nailing the 3911 to the wall and deploying it. Turns out it already wakes up mostly as an unmanaged switch. You only need one command, and that's only to allow the two fiber/copper uplink ports to participate as normal ports:
port set port 9-10 acceptable-frame-type all
config save
I rearranged the area and removed the old hub, and put the 3911 in place of the little NetGear. The NetGear was still required to pick up a couple of the (currently unused) ports. I checked, and the G5 is able to connect at 1Gbps. Unfortunately we have nothing else that can talk at 1Gbps, so that's not doing much good. It would be nice if the G4 file server could do so, but it's slow enough that I doubt it would do any good even if it could use 1Gbps.

Sunday, July 20, 2014

The G4's Rage 128 video card seems to have died today. Too much in-and-out of the system activity? I've been keeping it in a anti-static bag... Oh well, they're not highly valued anyway. I just have to be very careful not to disrupt any settings that will keep the G4 from being administered over the network. I bid on and won a SPAG44 Nvidia 6200-based card, about $11. With any luck it can be flashed for use in Macs, and will allow use of the Time Machine GUI.

Monday, July 21, 2014

I got the 3911 to be manageable via telnet. It's a slightly longer list of commands:
port set port 9-10 acceptable-frame-type all
port set port 1-8 pvid 127
vlan add vlan 127 port 1-8
dhcp client disable
interface remote set ip subnet
interface set gateway
config save
The main secret is that the Default VLAN is not good enough for the remote interface, so switching everything over to the Mgmt VLAN is the easiest way to allow the remote interface to be configured. The remote interface must be working before we can upgrade the firmware on the 3911.

To upgrade the firmware on the 3911 we have to get TFTP working somewhere accessible. On G4Server I copied the firmware to /private/tftpboot, and then used the shareware TftpServer utility to turn on the server. (It can be done by hand, but I gather that it's a bit of a pain.) The upgrade itself is straightforward:

software upgrade server package saos-6.10.2 package-path saos-6.10.2 service-disruption allow
and then wait. It takes quite awhile, in fact, and service is down for a fair number of minutes. 6.10.2 is the final version of software for this particular product. (Codename: Artamir. It's based on the Broadcom Raven switch chip, and uses Linux as an OS on the 300 MHz MIPS CPU that is embedded within the Raven. The unit is designed for extended temperature range operation.)

I then installed Wireshark on G4Server, I want to try to figure out why the linkstation is no longer accessible. Sniffing its MAC address (dug out of the 3911's flow mac show report) shows not much going on, and it was still using its default address. As an experiment I moved the Mikrotik and the linkstation to the unmanaged switch, and after that it was again accessible as a storage volume on the originally-assigned address, even after I moved things back to their normal ports. After a power-cycle on the linkstation, it again disappeared, presumably moving back to its default IP address.

So it appears that the 3911 is somehow interfering with the initial discovery process, possibly with DHCP.

Tuesday, July 22, 2014

I picked up a spare G4 Sawtooth off of Craigslist today, $20. Since we're now counting on using one of these for an Important Purpose, it makes sense to have a spare on hand. It was extremely dusty, I felt that with a little bit of watering maybe you could grow tomatoes in it! A little session with the air compressor outside took care of that. It came with MacOS 9.2.2 on it, and a Rage 128 Pro video card (with siezed fan). I removed the fan and cleaned it, and oiled the bearing. After that it worked correctly, and wasn't even all that loud. The MacOS 9.2.2 installation was pretty wonky, so I ended up doing a fresh install from CD. After that I could get it on the network, and I checked the board's boot firmare. It was already at 4.2.8, so it's up to date. The unit came with a 450 MHz CPU, 512 MB of RAM, a 20 GB hard disk, and the Rage 128 Pro, so it's moderately superior to the original machine. I tried the 'bad' Rage 128 card in it, and it worked, so I put it back into G4Server and it worked there again too. WTF? Oh well, we're back in business again, and I essentially have a full Sawtooth spare, modulo the RAM and hard disks (which would transfer in a pinch). I don't need a hot spare, having sufficient parts around to piece together a functioning system with an hour or two of effort is sufficient. This spare is really a power supply/motherboard/CPU source. I'm debating whether or not it is worth swapping things around so that the 400 MHz CPU is the spare. Seems like a lot of work for only a 12% CPU speed increase.

Wednesday, July 23, 2014

By teaching the Mikrotik that is also a local address the linkstation can now be pinged on its default address without the probe escaping outside of our home network. (Probably I should make the router treat the entire /16 as local-only but I hesitate to tinker too much with the thing as recovery from mistakes could be tricky.)
ip address add address= interface=bridge
I also enabled PostFix on G4Server and aimed the Xerox Phaser printer's email notifications directly at it instead of to one of our public addresses.

Thursday, July 24, 2014

The 6200 video board came yesterday. I used the Imation SuperDisk drive I have to make a floppy containing the nvflash.exe utility and a Mac firmware image that might work. This, in conjunction with the boot floppy I made for the SATA card flashing project, might do the trick. I'm taking it all to work in an effort to find an AGP-equipped PC with a floppy drive on the scrap pile. (This was much easier than what I did to make the boot floppy. I wish I'd remembered that I had this SuperDrive then!)

Friday, July 25, 2014

I brought home a scrap Micron PC from work that had a floppy and an AGP port, and got it running. (I had to put in some RAM.) Unfortunately I was unable to find a Mac firmware that nvflash could load onto the video card. Looks like there's only one make of 6200 card out there that could work, and this ain't one of 'em. Unhappy!

My problems with floppies have been greatly eased by first bulk-erasing them, then formatting them in the destination machine. Most, if not all, of the disk errors were eliminated by this. I theorize that the newer drives have very narrow heads when compared to the step pitch, so leftover out-of-track magnetic fields on the medium wreaks havoc. Bulk erasure gets rid of all that.

Tuesday, July 29, 2014

After much fooling around with the utilities and such I was able to make a ROM image that could be flashed to the card. Thereafter it disappeared from the PC's radar screen, and when placed in the spare G4 it booted into OS 9.2 and had usable video! I put it in the G4Server, and everything looked OK until the blue screen came up. That ended up turning grainy, and no windows showed up. The mouse was still OK, but that was it. The machine was not usable on the network, either. It appears that the memory and timing settings that the card uses post-boot are bad. Not really a surprise, I guess. Things just got a lot harder, since it won't work as a video card in a PC now. Unless there are MacOS flashing utilities I am going to have to run the PC blind. The down-clocked firmware version (for cooler running in a G4 cube) might work better, if I could get it loaded onto the card.

Wednesday, July 30, 2014

I tried to use the Tektronix 834 RS-232 protocol analyzer to watch the COM1 line on the PC, I figured I could redirect the output of nvflash to it to see what it was unhappy about. It was DOA, though, I think its 34-year-old EPROM faded/failed.

Thursday, July 31, 2014

I was able to borrow a PCI video card and watch the AGP flashing process directly. The problem is that the nvflash utility is crashing! That's not good...

The business with the serial port probably never would have worked, I don't think the crash reports can come out there.

Saturday, August 2, 2014

The flasher kept crashing, and an attempt to get an older one wasn't particularly successful, I had a lot of floppy problems. Since this boot floppy claims to support a CD I tested this, and it wasn't lying. I spent a bit of time gathering up every older version of nvflash I could find and made a CD of all of them along with the ROM firmware. After that it was uneventful. I was able to burn the down-clocked firmware onto the card using version 5.40, the latest 'old' flasher I had. (The problematic one was 5.63.)

After that it still booted into OS 9.2.2, but attempts to boot into OSX crashed with a panic screen this time. Still no joy!

Tuesday, August 5, 2014

It took a bunch of thrashing around, and eventual use of target disk mode and an all-day disk copy, but I eventually got a copy of OSX on a spare drive for use in the spare machine, so I can do my flashing testing without disrupting the main server.

Wednesday, September 10, 2014

The 8150DN printer has been putting out very mottled prints lately, but only really noticeable on photographs, so I swapped in the 'new' toner cartridge, but that didn't really do it. (It helped, though.) I then pulled the transfer roller and cleaned it thoroughly, with a bit of dish soap and with alcohol/ammonia window cleaner. There, much better. (You aren't supposed to do that, but it sure looks like it helped.)

Wednesday, September 17, 2014

We've had to put a cyan and a yellow ink block in the 8560 color printer. I finally got a clean cold power-up into an empty ink waste tray, and it dumps about 12g of ink on a cold start. As each block (of four) weighs only 25g, that's a fairly hefty percentage! They weren't kidding about this printer not being particularly suitable for intermittent home use.

Saturday, September 20, 2014

Recently fucking-Ptera unilaterally decided that any inward-directed traffic is 'business', and worthy of an extra $10/month in extortion fees. They turned on the firewall/router in their radio, which had been operating perfectly since installation in a bridging mode, using our router for firewall protection, and we were completely cut off. No warning, no advance notice, no discussion... just no more web site. I think an extra $120 a year to flip a switch back to where it had been before is pretty fucking unreasonable, and I am thoroughly pissed.

A kind soul has offered me a berth on his domain and server, and I am exploring the possibilities. An SSH socket relay looks like the least disruptive path right now:

ssh -R 8448: name@address -nNTx
Seems to be the basic magic incantation, executed on my server. This initiates a connection from inside my network, which means it will be allowed by fucking-Ptera, telling the more-publicly-available remote server to forward all port 8448 traffic to my server's (web) port 80. Now all I have to do is convince the remote server to forward all formicapeak subdomain traffic to its port 8448, and we're all done except for changing the formicapeak.com A-record! ... And it looks like mine host has already set up the Apache stuff. The heart of it is:
$ cat /etc/apache2/sites-available/formicapeak.com
<VirtualHost *:80>
        ServerAdmin webmaster@localhost
        ServerName formicapeak.com:80

        ProxyPreserveHost	On
	ProxyPass /
        ProxyPassReverse /
One little A-record redirect at the formicapeak.com host and it looks like we're back online! Thank you very much, Tim, and fuck you very much, Ptera.

Wednesday, October 1, 2014

I got a public key set up so that I can script the SSH relay, and created the auto-run relay script. Here is its essence:
while :; do ssh -R 8448: name@address -nNTx sleep 600 done

Sunday, December 7, 2014

Apple offers OSX 10.10 (Yosemite) free for all takers, so I upgraded our Macbook Pro with it. That took some doing, but it seems happy enough now. The new appearance and features do not please us much, but it does mean we're running completely up-to-date software: even if we bought a brand-new computer we wouldn't get anything better. (Maybe somewhat faster on the newer hardware, but that would be it.) No more too-old messages from dumb-ass web sites, at least for awhile. Of course, we're still stuck with the nasty keyboard of the MBP, and the newest OS did nothing for its wildly intermittent WiFi behavior. I'm sure it would be happier with more than 4GB of memory, too. Still, we've pushed back the inevitable a little bit longer.

Friday, December 26, 2014

The Xerox 8560 ran out of magenta ink and was whining, Jill needed to print our Christmas newsletter. All four ink buckets were low, I loaded it up with every spare ink block we had. (Three new actual Xerox magentas, the rest were all used blocks I had bought, and which probably are not authentic.) All we have left are two used magenta blocks, the next time we run out we'll be stuck. Printing is fairly streaky right now, that's a bit concerning.

Saturday, January 10, 2015

Bought another Panasonic DMP-BD35 Blu-ray player at Goodwill, $35. (A little much for a unit with no remote, if you ask me, but it's the same as one of our others, and can share its remote.) I brought it home and it seemed to work fine. It was able to upgrade its own firmware over the network just fine, from v1.3 to v2.4 (the latest/last from October 2011). This is for use on the projection system. This unit was barely able to cable into that hodgepodge of a system, and only because it had an optical digital audio output jack. There was one optical input left on the non-HDMI home theater receiver, and the projector had one as-yet-unused HDMI jack. So, in use you have to switch the projector to HDMI and away from the HT system, and tell the HT system to feed from the SA-CD input. (And then double-check the audio format decoding settings.) Messy, but it works. Sadly, this player doesn't have integrated Netflix et al., so we can't do that on the projector unless we move equipment around. Still, it's an improvement. It can use the same remote as the one on the TV on the other side of the room, so that lack was not a problem. I haven't yet decided whether or not to decommission the Sony DVD player in favor of this or not.

Saturday, January 17, 2015

Jill's been complaining for awhile now that her G5 Quad Mac has been roaring its fans at her. We've also experienced a few crashes that could possibly be thermally induced. So today I bit the bullet.

Googling for instructions I found how to remove the CPU assembly from the computer. This is not easy, I didn't have the correct long Allen screwdrivers. I managed to make do with some shorter ones, a long Torx screwdriver, some SAE Allen bits, and a 1/4" socket wrench set. That was a pain. Once I had the assembly out I looked it over and blew out the dust bunnies. It looked pretty good, no sign of leaks. It's got a second-generation Delphi/Cooligy cooling system. Not ideal, but so far so good.

That was the hard part. Turning the CPU unit over exposes 15 Allen screws to remove the heat sink from a CPU card. The paste between the two looked thin and dry, I wiped it off and replaced it with new. Repeat ×2 for both CPU's. Reassembly was in the reverse order, the most fiddly bit was getting the CPU assembly seated to the motherboard. Once assembled far enough to function I powered it on (but with the disks disconnected) long enough to hear the 'bong' that indicated functional CPU's, then I powered off and finished the reassembly. I replaced it in Jill's office and fired it up.

It came up just fine, and the fans did ramp up a bit while working harder during the boot, but then the got quiet again. That's new! Time will tell whether or not this is fixed, but so far I believe it to have been a, ahem, 'roaring' success!

Yes, I was able to save and reuse the anti-tamper pin that locks the cosmetic panel in place, and yes, I did drop a screw down into the works and had to do the dance with the machine held over my head, complete with appropriate incantations!

Sunday, May 10, 2015

Friday I hit a yard sale, and bought a Kindle Touch w/G3, for $10. With cable, charger, leatherette folder, and zippered Cordura case. After charging it and upgrading the SW (in three steps) to the latest, I find that its internal web browser seems to work OK, but only on wi-fi. Unfortunately it doesn't much like the open wifi in the house. I tried deploying a surplus D-Link DI-524 as a secondary access point, for it and for other devices that don't seem all that happy with the old AP.

Friday, May 22, 2015

The monitor on the troll has crapped out, and has a very squashed and distorted vertical display. (This was the spare monitor, already. They are custom, for an industrial computer manufactured by a company (ISC) I used to work for. Not readily replaceable by anything purchase-able, though a standard monitor could be made to work with some cabling effort.) I had already checked the non-polar electrolytic capacitor, and tested the big heat-sunk diode on the curve tracer, with no joy. Research suggested that the (common) TDA1170N vertical deflection IC is a bit fragile, so I bought two spares. I installed one this morning, with zero difference! Nuts. I really need to spend some quality time with http://www.repairfaq.org/sam/monfaq.htm.

Wednesday, August 12, 2015

Over the last week I've been trying to upgrade the HD on the backup/web server G4, the 20GB drive is awfully full and is starting to make intermittent bearing noise. Also, the drive sometimes fills up with logs (?) and chokes, requiring a reboot. I'd recently bagged some cast-off IDE (PATA) drives from work, and elsewhere, and so I had some materials to work with. (Formerly all I had was some 20GB drives, which weren't any bigger than what was in use.)

Research showed that the effective drive limit for this system was 128GB or so, due to the 48-bit LBA business. While there is a commercial driver for the G4, it wasn't worth the investment. Anything larger than 20GB will do, really.

So, rummaging found an 80GB WD800 that looked nice enough, and had a lower power rating than the bigger drives. I put it on, and WTF? It only showed around 30GB. While 30GB is in fact big enough to make a substantial difference, I wanted all 80GB. I did some cursory research, and found mention of the Host Protected Area (HPA). Huh, I guess this old drive had been used with an HPA in its former life. Thus began a round of trying to find out how to turn this off, culminating in my resurrecting one of the PC's I'd use to flash SATA and Video cards for G4Server, and getting Linux working on it again. After downloading both the C development kit (for Wary Puppy 5.5) and the hdparm sources I finally prevailed, only to have hdparm tell me that the HPA was not in use. Crap, there goes a day... Linux also told me that there was a fixed 32GB partition on the 80G drive that I could not enlarge. After even more flailing around and research, I finally found out that this particular drive (the WD800) has a lovely 'feature' where if you have both a CS and a Master/Slave jumper it'll automatically cap the size to 32GB! Crap! So I removed the CS jumper and the drive started working perfectly, both on Linux and on OSX.

Next I was able to format it in the Mac, and then, after a bunch of flailing around getting it booted from an optical disc (the burnable copy I have doesn't really work on its DVD-ROM drive, but the scratched pressed master I have, which won't complete an Install, worked fine to do disk repairs and restores) I was able to do a full system restore from the Time Machine backups.

So now it boots! Somehow, though, the RAID array managed to lose sync, so it wants a days-long repair. That'll wait. What I found was that while it seemed to come up OK the web server wasn't working. After a lot of flailing around I finally figured out, from examining the Console log and some surfing, that Apache wasn't starting up because it didn't have a log directory. This fixed it:

	# sudo mkdir -m 0755 /private/var/log/apache2
After that we were back in business. All that remains is to manually transfer the bulky stuff that's not part of the Time Machine backup, get the RAID and Time Machine back online, and button things up. The system claims, at this time, that it'll take about 29 hours to bring the RAID back up.

Thursday, August 13, 2015

It was only about 24 hours for the RAID to rebuild. I turned on Time Machine again and called it good. We should be back to normal, but minus the periodic disk space episodes.

Friday, August 14, 2015

The POS 12V power supplies that came with WWP's (now Ciena) LE-58H (CN3911) don't last very long. The one I'm using as a main household switch died, so I nabbed four more from work. (Get 'em while you can, the 3911 has been discontinued! This is the fourth such power supply that's died on me, here and at work.) Anyway, I installed the replacement and restored the network wiring to normal.

Saturday, August 15, 2015

In recent months printing has gotten pretty difficult around here, jobs don't complete, get hung, etc. Very frustrating. I did some Googling and found one mention that DNS server misconfiguration could cause these problems. So I queued up a big bitmap print job from Jill's G5 to the HP 8150DN, and it couldn't finish. I then unplugged the link to the Mikrotik router, leaving our home with no addressing services or Internet whatsoever, merely coasting on whatever IP addresses or content was cached, and then it printed fine. One data point, at least.

Wednesday, October 28, 2015

I ordered a third WD Green 2TB drive to complete our backup strategy. The plan is to periodically (quarterly?) swap out one of the drives in the G4's RAID and put the extra one in secure off-site storage. The RAID will rebuild the newly-added drive, which takes about a day, and the up-to-date removed one will be offsite immediately. So long as the RAID rebuild goes smoothly, this should work. $85 for the last (of the three) drives.

Wednesday, November 4, 2015

The drive came yesterday, and today I installed it. Dragging the drive onto the RAID set added it, the system claims about a day to restore the RAID to full operation. I put the old drive into an anti-static bag, put that and a card identifying it into an old plastic first-aid kit box, one with a rubber lid seal, put that into a large sealed Baggie, and put the whole mess into off-site fire storage. (Not humidity controlled, hence the bagging.)

Saturday, November 14, 2015

I find that our 2007 MacBook Pro can actually utilize 6 GB of RAM (but not 8) in its two slots, which is a rather odd size. Still, it's 50% better that what we have now so I ordered a 4GB stick for it from OWC: $117. (Spendy, but that's what this older style costs in this size.) Also, everybody states that a SSD would be a very dramatic improvement, the single best thing you can do for a machine, so I ordered an inexpensive one of these from OWC. 120GB for $65. This is essentially the same size that is in it now, but so far we haven't really been space-limited. If this all works, that should end the complaints about the slowness of this machine. For now.

Saturday, November 21, 2015

The care package from OWC came. Installing the RAM was trivial, and it just worked. 6GB, and I think the machine is snappier.

Sunday, November 22, 2015

The hard disk is another proposition. Many, many little screws to remove to get it swapped. Before beginning I re-downloaded the Yosemite installer, and put it on a 8GB CF, using the 'createinstallmedia' script to make it bootable:
sudo /Applications/Install\ OS\ X\ Yosemite.app/Contents/Resources/createinstallmedia --volume /Volumes/myvolume --applicationpath /Applications/Install\ OS\ X\ Yosemite.app
Yes, the MBP can boot from it, using the Lexar Firewire CF reader. I got the drives swapped, and tried an install. The installer is unable to access the network for a Time Machine Restore, so I went with a fresh install for now. But, it didn't work. The process failed, and after that the SSD disk had disappeared from the system. A reset didn't bring it back, only a power-cycle would do so, and it's not got anything usable on it. Total waste of time. I did this more than once, it was perfectly repeatable.

Monday, November 23, 2015

Total frustration. The Yosemite install freezes every time. I tried booting the CF from a USB reader instead, so that I could free up the Firewire connection, in a feeble attempt to use Target Disk Mode on the Time Machine server. Much slower, but no difference. (And pointless, you can't use TM that way.)

Tuesday, November 24, 2015

I found some instructions on getting the networked Time Machine database accessible for a restore, and it worked well. From the Terminal:
mkdir /Volumes/Backups
mount -t afp afp://adminname:password@serveripaddress/Backups /Volumes/Backups
hdid /Volumes/Backups/nameofyoursparsebundle
After that the Installer could see it. It estimates 3+ hours to do a full restore. I am skeptical, the problem was always at the end of the procedure, but we shall see.

...and it worked! Sort of. The thing got stuck at the end, and the log showed some kind of disk queue timeout during the 'bless' operation. But I rebooted it and it came up on the SSD anyway. It looked completely normal at that point.

Next was to restore the 10.6 partition, which worked about the same. I had to jigger the partition sizes a bit to get the TM Restore to start, but that worked too. Afterwords the machine could boot up either OS, and it is a lot faster in operation. With the additional RAM we can again have all three user accounts active at once, and work normally on any one of them. The machine seems entirely satisfactory again.

Friday, January 15, 2016

I bought 2 Apple TV's, a V4 and a V3, plus HDMI cables at Best Buy. $260. The V4 is fancier, but the V3 (for the downstairs TV) might be something we like better. We'll see!

Saturday, February 6, 2016

My brother, a die-hard PC guy (probably due to cost) finally got fed up with all the Windows BS after the umpteenth time his PC got all choked up and virusy, and was considering a Mac. (Perhaps the fact that I'm using more-than-decade-old HW and SW, with no slowdown or virus problems was a factor?) After discussions with me regarding his budget and needs he finally found a very good deal on Craigslist: a stock four-core first-generation Mac Pro, with a nice 23" cinema display, keyboard, and mouse. ($150!) To meet his long-term needs it's going to have to run El Capitan, which it can do if you stuff enough RAM in it and the video card is upgraded properly. This whole process is... tricky. Apple doesn't want you to do this on quite such old hardware, but it can be done and by all accounts works very well.

Anyway, it seemed like such a good idea that I bought one too. (Jill's G5 quad is adequate, but the whole browser situation is not that good, and getting worse by the year.) I bought it on eBay a week ago, $300, and it came today. Perhaps less of a good deal, perhaps not. This one is a second-generation eight-core unit, complete with 16GB of RAM and an upgraded ATI HD 4870 video card with a noisy fan. The idea is that I can pre-flight all the necessary work for my brother's computer on this one, and just deliver to him a new SSD with El Capitan on it and a video card all ready to go. (I've already upgraded his RAM and put a 2TB Time Machine backup drive in it.) Both machines are running 10.7.5 Lion, the last OS that is officially supported by the first two generation Mac Pros. (Everything subsequent can run the latest OS without any fuss.)

I bought a wifi card for it, and it works fine. (If it's properly installed, the two antennae are very fiddly to get connected right.) The very first thing I did was to drill a hole in the video card's fan hub and drip some oil on the bearings. That helped a lot! The hideous shrieking and groaning is all gone. If this video card is good enough for El Capitan I'll replace the fan. If not, I won't waste my time and money on it.

Saturday, February 13, 2016

Step 1 of the Mac Pro upgrade was to scare up a scratch hard drive upon which to experiment. I had a 320GB SATA unit in the drawer that looked like it would work. I partitioned it into two 120GB partitions, El Capitan and Yosemite, and a scratch drive for the remainder. The idea is to preflight the SSD size I intend to buy. The ultimate plan is to use a smaller, less expensive SSD for the OS and applications, and keep the original Lion home directories on the larger HD. Unless you're doing A/V editing/mastering, user files are rarely the slow part of using a computer these days; the effective speed of this hybrid arrangement should be nearly that of an SSD alone, at least as we'll use it. That also preserves a completely usable Lion system, for fallback and any legacy software needs.

Step 2 was to use target disk mode and the laptop, our only official El-Capitan-capable machine, to actually install both Yosemite and El Capitan on these partitions. That was a giant pain, it's a slow process to begin with, and there were a lot of false starts and crashes. Eventually I prevailed, the main problem was that the 320GB drive was bad! (This was not easy to determine, at first, but it eventually got enough worse that it became clear that the disk was junk.) A friend supplied a 250GB SATA unit that he didn't want back. Once I switched to that everything went smoothly. (I didn't want to buy an HD for this temporary use, and I didn't want to buy the SSD's if this wasn't going to work out.) I ran Software Update repeatedly on both OS's until both installations were completely up-to-date.

Step 3 was to replace the two boot.efi files (per partition) with the magic ones that let the old Mac Pro's work. (Kudos to 'Tiamo' et al.) This is actually quite easy, running from a root shell in Lion. El Capitan then booted and worked fine, Yosemite never did. (We don't need Yosemite, it was just a backup plan in case El Capitan didn't work out.)

At this point we have a fully-usable early Mac Pro running El Capitan. The experiment is a success.

Saturday, February 20, 2016

The next step of the Mac Pro upgrade was to suck Jill's G5 dry using Migration Assistant. I did this from Lion, not El Capitan. Mysteriously, the HD 4870 was unable to drive the 30" monitor at all. I used a different, smaller monitor for this lengthy process. Very disturbing, I would have thought that the 30" would work. I hope this isn't another sad story like we had with the G5 quad, trying to find a video card that would both work in the system and drive the big monitor.

From El Capitan I then used Migration Assistant to copy up the applications and system settings from Lion, but not the user accounts. I then created user accounts manually, and used the option-click to change the UID's and home directory paths to point to the Lion HD. Thus we have two different OS's that use the same exact user accounts and data. You can option-boot into either one, your choice. It all worked smoothly.

I got to wondering if the missing power cable (one of two) for the video card was significant? The system nags about it every time it boots, and there are odd video anomalies at times. I couldn't find a cable in town, or at work, but then I remembered that Jill's G5 quad had one of these cables in it. I borrowed it, and found that the HD4870 then worked perfectly. The 30" monitor works, you can log in multiple accounts, the DVD player works... And Unigine's Valley demo program will knock your socks off. Why the hell was this system, which came with this video card, missing this cable? I ordered one, immediately.

Friday, February 25, 2016

The $5 extra PCIe power cable I needed for the video card came, and when I installed it everything now works perfectly. All systems are Go!

Saturday, February 27, 2016

I was at my brother's, and I brought my HD and the video card to try out. I put them in his system, and everything worked great. No problems expected with the plan at this point. His stock 7300 video card is indeed too old for El Capitan, but it was able to function—just no 'video' applications. It acted almost exactly like the 4870 did when it had only one power cable.

The next step is to find a second 4870 for my brother. Since I have an offical Apple unit, we should be able to make do with a cheap PC unit and some flashing utilities to clone the firmware.

Tuesday, March 8, 2016

The $10 replacement fan for the ATI HD 4870 came yesterday, and I installed it today. It's easy to do, but there are a lot of screws. I needed to refresh the heat sink grease while it was apart.

It works well, and is quiet. I did need to use one washer to keep a screw from dragging on the fan, its mounting base is a little thinner than the original.

Wednesday, March 9, 2016

Our work office is moving, and there's a lot of cleanup. I scored an NVIDIA Quadro NVS 295 video card, it has two Display Port connectors, and no fan or exra power cable requirements. I'm not sure if it'll work at all, but it was free. I ordered a cheap DP-to-HDMI cable from Walmart, on clearance. We'll see if it can drive the TV with it. (Just curious, no real need.) If you put it in Slot 4 you can still get x8 PCI connectivity for it, the only x16 lane (Slot 1) is in use by the main video card.

Sunday, March 13, 2016

I found these ultra-simple instructions for flashing PC HD 4870 cards using only the target Mac, which I have paraphrased here. You need a different, non-ATI video card to execute the procedure, or else the ability to ssh into, or use headless, the target Mac, so that nothing is modified except the target video card.
You'll need the iMac Graphics Update 1.0.2 and Pacifist.

First, use Pacifist to open the graphics update image. Extract two files: ATIROMFlasher.kext and ATIFacelessFlash.app.

After extracting them we make sure the kernel extension is able to load:

chown -R root:wheel ATIROMFlasher.kext
chmod -R 0644       ATIROMFlasher.kext
kextutil -nt        ATIROMFlasher.kext

Strip everything out of the firmware updater and replace it with the image intended for the 4870:

rm -f       ATIFacelessFlash.app/Contents/Resources/*IMG
cp 4870.ROM ATIFacelessFlash.app/Contents/Resources/

Finally, flash the card:

kextload ATIROMFlasher.kext
open ATIFacelessFlash.app
The app should appear in your Dock, wait for it to complete. Then:
kextunload ATIROMFlasher.kext

At the next reboot the 4870 should be working normally.

Wednesday, March 16, 2016

The cheap DP-to-HDMI cable showed up, and I hauled the whole mess into the living room to try it on the TV, feeding it from the free NVIDIA Quadro NVS 295 video card. With it cabled up it indeed does not have boot video capability, but once the OS is up all the way the system seems to work fine. I was able to have it play a DVD on the TV while the Valley demo was running on the other display. The system would also work with the 295 as the only video card. The Valley demo would actually run on it, but the frame rate was 1–2 FPS, which is dreadful. Other than that, though, everything seemed to work fine.

The other free DP-only video card, the ATI 3470, was non-functional.

Saturday, April 30, 2016

The flat chiclet keyboard on the Mac Pro was disgusting, so I tried to clean it with windex, carefully. (Holding upside down while wiping it with a dampened rag.) I guess I wasn't careful enough, the keyboard was non-functional afterwards. The M1243 keyboard is glued together, and cannot be disassembled without destroying it, so with nothing left to lose I submerged it in a sink of warm water and rinsed it out thoroughly, shook the water out and vacuumed it dry as best as I could, then placed it in the oven at 250°F for an hour to dry, then let it cool. Success! It works perfectly, and is clean. Maybe not the best way to go about it, but...

Wednesday, May 18, 2016

Happy volcano day to me! I'd been seeing this bedraggled Mac Pro on craigslist for awhile, offered at $75. Today I finally went downtown and got it. It's a stock 2.66 GHz quad Mac Pro 1,1 with a 7300 video card and 2GB of RAM and a sickly 1T Seagate Barracuda drive, running OSX 10.6 but it does have both the WiFi and the Bluetooth cards. The PO had managed to break off the lithium coin cell retaining clip, which I guess is pretty easy to do. So, it's not in perfect shape. It came with a keyboard and mouse. This was his spare system, I think he was trying to free up a little space and a little cash too.

Eminently upgradeable, and the broken battery holder doesn't matter if the thing is kept plugged in to a UPS.

Work had discarded a bunch of suitable ECC RAM that I grabbed, I pawed through what I had left (after upgrading the other Mac Pro) and gave it another 9GB of RAM, filling all the slots and bringing it up to 11GB. Should be plenty! Jill's will have 24GB, mine will have 11GB.

Thursday, May 19, 2016

Though both Mac Pros are slated to run 10.11 El Capitan, the last OS that is officially supported on them is 10.7 Lion and we'll need that as a fallback position. I ordered an official copy from Apple, which I intend to burn to DVD. I intend that both units have a main 10.7 Lion partition containing all user accounts, a small 10.6 Snow Leopard partition just to keep access to Rosetta (Power PC emulation), whenever that should turn out to be necessary, and an SSD that has 10.11 El Capitan on it which has all its regular user accounts aimed at the 10.7 installation. (10.6 will probably do the same thing.)

I hate Intel, but these 10-year-old machines are fast, capable, up-to-date, and affordable. Gotta go with it, I guess.

Saturday, May 21, 2016

It's time to get off the stick on these computer upgrades, I ordered a bunch of stuff for the two Mac Pros. Two 120GB SSD's, a replacement 1TB drive for #2, two "Big Jim" SLAED X5365 quad-core 3GHz CPU's for #2, some good silver heat sink paste, and two video card power cords. I still need to scare up another 4870 video card, and I really should upgrade #1's 500GB drive to a 1TB, so that both machines match. The first Mac Pro has been operating very well, fast and quiet, it's time to retire the G5 Quads (especially the flaky one) in favor of something a little newer. The browser situation is getting untenable.

Daniel will continue to use the G5 Dual, mostly because its Studio display only hooks to a video card that won't go into one of the Quads.

Sunday, May 22, 2016

I upgraded the firmware to 2,1 on the newest Mac Pro, it should be ready for a CPU upgrade now. I also downloaded my purchased (!) copy of hoary old Lion.

Tuesday, May 24, 2016

I found a 1GB HD 4870 on eBay, for $37.35 shipped. PC, not Apple, but you can re-flash them to add the EFI.

Friday, May 27, 2016

The parts have started to roll in, the 4870 video came today. I disassembled it and cleaned out the dust bunnnies, and re-applied heat-sink grease. It works! It still has PC firmware, so the boot process into El Capitan was blind, but other than that...

Details from the various labels on the card:

HD-487A-ZH HD-487A-ZHFC VA.5
S/N K7H045438
Advanced Micro Devices Model b507

The "ZHFC" card is one that has been known to work in a Mac.

Monday, May 30, 2016

The last few days I've been messing with trying to re-flash the 4870, with poor success. (Crashes, 'bong' cycles, video artifacts, etc.) The 'pipomolo' instructions were almost good enough, but what was really necesary was to hand-build a composite ROM image, preserving the original on-card BIOS section, so that all three required sections of ROM end up where they need to be. I used emacs' hexl-mode to rummage through the various images to see what needed to be where. (The Apple EFI image was just a bit too large to insert into the hole that was left for it, so the 'third' section of the ROM had to be moved down 512 bytes. That was the key.)

After that, the card seems to be working perfectly, boot screens, fans, and all. On OS 10.6, 10.7, and 10.11, boot CD's and DVD's, fake 'DOS', etc. Here is the script I made to build the composite ROM image:

# Start with a copy of the file taken from the card via atiflash:
cp -p HD48702.ROM pc4870.rom

# Extract the Apple EFI from the official 4870 card image:
dd if=HD4870.ROM of=efi.part bs=1 skip=63488 count=49152

# Copy the EFI into the new card image (this destroys second MCuC section):
blocknum=`printf %d "'\`dd if=pc4870.rom bs=1 skip=2 count=1 2>/dev/null\`"`
size=$(($blocknum * 512))
dd if=efi.part of=pc4870.rom bs=$size seek=1 conv=notrunc
rm efi.part

# Copy the second MCuC section from original image to its new home:
dd if=HD48702.ROM of=pc4870.rom conv=notrunc bs=1 iseek=112640 oseek=113152 count=17920
# HACK!  The second MCuC is now in a different place (0x1BA00 vs 0x1B800)
# so we have to patch the address near the first MCuC to match.  Easiest way
# is to just copy some random 0xBA I found over the now-incorrect 0xB8.  Ugh.
dd if=pc4870.rom of=pc4870.rom conv=notrunc bs=1 count=1 iseek=1467 oseek=48220

# Now fix up the headers, etc.
python fixrom.py pc4870.rom pc4870.rom

# This reporting tool should be happy with the result.
./efirom -d pc4870.rom

NB: This script cannot be used unmodified for your card! It was hand-tuned for my card.

At this point I used a DOS-ish boot CD to get to where I could use a standard ATI flash utility. I have temporarily mounted a FAT-32 IDE disk in the MP's second optical drive bay, 'C:', which holds the flashing utilities and the flash image made above. Tedious, but straightforward. Essentially you are booting into DOS and using a correct-vintage "atiflash" utility: atiflash -p 0 pc4870.rom.

I also used the Disk Utility to 'restore' the running El Capitan HD to the new SSD, and sucked over the user accounts and files from the main G5 quad, while running Lion. I then ran the upgrade to 10.11.5 on the SSD, re-patched the boot.efi files, and set up the 2-disk boot environment for El Capitan and its Admin account on the SSD, and another Admin account and the remaining user accounts on the 1T 10.7 disk. So it's dual-boot, 10.7.5 and 10.11.5, same user accounts on both. Seems to be working fine.

Actually, it's triple-boot. I made a 20GB partition on the 1T drive for 10.6, so that Rosetta would be available if ever needed.

Tuesday, May 31, 2016

The last package came, bringing with it the silver heat sink grease I need for the CPU upgrades, the power cables for the other 4870 card, and the other 1T drive. I put the cables and the drive into the first MacPro, for Jill's use.

Wednesday, June 1, 2016

I got the triple OS installed on the MacPro's new drives, and got the updates all applied. That was tedious.

Thursday, June 2, 2016

I deployed the MacPro in place of Jill's G5 quad. I configured the printers, and got it hooked to Time Machine. (The first backup took all night, roughly 200GB.)

Friday, June 3, 2016

I tackled the second MacPro and opened it up, removing all the removables and getting the CPU heatsinks out. Lots of dust bunnies! I didn't have the correct long 3mm Allen, but with the memory card cage removed I could get sufficient access with a shorter one.

Saturday, June 4, 2016

I removed the logic board from the case and removed the broken battery clip. The negative terminal is heavily tied to the ground plane, and it really didn't want to come out. The solder sucker opened up the positive hole just fine, but I just couldn't get enough heat into the negative hole to clear it. I ended up drilling out the solder with a pin-vise drill. The Radio Shack battery holder was a perfect fit, and soldered in without an excess of trouble. (That dratted negative terminal was a bit of an issue.) With a new battery clipped into place I blew the dust bunnies out of the case, and put the logic board back. Next step: CPU upgrades!

...After kayaking I finished up. The CPU's were easy to swap, a little bit of isopropyl alcohol and a paper towel removed the old heat-sink grease. The ZIF sockets were easy to manipulate, and the new CPU's dropped right in. I used new Arctic Silver heat sink grease, and smeared a thin layer on each CPU and its mating heat sink surface, then put a thin bead down the middle of each CPU and put on the sinks. I then tightened them down, and put the unit together far enough (sans disks) to hear the normal power-up 'bong'. I put it the rest of the way together, and found that it wouldn't boot. Blinking file folder time. Crap! I took it back apart, and found that the SATA data plug wasn't quite seated all the way to the logic board, so all the HD's were dead. After that was dealt with the unit worked perfectly, and identified itself as a MacPro 2,1 with 8 3GHz cores. Success!

It's time to retire the G5 quads. I prepped them for moving to storage. I stole the superdrive from the flakey one and put it in the second MP as its second optical drive. (The other one has a Region 2 DVD-ROM drive in there, which works but is not well integrated with the GUI.)

Tuesday, June 7, 2016

The troll has not been running recently, and it's very hot; the sprinkler system it controls is vital. The problem turns out to be the power supply, a then-standard off-the-shelf PC/AT power supply. Not nearly as reliable as the rest of the computer, which was designed to last the ages. It keeps throwing a power-failure state, and sometimes didn't have the voltages come up right. I swung by the recycling place, and they had exactly one such supply, still in the machine's carcass. $5, if I removed it. Done.

After work I plugged it in place, and the system seems fine now.

Thursday, June 16, 2016

The MacPro project has hit a snag, with Time Machine! Turns out the boot.efi patching results in both machines having the same Hardware UUID, and TM uses that for its database key. So, the two machines' backup sets were stepping on each other, making a mess. I fiddled with this a lot, even going so far as writing a binary search/replace program for the UUID on the raw ElCap volume, with no success. I ended up essentially starting over on the second machine, to reinstall everything after the mess I made. The second machine is now not backing itself up until this can be resolved. I had to throw away all the mangled backups and start TM over on the first machine. It, though, should be protected.

Sunday, June 19, 2016

Jill's been complaining about having to share the laptop, so I checked Craigslist yesterday and found a late-2008 MBP with 8MB RAM. Jill picked it up today, for $220. In the evening I wiped it and put on a fresh copy of El Capitan, and used Migration Assistant to set it up from the other MBP. It's unclear which she will prefer, the older one is much faster since it has an SSD, but one of them can be all hers. (Her choice, but I'm going to assume she'll prefer the newer one.)

Monday, June 20, 2016

I cleaned the new MBP thoroughly, getting off all the tattoo parlor stickers and residue and trying to eliminate the smoke smell and all the dog hair. There was electrical tape around the cord where it comes out of the charger (the insulation was splitting there), there was a wad of dog hair caught in the adhesive. I removed all that and used some heat-shrink tubing in its place. Much tidier, and hair-free! I used soapy warm water on a damp rag to clean everything, and gouged the guck out of the crevices in the battery hatch, etc. I pulled the back and scraped out the dust bunnies, and removed the two fans and cleaned them out. I pulled out the DVD drive and opened it up to clean it out, and found a soft plastic foreign object that would have prevented it from working. All better now. The entire machine got a gentle wash and dry, and now smells pleasantly orangey (from the Goo Gone) instead of like a stale dog with a pack-a-day habit.

I set it to making its first full backup, that'll take awhile.

While I was at it I decommissioned my Pismo. I rarely use it, as its browsers are effectively useless now. Now that we have two newer laptops the older one will take the Pismo's place, and Daniel will timeshare on it for homework.

Friday, December 16, 2016

The Sony CDP-CA7ES CD player packed it in yesterday. While it has had a tendency to get stuck in recent years, needing manual un-jamming, this was different. It acts like perhaps the motor power supply is failing, and the platter can't move well. Also, a white plastic piece fell out of the mechanism. I read up, and they're not mechanically all that solid. As we've got a party tonight, I dug out the old Kyocera DA-01, which works as well as it ever did. Nice looking unit, $850 on sale in 1983, but it's fairly persnickety. Always was.

For the evening, it worked well enough to get by. But we want a changer, for programming an evening's music. I'm not ready to do the modern music-server thing yet.

Friday, December 23, 2016

Wednesday I stopped at the thrift shop, and found a Panasonic DVD-F87 5-disc DVD/CD carousel changer. I tried it today, and it works. If you push the CD button on the front it saves time, since I think the auto-media-detection starts with DVD and works downwards.

Monday, January 2, 2017

Yesterday the 2007 MacBook Pro wigged out on me, and then wouldn't start. Blank white screen, a bit of clicking, and a repeated 3-blink LED pattern. The googletubes says that's bad RAM. I reseated it, and nothing. Divide and conquer, it turns out that the salvaged 2G stick was bad, not the expensive 4G stick. That RAM had been in there since the first RAM upgrade in 2014. Well, I have more of those! I dug out the other 2G salvaged stick, which had been retired in 2015 in favor of the 4G stick, and all is well again, with 6 GB of RAM total.

Friday, January 20, 2017

Bought another laptop, to relieve the congestion at homework time. (Which lasts all evening.) A 2008 2.4GHz Core 2 Duo Macbook Pro 4,1 with a 200G hard disk and 2G of RAM. $150. Slightly high, perhaps, but it was in good condition, etc., with a new (-ish) battery claimed.

Saturday, January 21, 2017

Cleaned the new computer (it wasn't bad at all) and named it MBP3, how imaginative. I got 10.6 and 10.11 (El Capitan) both installed on it, and fully updated. User accounts and settings migrated from MBP. I checked the scrap pile and found 4GB of RAM to put in it, doubling what was there. I pointed Time Machine at the backup server and slapped it on the ass. This one's pretty much ready to deploy. I expect that it will be the one Daniel uses for homework.

Monday, January 30, 2017

Bought another Mac Pro off craigslist today, $50 with no RAM or HD. Looks very clean. I put the 2GB of leftover RAM I had in it, and it seems to want to boot. If I can get the power draw down on it enough I'll use it to replace the G4 server. It'll be faster, and run newer software, and the RAID mechanics will be better. If not, maybe it'll replace the G5 quad cheese grater Daniel uses sometimes.

Friday, February 24, 2017

At the thrift shop I picked up a Technics SL-MC4 60+1 CD jukebox changer, $10. That should hold the whole Christmas collection for next year! It's kind of clunky, compared to the Sony, but so what. I'm just not that fond of the Panasonic changer, maybe this will be better.

Sunday, February 26, 2017

Drove to Pullman and picked up another Apple 30" monitor. A bit dim, but clean, and came with all the packaging, and a NIB Dual-Link Mini Display Port adapter. $300.

Thursday, March 2, 2017

The AT&T 5471 cordless phone which I've had forever started beeping and showing a low battery light. I checked the 4051 3.6V NiCd battery pack, and the center cell was dead and couldn't take a charge. I ordered a new one, $2.68!

Thursday, March 9, 2017

The 4051-equivalent came today, and I put it in the phone. It seems to work, time will tell.

Monday, March 20, 2017

The G4Server's WD800 HDD that I put in a year and a half ago went into the clicking mode of death today. No recovery, even after cooling down all day. Bummer.

Tuesday, March 21, 2017

I dug in the drive box and found another scrapped 80GB IDE drive, a Maxtor this time. It installed easily enough, and I formatted it via the too-small backup HDD that's in the G4Server for just these kinds of duties. I then had to use the Open Firmware trick to allow the Leopard installer DVD to run. Once I got that up I told it to do a full restore from Time Machine. It thinks it's going to take a day and a half. Sometimes it's right, sometimes it's wrong; we'll see. ... It finished overnight, so it was clearly less than about seven hours.

Wednesday, March 22, 2017

Unfortunately, it would not boot from the restored drive. After I fooled with it for quite awhile I eventually figured out that I'd not correctly formatted it, so it still thought it was a DOS disk, which it cannot boot from. (Wrong kind of partition table.) So, I get to start completely over. Rats. Next time pay closer attention, the 'erase' step is not a gimme!

After again getting the system restored and rebooted, around 6 or so hours, I again found that Apache wasn't starting up because it didn't have a log directory. This fixed it (again):

	# sudo mkdir -m 0755 /private/var/log/apache2
What a pain! I also have its Time Machine backups turned off, because I want it to adopt the old set, the old drive is gone never to return. Newer OS's have a simple command to do this, but 10.5 requires an elaborate dance.

I also found that Mailserve Pro wasn't starting up correctly, and it turns out that it also was missing a directory:

	# sudo mkdir -m 0755 /private/var/spool/postfix
Strange that these weren't part of a 'full' backup.

Saturday, March 25, 2017

Craigslist struck again! There was another Apple 30" monitor, locally this time. This one is not dim, and is clean, and came with all the packaging, $250. A much better deal. So I swapped it for the one I was using, ol' yaller can be the backup. My son expressed interest in it, and I guess it can be in occasional use rather than in a box, so long as it is understood that it's gone if one of the other two monitors croaks, so we tried it on the G5 dual he's been using. Its video card doesn't do dual-link DVI, so it only had half resolution, which is terrible on such a large monitor. So, we dusted off Jill's G5 quad that had the fancy video card in it, and copied all his files to it, and put his machine down in storage. (The only reason he was using the dual is that its monitor has an ADC connector, and the only video card we had that would drive it was in that computer, and that card in turn would only fit in that computer.) Some physical rearrangement was required to get such a huge monitor in that space, but we prevailed. (I made him do all the heavy lifting!) We tried out all his games, and everything seemed to work OK. None of them would drive the monitor at maximum resolution, I don't think any computer on the planet had that kind of resolution when those games were all written. Regardless, they all looked good and worked well.

The only problem is that I can't get it to hook up to Time Machine, it purely refuses to see the G4 file server's disks. (You can ping it, and use the Screen Share application, so it's not something simple.)

Tuesday, April 4, 2017

The RAID set on G4Server again was offline, I'm not sure what's up with that. The MBP claims that it's been unable to make a backup for 10 days, which basically means since the loss of the main HDD. I think I have some flakey disk cabling, and every time I open up G4Server I risk knocking it off its perch. The thing is just rickety and sluggish enough that I threw caution to the winds, and put both drives into the new MacPro. It booted and recognized the offline RAID set, and began repairing it. It's claiming 7 hours, and is certainly a lot more responsive while it's doing so than the G4.

... The repaired RAID set dropped back into G4Server with no drama.

Thursday, April 20, 2017

I've been waiting and watching for an ultra-low power Intel Xeon L5215 (TDP 20W, 2 cores) to pop up for sale, and I've only seen one in months, at $100. Too rich. Today I caved and ordered an L5408 (TDP 40W, 4 cores) for $18. The plan is to replace the two duals in the new MacPro with one of these, to see if that machine can get anywhere near the low power draw of the G4. If it can, I'll replace the G4 with the MP. Currently the MP, with monitor, draws about 200W when you're using it.

Saturday, April 29, 2017

I should have done more homework. Once my long 3mm allen screwdriver showed up I finally tackled the CPU swap. And.... nothing. Turns out the 54xx CPU's are not compatible with the MacPro 1,1 and 2,1 machines, and neither are the 52xx CPU's; you can only use 51xx and 53xx CPU's in them. Doh! It can, however, be saved for a MacPro 3,1 if I should ever get one.

Reading some different web pages, it turns out that somebody has actually measured idle power of various CPU's in the MacPro, and the SLAG9 5160 is a 3GHz dual that idles at only 8W, substantially lower than anything else measured. And, they're cheap. I ordered one, about $2.50 with shipping included. Its TDP is not any lower than the 65W SLABM 5150 that's in there, but the SLABM idles at 24W, 3× the power of the SLAG9. (Each, there's usually two.)

Unfortunately for the end goal of the project, it appears that the raw power draw of the power supply and motherboard is substantially above that of the G4 that I'm hoping to replace. It's looking like I can only cut about 40W out by removing one CPU altogether and replacing the 5150 with a 5160, which isn't that much out of the 200W the whole thing uses. Still, for $2.50 it's worth getting a definitive answer.

Meanwhile I can put one of the 5150's back and keep tinkering, knowing that I can only pull out maybe 16W with the final CPU swap. Sigh. Stripped, only 1 5150 SLABM CPU and headless with no optical drive, with only the minimum of 1GB of RAM installed, the power draw is 111W. So, the 'final' solution will only idle at maybe just under 100W, a far cry from the 35W or so of the G4Server. Adding back the other 1GB of RAM it's 115W, with the optical drive it's 117W, with the 7300 video card it's 136W, with keyboard and mouse it's 137W, with the monitor too it's back to 'normal', and 164W.

Saturday, May 6, 2017

The SLAG9 5160 came, and I installed it. It works just fine and seems snappy. I stripped the unit to the minimum, but kept the 2G of RAM; the power draw is more than before: a disappointing 120W. That's easily 3× the power of the G4. Feeling around by hand shows that most of this waste heat is in the Northbridge (?) and the RAM itself, as the CPU is basically cold. So, though the 5160 supposedly uses less power than the 5150 at idle, its greater clock rate probably is resulting in a faster system bus, more than making up for the power savings. Pulling out half the RAM by way of removing the riser card dropped consumption to 98W, but the system was very unhappy then. Unless there is a way to dramatically underclock the system bus, this probably means that this project is ultimately a failure.

It's a pity, the greater ease of drive swapping in the new chassis would be nice, as would the higher network speed, and gaining the actual ability to use Time Machine on the server to access its own backup set. (The G4 lacks a sufficiently powerful video card, and I was never able to find one that would definitely work for cheap enough.) Also, the newer OS's support some additional Time Machine features that would be nice to have. C'est la vie.

Sunday, May 7, 2017

I guess it's worth one more try. I ordered a 5113 (SLAG7) CPU for $9. This has the lowest specifications I could find, yet is the latest (G0) process stepping. In particular this has an 800MHz bus speed, which should cut the power being burned in the Northbridge and the RAM. Well, we'll see.

Thursday, May 11, 2017

The SLAG7 5113 came, and I installed it. It worked, and claims 1.6GHz. The full-system power draw is down only slightly: 156W compared to the 164W with the stock 5150. Stripped, on half RAM, it's 94W, down only slightly from 111W. I wonder if there is lower-power, slower RAM that could drag the bus clock down, and thus save power?

Tuesday, May 16, 2017

I ordered 2GB of 533MHz Hynix memory, thinking that slower RAM would reduce the power draw. It does not. 97W was the best I could see.

Tuesday, May 16, 2017

Some time ago I'd scored a Dell 4GB stick of 667MHz RAM from the dumpster at work, but it was a singleton, and so sat unused in a box. I thought these had to go in in pairs, but on a lark I tried it, by itself. It boots just fine, and the power draw is 89W, the lowest ever. But, still too high.

It looks like it's game-over time, unless I can figure out a way to down-clock the Northbridge and RAM.

Thursday, June 1, 2017

I did the 'quarterly' RAID disk swap in the G4Server. 14 months since the last one! It was a giant PITA, the rickety cabling and general sluggishness led me to believe it was not rebuilding, even though I eventually figured out it was by looking at the Activity Monitor's disk I/O report. I really wish I could find a way to upgrade to the MPServer, but the doubled operating cost is a real sticking point.

Saturday, August 26, 2017

I have been testing a new idea for the 90W MP server: I have it sleep all the time and just wake up for backup duty. (Time Machine supports this operating mode. I have one of the other MP's aimed at it. This worked only sometimes, because the sleep proxy server (our Apple TV) would not remember the address long enough. The MP is supposed to wake up periodically and refresh the address, but that didn't seem to be reliable, and more often than not by the next day the backup server was not reachable. The breakthrough came when I told the MP to wake up daily on its own timer, in addition to the on-demand wakeup. Most days the backup server was then reachable on demand. So, done this way we could have a mostly-reliable backup server, faster and easier to administer than the G4, which I think would be good enough. (I believe I can transfer the SATA RAID to the MP directly, with no loss of functionality or backup history.)

Doing it this way, though, means that the MP would only be the backup server, and not a file or web server of any sort. Which means the G4 server could not be retired. Annoying, but not necessarily a deal breaker. The real question, now that this seems to work, is just how much additional power running both servers will cost. I've had a Kill-A-Watt on the MP server for some time, and the numbers are: 11.7kWH over 1357 hours. That works out to about 76kWH a year, or $4.21 a year at our 5.5¢ per kWH rate. Not too unreasonable, really. I think this might just work!

Sunday, August 27, 2017

Daniel's Mac G3 "Wallstreet" has had a very flakey power cord for some time. Today we took it apart, and found that the power cord had been pushed in, bending the mounting ears and popping loose the surface-mount connections from the power socket. I hammered the body back into plane and resoldered things. After reassembly everything worked, though it was no more reliable at starting up than it has ever been. (It has an aftermarket Powerlogix G4 upgrade in it.)

The trickiest bit was finding out how to get the keyboard out! Where the later Pismo has two slide latches on top, the Wallstreet has two slide latches inside the two bays. Everything else was just screwdriver work.

Thursday, August 31, 2017

I have just not been that happy with the WWP's (now Ciena) 12-port 10/100/1000 LE-58H (CN3911) Service Delivery Switch (highly managed) that I brought home and installed when we discontinued them, so I broke down and ordered a 24-port unmanaged 10/100/1000 switch, a Netgear JGS524. Used, $50 with shipping. Today it came, and I installed it in place of the 12-port 3911 and the 5-port FS105 extender had I needed to use to cover all the ports in the house. We'll see if it behaves any better, I just think the 58H doesn't always pass all the traffic. Our network has been cranky for quite some time, we'll see if this helps.

Friday, September 15, 2017

The 90W MP server seems to be working fine as a backup server, but most of the time I cannot connect to it via screen sharing, and I end up forcing a power cycle on it in order to poke around. As an experiment I'm putting in more RAM chips, 4x512MB for 2GB (half what was in there for the low power draw, but in pairs as is recommended), and the video card. The quiescent draw is about 30W more, or around 120W that way. I'm going to see what the cumulative power draw is, and see if I can get in after it's been running a few days.

If I can find a 2014 Mac Mini (Macmini 7,1) for a reasonable price I can replace the trusty G4 server with the MP server and the MM; while the MP draws a lot more power, it shouldn't be on that much of the time, and the newer MM has only a 6W draw at idle (substantially less than the G4), for a net power savings since it'll be running 24x7. And, the pair of them will be easier to administer since they'll both be running newer software.

Wednesday, October 4, 2017

The Kill-A-Watt on the MP server, with the video card and RAM in it now, shows the numbers to be: 2.36kWH over 336 hours. That works out to about 61kWH a year, which is less than before! Clearly there are duty cycle factors at work here that overwhelm the 25% increase in power draw. More research is indicated. So far the machine has been more reliable in this configuration.

Thursday, January 11, 2018

Mid-December my MBP Macbook Pro (3,1) crapped out, and would no longer boot. Wouldn't even make a noise, dead dead dead. No amount of fiddling with it brought it back, so about a week ago I bought a used one off eBay, about $90 (shipped) for one with no battery or charger, but otherwise clean and working. It took awhile to make the decision, and to find one the same vintage so that all parts would be interchangeable. It arrived today, and booted. Seems to work. Has a 200GB HD and 4GB of RAM. The power port is a bit dented up, but functional.

Friday, January 12, 2018

I swapped in the 6GB RAM, battery, and 120GB SSD from the defunct 3,1. That went fairly easily, and seems to work fine. The new unit is 2.6GHz as opposed to the original's 2.2GHz, so that's a plus. I like that these units have LED backlights, no more aged purple/dim screens on these now-older machines.

Saturday, January 13, 2018

Somebody bought a new 4k LG TV, and sold their Panasonic TC-P58VT25 3D Plasma TV on Craigslist, I got it for $250. "It's in perfect condition." With remote and 3 pairs of active-shutter 3D glasses (2 rechargeable Panasonic TY-EW3D2M, and 1 Panasonic coin-cell based TY-EW3D10), but no stand. (The missing stand is a TBLX0158, the 'poles' are TBLA3689 and TBLA3690.) It's a $3,200 2010 model, basically the same thing we're using now but larger, and with 3D. (Yawn.) It's nice to have a spare, plasmas are gone from the market but the OLED's aren't quite where I'd want them to be if I needed to buy a replacement TV.

Thursday, January 25, 2018

Disaster! There was a new 2008.001 security update from Apple, and it kills the hacked Mac Pro units! I spent a couple of days trying to recover it from the backup, which did not work because the repair disk wouldn't work either, and apparently you can't restore El Capitan from an older OS. I ended up duping the El Capitan HD from the UN-updated machine, and then I was able to do a restore from Time Machine. The patch for the Meltdown bug is ironically named, for sure!

Friday, January 26, 2018

I made some insulation for the new speaker jacks for the 'new' Panny TV. (The built-in speakers are lame, using external speakers works very well without having to buy a sound bar.) I used a shotgun shell and cut out plastic disks, and used a hole punch on them. A little ring of heat-shrink will serve to insulate the bore of the hole in the back.

Saturday, January 27, 2018

I soldered extension wires on the jacks, then installed them in the back of the TV. The multimeter showed that they were fully isolated from the chassis, as intended. I cut the wires to the speakers and hooked them up to the jacks, it's all ready to button back up.

Sunday, January 28, 2018

I got the TV reassembled. No leftover parts!

I have, leftover from the TC-42PX24, its TBLX0134 stand, with the TBLA3685 and TBLA3686 poles. While the stand is too small the poles seem to fit well enough, and the bare metal feet (TBLA3822/TBLA3823) come out of the plastic base easily. I found the correct TBLX0158 online, but it's $100. A bit dear, considering the price of the TV itself. I think I'm going to see how it works with just the bare feet. In the worst case I can enlarge them with extensions of some sort. I need some mounting screws, though.

Tuesday, January 30, 2018

Yesterday I bought screws and washers for the TV stand, 4 M5×25 for putting the poles on the feet, and 4 M5×30 for attaching the poles to the TV. (From the owner's manual setup instructions.) About $4.50 at the hardware store. I screwed it all together today, now I just have to put it on the floor and see if it's stable enough as-is, or whether I need to make some wooden skis to extend the feet more. It's really too heavy to manhandle alone (unlike the 50"), I don't want to risk damaging it, so I'll wait 'til I have some help...

My son and I got a chance in the evening, and swapped TV's. The 58" TV is enough wider, and heavier, that you really can't reach around the back easily anymore. Oh well! It took some fiddling to get all the sources hooked back up and labeled, channels programmed, etc., but I eventually got there. It has the same connector set as the smaller one, though it does have one more HDMI input, so there was no lasting trauma.

I tried out the 3D, and it works.

Thursday, March 15, 2018

Jill's new iPad was having trouble connecting to the D-link DI-524 wifi ("cathey2") unit, so I deployed a thrift-shop Apple Airport Extreme that I got awhile ago for $10. (This is a 4th-gen unit, I believe.) I configured it for our devices that don't seem all that happy with the other two AP's, maybe it'll work better.

I think the DI-524 might be band-flipping or something, causing a periodic drop in service. I saw, using option-click on the wifi menubar icon, that it seemed to flip between 2.4GHz and 5GHz bands.

Wednesday, June 6, 2018

Jill expressed great interest in being able to print to the Canon MF4350d copier/printer that lives upstairs. The problems are two: it's USB only, and there is no network where it lives. Enter a cheap old Apple Airport Express, model A1084, which I got on the 'Bay for $12. This should have been a "just works, no trouble at all" item, being Apple consumer gear, talking to other Apple consumer gear, but the New Apple is not the Old Apple. Its problems were two: 1) The online documentation explaining how to do a reset to factory defaults is wrong, the pattern of lights does not match so it's unclear whether or not you've actually caused it to become configurable from scratch. Eventually I was able to get it to go into configuration mode. 2) Apple dropped support for this unit in the latest utility program. The utility can see it, and if you can get it programmed it will work as always, but the utility simply refuses to do anything with it. A number of suggested workarounds either failed or looked very complex/fragile. Eventually I caved in and used the old G5 quad, running older tools, to program it as a print server network extension to the "cathey3" network. Once that was done it all simply 'just worked' as I expected. (I should have remembered that my MBP can alternately boot 10.6, which has the older utility. It would have been much simpler!)

At that point I downloaded Canon's print driver for the unit and installed it on most of the machines. It printed fine. I lost the entire evening to Apple's failings in this area, it was nearly 1 AM when I finished.

Saturday, June 9, 2018

The web relay went down, there was a crash/reinstall on the relay side. I had to re-setup things, primarily the certificate-based login by copying ~/.ssh/id_rsa.pub from G4Server to ~/.ssh/authorized_keys on the relay. After that we were up again.

Monday, July 16, 2018

The G3's main hard disk died today. There was, apparently, no real backup. (Nothing vital was lost, except my e-mail archives.) The machine was too old to participate in the automated Time Machine backup system that everything else was using, and I just didn't really think about it. It just worked, pretty much flawlessly since we got it 2000. Such is the price you pay for neglect.

I wasted a lot of time trying to recover the system, the dead drive was still partially alive and there were vestiges of backup data tucked here and there, but it was pretty much a loss. Maintaining this log also stopped, because the G3 was the site master, and until I gave up on it I didn't want to move to mastering directly on the server. That took... awhile to resolve.

The G3 is alive again, with a replacement hard disk, but it's not the same.

Friday, September 7, 2018

Apparently I have inherited my dad's brand-new 2018 Macbook Pro. I will call it MBP4.

Monday, November 5, 2018

I started a new job, working for Silverwood Theme Park, and will be needing MBP4 when traveling for work.

Tuesday, November 6, 2018

The new job is a work-from-home deal. The home office needs some drastic work to make it a good place that can be used daily. The dead-but-ressurrected G3 is being dethroned from pride of place, because it is not adequate for my needs. (And hasn't been for awhile.) I removed its Sony Trinitron monitor, destined for storage, and moved MacPro2 and its 30" monitor from upstairs into its place. (I have begun re-mastering this site on MacPro2, starting with the site data that existed on G4Server.) My old job cast off a 4-port IOgear Miniview GCS1644 KVMP switch when they closed our office, which I grabbed. They also cast off a bunch of flat-screen monitors, one of which I put up on a shelf above the main monitor, and hooked up. (This KVMP switch does USB, audio, and two dual-link DVI ports, for four computers. Nice.) The second monitor is there mostly for show, but its second video input, VGA, can go to the old ADB KVM switch, so the G3 (and others) can still be used, if necessary. (Probably won't be, but...)

I also moved MPServer to the other side of the room, next to MacPro2, so that the KVM switch can reach it, too. I had to run some Ethernet cables across the room directly from the main switch, the little hub at the desk was very old, and limited to 10MB! (I need to keep it, as it's the coax Ethernet (a.k.a. thinwire, a.k.a cheapernet) converter, which the older gear needs.)

The final change, besides cleaning the room back into usability, was to put the Silverwood computer on as the KVM switch's third machine. It doesn't have a DVI plug, but does have HDMI, DisplayPort, and VGA. I have a HDMI-to-DVI cable, but you can't drive the full resolution of the monitor this way. I have an Apple mini-DisplayPort to DL-DVI converter meant for driving this exact monitor, but I need to get a DisplayPort-to-mini-DisplayPort adapter cable first. Once I've done that, the 30#34; monitor can be used at full resolution on all three computers, via the KVM switch.

Monday, November 12, 2018

I set up a "cathey3" network repeater in the bedroom, fed from a powerline Ethernet adapter. (Liberated by MacPro2's move downstairs.) The data rate is slow, but the wifi signal strength is good. Maybe Jill's phone will be happier? We really need to run an Ethernet wire, but that is not easy to do.

A significant delay was that the MBP, necessary for programming the second Airport Extreme, crashed and seemed to have corrupted the file system on its SSD. After much fooling around I resorted to the recovery disk mode, and restored from the latest Time Machine backup. It took awhile, but worked again afterwards. Hey, the system works! Nice to know.

Tuesday, November 13, 2018

The (used, $180) OWC 13-port Thunderbolt 3 dock for the MBP4 laptop, model OWCTB3DK13PSL, arrived today. The monitor and USB ports, at least, seem to work fine. Still need to test the Ethernet and Firewire ports. This will be necessary for good productivity when working from sites other than home, as laptops are very portable but never as nice as desktop gear. (On-call on vacations, etc.)

Wednesday, November 14, 2018

I tested the Ethernet and Firewire ports on the dock, and they seem to work. To actually be able to use Ethernet, you really have to play games with the Network control panel, since the Ethernet port is now 'remote' on these Thunderbolt Macs. My first experience with this, not fun. Feels even more like a PC.

Thursday, November 15, 2018

Some time ago I moved the RAID array from G4Server to the new MPServer, with it set up with a short auto-sleep, but with daily auto-wakeup and wake on network activity. Though MPServer still runs fairly hot, this way it's asleep most of the time, except when somebody is backing up, and so the net power draw is modest. Thus, we're getting the low overall power usage of G4Server but with the performance and feature improvements of the newer MacPro. (All the earlier machinations trying to get MPServer's power draw down are insignificant, compared to simply getting its duty cycle down, and were probably wasted effort.)

The daily wakeup is to get around a persistent system bug, where the distributed sleep proxy logic loses track of MPServer. If MPServer wakes up on its own, the proxy server(s) will re-discover it even if they lost track of it. (The sleep proxy system is not terribly well-designed, because the proxy pings the target once shortly before timing out. Proxies only remember targets for a few hours, and ping their targets before forgetting about them. If the target wakes up and responds, and the proxy sees the response, then the target remains available via proxy. One lost packet either way, though, and the target is lost to all its clients until it is woken for some other reason. Hence the daily auto-wakeup as a fail-safe.) I have replaced the problematic Ciena 3911 with a simple Netgear switch, which I think is doing a better job of not losing packets, and we now have two Apple TV's and two Airport Extreme units, all of which run sleep proxy services. Backups seem to be happening more reliably now.

While G4Server has been a stalwart workhorse, and in fact is still operating as the 24×7 web server, MPServer is a far better Time Machine backup server. (Once I can find the right deal on the right Mac Mini [2014, 7,1] to be the web server, which will have even lower net power consumption than G4Server, the G4 can be honorably retired. With the death of the G3, mail relay service is no longer needed.)

The slower G4's have an excellent reliability record, but really don't want to run 10.5, necessary for Time Machine. Faster G4's are available, and could better host 10.5, SATA drives, and Gigabit Ethernet, but they are much hotter, louder, and less reliable. At that point a Mac Pro 2,1 is a better choice, because it will run 10.7, supports more RAM, and it has those nice drive sleds. And, the power draw can be mitigated as described above, and they're getting pretty cheap due to their age and inability to run the newest software.

I also procured a fourth 2TB drive, for two independent offline slices. All four drives thus never need be physically co-located, thus improving overall system reliability. (A fire on slice rotation day still leaves one intact drive, elsewhere, from which to recover.) In fact, there is no point to all the drives ever being physically present together, because there are only three open drive bays for the RAID in MPServer.

Trying to make a 4-slice RAID, though, in a unit that has only 3 open drive bays is tricky. Apple's RAID GUI has enough bugs that you can't do it that way, but fortunately the CLI can get the job done. You can use the GUI to build the basic 2-drive mirroring RAID and get things all set up and working, then install a blank (matching) drive and:

diskutil list
diskutil appleraid list
diskutil appleraid add member disk3 disk4
I've done this twice, now. ("disk3" is the new blank disk, and disk4 is the virtual drive that is the RAID, to which you are adding a new slice. Make sure these are the correct names for your situation!) The system will add the slice, and start copying all the mirror data to it. Once it is rebuilt you can pull it and store it off-site. The RAID will always think it is in a degraded condition, because all the slices will never be present together, but this seems harmless. I believe that swapping out one online slice with one of these offline slices will result in an automatic RAID rebuild on the newly inserted slice. If not, you can always get it done manually, via diskutil. (The intent is to do this rotation quarterly.)

These are the salient features of MPServer and its 2/4-drive mirroring RAID:

On another topic, Jill has not been happy with the printer situation, especially color. Today on Craigslist was an HP CP2025dn, $50 and claimed to be working fine. At that price I bit. At lunch time I went out and bought it, and lunch.

I installed the CP2025dn in place of the failing (feed problems) 2200dn, and cabled it directly to the house network instead of off the second Ethernet port on MacPro, so it can be easily shared. It worked perfectly.

On a roll, Daniel and I cleaned out two holes in my office and moved the big black and white (HP) and color (Xerox Phaser) printers into it, out of the family room. (A move Jill has been lobbying for forever.) The HP is happy there, and seems to work. It'll be a real pain when it jams, though, because it has to be pulled out to open any side doors. Oh well.

Friday, November 16, 2018

I finished decommissioning the family room printer cluster, moving it and the powerline Ethernet adapter into the office. (That 24-port Ethernet switch is really filling up.) The biggest problem with the Xerox Phaser color printer is that it takes forever to warm up, and wastes huge amounts of expensive wax while doing so. It really wants to be left on 24×7, but asleep. We tried that once, but Jill objected to the wax smell, which should not be a problem in my office, and every time the power went out the printer would spew a bunch of wax into the waste tray when it came back on, and that wax is expensive. With the move of MacPro2 to the office it's now on the big UPS, which freed up the APC1500 that it had been using. I put the UPS on the Xerox, which is definitely against recommendation. (Heating elements are high power! It definitely threw overload indications while the printer came up.) We'll try leaving it on this way to see if it's any happier. Otherwise, it'll probably have to go. I'd like to keep it, because it's a multi-function machine and has a very good color scanner, and can be a duplex color copier too. So, we'll see how this attempt goes, the printer was quite the strain on the UPS. We don't want the UPS for when it's operating, what I'm after is keeping the printer alive while it's asleep should the power go out. So long as the printer can operate OK through that UPS, when the power is on, we should be good. (The printer wants more wax, there's a shock.)

Saturday, November 17, 2018

I upgraded the Xerox Phaser 8560MFP firmware. It went from 8.26/4.4.0/19.P3.4.2.0/ to 8.54/4.9.4/19.P3.4.9.4/

I've been reading, and apparently every time you exit power saver mode it burns ink. I've set it to 4 hours to enter that mode, the longest available time. I had thought that power-saver mode prevented ink burn, but it does not, and the firmware upgrade offered no relief. If only the ink were cheaper!

Saturday, November 24, 2018

The MBP3 machine died some time ago, black screen no reaction. This is the second MBP of that vintage to go out. Some reading shows that these are particularly prone to thermal fatigue on the BGA devices (CPU, GPU), and that a home reflow can often bring them back. These devices are sufficiently old that no commercial repair is economically justifiable, and so with nothing to lose I tore into it. The secret is 400°F for 7 minutes in the oven, followed by a gentle cooling. Only the logic board, nothing else can take the heat, including the little black plastic spacers at the hinge end of the logic board. (Oops.)

As always, the hardest part is disassembly/reassembly. I got the logic board out, and found that a pop rivet from the junk box was a perfect fit to one of the central mounting holes. I slipped it in that, and clamped the stem of the rivet in some all-metal vise grips, which served as a heavy base. This sat nicely on a cookie sheet, with the BGA devices up.

After baking and letting it cool to the touch in the oven with the door open, so as not to jostle the hot solder, I put it together far enough to try testing. Success! It came back on. Unfortunately I was ham-handed in removing the little connectors, and broke some of them, so while the operation was a success the patient died anyway. (Those tiny connectors pry up off the board, they don't slide out like the older, bigger connectors they resemble. I tried to recover things, but it really didn't work out. The machine worked well enough to prove the concept, but was essentially ruined.

Sunday, November 25, 2018

However, there were two of these in the broken MBP pile, so with my hard-earned experience in hand I tackled the first one that had died years ago. Knowing the tricks this time it went a lot smoother and, after baking, this other one also seemed to come back to life. Using pieces from both 'dead' MBP's I was able to assemble one unit that was complete and working. (Getting it all back together cleanly, with no gaps in the case etc. was the most difficult part.) The original HD from MBP3 also seemed perfectly usable, so I didn't even have to do a restore from backup step. Once it was booted and running again I had to do a "tmutil inheritbackup ..." step, because to the rest of the world it looked like I had put in a replacement logic board, which is in fact exactly what I had done. This is in order to allow the backup set to continue on from where it left off, rather than starting over.

Had I been more careful, both laptops would have been fully resurrected. Oh well!

Friday, January 11, 2019

The MBP's Time Machine backup has been whining lately about needing to start a new backup set. The error I see is "Time Machine completed a verification of your backups. To improve reliability, Time Machine must create a new backup for you." I Googled, and this seems to be the standard reaction to what it perceives as a corrupted virtual filesystem, the sparsebundle that actually holds the backup on the server. I don't want to start the backups from scratch, although this wouldn't be the worst thing in the world. Fortunately there are some suggested fixes on garth.org, involving fsck on the file server, which I applied. (This will take hours, plan accordingly.)
# Gain privilege:
su admin
sudo sh

# Get to the problematic sparsebundle:
cd /Volumes/Backups

# Remove the 'bad' marking that TM has put there.
chflags -R nouchg {name}.sparsebundle

# Mount the bundle as a virtual disk volume.  Note the device name
# of the HFS partition, reported as something like /dev/disk9s2.
hdiutil attach -nomount -noverify -noautofsck {name}.sparsebundle

# If it didn't already start auto-repairing (it shouldn't), fix it.  Patience!
fsck_hfs -drfy /dev/disk9s2

# Unmount the virtual volume, post-repair.
hdiutil detach /dev/disk9s2
You're not done yet. Now we have to tell TM it's OK to use it again. You need to edit a plist file within the sparsebundle that records the state of the backup. At the top level of the sparsebundle directory find a file called com.apple.TimeMachine.MachineID.plist. Edit it and remove these two XML nodes:
Finally, you want to change:
# Fix Time Machine status.  Apply the edits specified above.
emacs -nw {name}.sparsebundle/com.apple.TimeMachine.MachineID.plist

Sadly, all this effort did not work out. (Perhaps I missed a step? Regardless, I'm tired of fooling with it.) The next backup on the laptop offered to adopt one of the other laptops' backup sets, or start from scratch. Scratch it is, then. On the plus side, that freed up some space.

Thursday, February 7, 2019

The Park was throwing away a MacPro (?) ATI Radeon HD 5770 video card (they'd upgraded a Mac Pro to an RX 560). I nabbed it for a spare, in case one of our two 4870's goes out. (Supposedly very comparable in performance, but consumes a little less power.)

Thursday, February 14, 2019

Recently the used Panasonic DMP-BDT220 3D Blu-ray player I got to go with the 3D TV crapped out. (I put the old DMP-BD35 back temporarily, and we finished the movie in 2D.) Investigation showed that it would play DVD's and CD's well, but the blue laser was apparently done. Cleaning the lenses did no good. (Some research showed that not too long ago they were trumpeting a major breakthrough in blue laser lifetime, up to 10× the previous state of the art and getting into the 3,000 hour range, which is not really all that long. Especially if you're prone to wandering off while the movie is playing its main menu choice video loop...) I ordered a Panasonic DMP-BDT230GA 3D Blu-ray player from B&H electronics, $81.55 NOS. Today it came, and I tried it out. The 3D works. The unusual thing was that it came with two non-USA power cords, but I scrounged up a cord that fit and it seemed happy enough. Is that the GA option? Perhaps that's why it was substantially less money than the alternatives. Good thing it seems to have a universal power supply in it, and is still using NTSC format.

Memo to self: stop buying used laser gear!

Wednesday, March 20, 2019

Yesterday I picked up the second Mini-DisplayPort-to-Dual-Link-DVI Apple adapter (MB571Z/A) at the Park, $50 used off of eBay, and today I cabled it in to the desktop cluster. This was the last step necessary to get Dad's laptop hooked up as a full peer of the other machines I use on a daily basis, using the fourth and last slot on the KVM switch. (And thus avoiding the Screen Sharing application that I've been having continual re-authentication trouble with.) The laptop is connected to the OWC dock, and that goes to the new adapter, and thence through the KVM switch to the big 30" monitor. With a bunch of fooling around I got everything cabled back up, and in such a manner that both the laptop and the dock may be easily taken for transport. (I hope to find a second dock at a good price, it'll be nice to have a backup and also to reduce the necessity of moving the dock too on remote days. With a second dock all I'd have to do is grab the laptop and remove the one cable that goes to it. [And don't forget the external Silverwood disk drive.] Currently the 13-port OWC is a discontinued product, but still sells for $250, vs the $180 I paid for this one.)

Wednesday, April 10, 2019

Bought a 2011 Mac Mini, with 1.5TB external drive and wireless trackpad. $175 off Craigslist (Portland), 2.5GHz, 16GB/500GB. For use as a backup in case the Macbook Pro should fail. In theory VMWare Fusion 11 will work on it. It's currently scrubbed clean, and loaded with High Sierra.

Thursday, April 11, 2019

Got VMWare Fusion 11 installed on it, and working, so it's a go as a viable backup system for work. It's nice when plans actually work!

Thursday, May 9, 2019

Won (eBay) a 2014 Mac Mini. 4GB/500GB claimed. $250, shipped. We'll see if it's as described, I've seen Minis for sale advertised as 2014 that are actually older than that. The listing didn't give enough information to cross-check the claim. (That's what the "not as described" process is for, if necessary. I used Paypal.) As is my usual now for more expensive items, I had it shipped to the Park to avoid the latest tax grab.

Wednesday, May 22, 2019

Picked up the Mac Mini at the Park. Powered up it is indeed a 2014, the slowest 1.4 GHz model and with the claimed storage, but the Ethernet wasn't working, even after deleting and recreating the interface. After an OS update the Ethernet worked again, so that's a big relief. There was enough residue on the hard disk, though, that I'm formatting the HD and starting from scratch with High Sierra, and calling it "MMServer." I don't need anything that might have been left on it, and I don't want any complications.

Thursday, May 23, 2019

Got all new High Sierra software on the new Mini, and fully updated. I Googled for instructions on enabling the web server, and they're fairly basic, since I don't want server-side execution like PHP or PERL: To get FTP back, so that my mastering flow doesn't need to change, I first had to install the Homebrew tools package. Then brew install inetutils, brew install pure-ftpd, and sudo brew services start pure-ftpd. Google for it. After messing with this I found that ftpd wasn't all that easy to get running, so I punted and just switched to using ssh tools instead.

I decommissioned the faithful G4 Server (RIP!) and put the MM in its place. After tweaking its network and startup options appropriately, it seems to be functioning as our Web server, just as the G4 did. The amusing thing is that after all this the fanage in the office isn't all that less obnoxious. That's because the G4's fan was actually well designed, and though not exactly quiet it was a dull roar that was not too bothersome. The higher-pitched, more-annoying fan in that corner is in the Netgear JGS524 switch. I searched for fanless equivalents, and came up with a $40 (used) Netgear GS324 switch. I ordered it off eBay, we'll see if that resolves the issue when it comes. Though 24 ports is barely enough, the 48-port big brothers are stupid expensive, and not common.

Dad's laptop has been unable to install OS updates since I got it, I think Dad crapped it up pretty good during his decline. Having just installed a new OS from the network on the 2014 Mini, and with the Park in a semi-stable position, I felt more confident doing this now on the laptop. There was a significant setback, though, as whatever Dad (or perhaps I) had done with the machine had caused the loss of the recovery partition, so it couldn't self-host its own recovery.

Being the Clever Dick that I am, however, I decided that with the help of the 2011 Mac Mini I'd bought as a spare workstation I can still get the job done myself without resorting to a trip to the Apple store. (I can't do it with the 2014 Mini as it has neither Firewire, nor Thunderbolt 3.) Steps:

A fun (re-)discovery was the HDMI⇆DVI adapter stored in the 2011 Mini's box. Exactly what I needed on the 2014 Mini to talk to the 20" Apple Cinema display that is there for use on that machine, on the rare occasions where that sort of thing is necessary.

There was a delay, however, as the recovery software insisted that it needed an update first... Having done that it wants to do it again. And again. Something is wrong with this process.

Friday, May 24, 2019

I'm going to try a different tack. I should be able to make the external 2TB drive box bootable. Booting from that I should be able to install a fresh OS on the laptop proper. First step is to download a full High Sierra installer. That's a bit tricky, Apple doesn't really encourage this anymore. I'm using the aftermarket High Sierra Patcher to accomplish the 5+GB download...

That took quite some time, there were failures in the middle and I had to re-start (resume) to get it going again. I copied the installer to the 2TB drive, and also ran it and installed the OS on the 2TB drive so it is bootable too.

With that done, the laptop could be booted from the 2TB drive... Not. The laptop's default 'T2' security chip setting is to refuse to boot from external devices. The only way to change that setting is to use recovery mode, which does not work. Catch-22!

It turns out, though, that the Internet Recovery mode did work. By holding down CMD-ALT-R I was able to (eventually) get to the Mojave (!) utility that lets you turn down the security settings to the traditional level. This allowed booting from the external drive, except that the laptop still refused! It would, however, tolerate booting from the Mac Mini via its target disk mode, but that refused to install the 'older' (a lie) High Sierra on the laptop, and suggested I use Recovery mode. That, however, still refuses to boot without an update, and I'm not falling for that again. Trying to use Internet Recovery itself to install an OS only offers me Mojave, which I do not want.

Total frustration, a solid day wasted, and still no repaired laptop.

I took a break, and then came back at it. I tried partitioning the 2TB drive so that I could copy everything off the laptop to it, prior to a wipe and clean install of the laptop. That didn't go well, and in frustration I tried CMD-R again. It worked! I don't know what exactly in all the thrashing around I did that 'repaired' the recovery partition, but I'll take it. It's offering to re-install High Sierra on the laptop's own HD, which is all I really wanted in the first place. It claims it'll need 2+ hours, which I'm prepared to give it at this point.

It was a lot longer than estimated, and there were some updates afterwards, but it looks like I've finally got a clean OS on the laptop. The stupid system apparently will only use wi-fi, and not the cabled Ethernet, I'm sure that cost some time.

Also, at the moment, my 2TB work drive is solidly crapped up with partitions and such that I do not want.

Saturday, May 25, 2019

Using the laptop, removing the extra partitioning on the 2TB drive was quick and painless. The laptop, however, will not boot from it, simply showing me a barred circle, nor will it install the OS on it using the full installer I downloaded. Says it's 'too old'.

The 2011 Mini, though, will boot from the 2TB drive. I did so, to the recovery partition, and re-installed the OS on that drive, and then updated it. It took forever, but I wasn't there to watch it so it didn't matter.

Sunday, May 26, 2019

I booted the laptop from the refreshed 2TB drive, it worked fine this time. As of now, the laptop and the Mini backup both are fully up to date, and the 2TB drive can be used to boot either one of them in a pinch. We're in good shape.

Tuesday, May 28, 2019

I managed to eradicate "Clean My Mac 3" from the laptop. Finally! Besides killing any active processes, the major tool was "find", looking for "*CleanMy*" and "*macpaw*" files and directories. There were a lot.

Wednesday, May 29, 2019

I compared the two Mac Minis' video performance by running Unigen's Valley demo program on both, using the same HDMI-connected monitor. Even though the 2011 has a faster CPU, more memory, and a separate AMD video controller, the 2014 kicked its butt. Huh. The old Mac Pros, though, are still the solid winners, by a wide margin.

Saturday, June 1, 2019

I picked up the GS324 ethernet switch at the Park, and installed it in place of the JGS524. Ahhh! Blessed silence! The GS324 is rated at 11W (maximum), vs the JGS524's 18W (maximum), so that's also a win. It also has twice the buffering, 3× the MTBF, one third the packet latency, and even cost less. Win, win, win, and win.

Monday, June 3, 2019

Long past time for the 'quarterly' rotation of the offline/offsite RAID storage. I had been waiting, though, for the MMServer change to settle down before I did this, which was last done 11/18. I rotated the northern backup this time. Once the RAID started rebuilding I deleted the online G4 server backups, which will free up some space for the new MMServer backups. (We don't need archival backups of decommissioned machines, the machines themselves are still accessible, in storage.) The RAID rebuild began on its own, as expected. All I had to do was shutdown, swap out the #4 RAID drive sled, and fire it back up again. The automatic rebuild is estimated to take about 9 hours.

Tuesday, June 4, 2019

Now that MMServer has stabilized in its role a bit, it has used 0.29 kWH over 64.5 hours, which would extrapolate to around 40 kWH/year, or $2.16 in electricity at our 5.5¢ per kWH rate. (I'll refine these numbers later.) That's around half what the backup server consumes, and the pair of them together (MMServer and MPServer) still consume far less power than G4Server did, when it was serving both roles. That 24×7 multiplier really adds up...

Wednesday, June 5, 2019

Jill's MacPro has been acting flakey lately, and when I tried to boot from its recovery partition to do a Disk First Aid it went insane, and couldn't even boot normally after that, so it was clearly in trouble. I did some flailing around, and what worked best/easiest was to:
  1. Pull the user files (Lion) disk from the MacPro, for safety;
  2. Insert one of the (recent) offline/offsite RAID backup drives to serve as a data source;
  3. Grab one of the ElCapitan MacBook Pro's and boot it into recovery mode;
  4. Connect the two systems via a Firewire 800 cable; (Firewire 800 being faster, and certainly easier, than getting the network hooked up to the online backup server so that I could do the restoration without the help of the laptop. I didn't need or want the current backups anyway, this system has been unhappy for awhile.)
  5. Boot the MacPro in target disk mode;
  6. Restore the SSD from the time machine backup, using the April 24 backup; (Somewhat randomly chosen, as being not extremely recent, but unlikely to be too far out of date.)
  7. Reconfigure the MacPro back to normal, and reboot.
This will have no effect on Jill's files, which are stored on the Lion volume. The only things on the SSD itself are the OS (unchanging), Applications (fairly stable), and system settings (re-creatable). The whole process takes (claimed) about an hour and a half. For extra credit, use Time Machine proper to restore the Applications folder afterwards.

Saturday, June 8, 2019

Further MMServer numbers: 0.72 kWH over 160 hours, which still extrapolates to around 40 kWH/year.

Tuesday, June 11, 2019

Final MMServer numbers: 1.09 kWH over 243 hours, which still extrapolates to around 40 kWH/year. The machine wanted to update, which required a restart. This has been tying up the Kill-A-Watt for long enough, so since it was an opportune time I powered it down and removed the tester from the circuit.

Friday, August 2, 2019

My main Mac Pro (MacPro2) puked, and I ended up having to do a format and full restore from the Time Machine backup. (It was running less and less time between crashes, and finally got to the point where it couldn't finish booting up without crashing again.) The problem seems to be tied to the SSD boot drive, which is all I restored. (I had to do this same thing to Jill's earlier.) This is easiest/best done using a laptop against the Macpro running target disk mode. (Both machines run El Capitan, matching the OS's is Important else the Restore function can't work.) I restored over the network this time, rather than grab from an offline RAID drive. The secret is to boot the laptop with Cmd-R, then engage the Terminal and run these commands first before engaging the Restore function:
mkdir /Volumes/TimeMachine
mount -t afp afp://user:password@MPServer.local/Backups /Volumes/TimeMachine
hdid /Volumes/TimeMachine/MacPro2.sparsebundle
That gives the restoring laptop access to the remote backup set. I went back into July for the restore point. You must leave the Restore session alone for awhile before the restore data set will become visible on its own; don't be impatient!

As part of the crash diagnostic procedure, involving removing memory, etc., I had replaced the 4870 video card with the 5770 I had scored earlier. (The machine continued to crash with the 5770 in place.) I kept the 5770, it's lower power and seems to be quieter fan-wise, and it was more work to swap it back which didn't seem necessary. It doesn't have the second DVI connector I was using for an auxiliary non-KVM-switched monitor, so I've ordered a cheap mDP→DVI adapter. For the laptop I also ordered another USB-C converter, and an older 10-port OWC dock.

Monday, August 5, 2019

These last few days have been a nightmare but we've finally gotten all three cell phones switched to T-Mobile. (The big problem was getting AT&T to unlock Dad's old iPhone 6, to which Daniel is upgrading as his new semi-leadership role in marching band pretty much requires this.) Much better data plan than before, for about the same cost; instead of 4GB shared among three phones each phone gets 25GB/mo, and if overrun the particular phone's connection slows down rather than getting hammered with a stiff fee.

Monday, August 12, 2019

So far the MacPro has been reliable since its surgery.

The recently-ordered Maccessories have arrived. The third USB adapter for the MBP4 laptop seems to work ($7.08, eBay), as does the mDP/DVI adapter ($6.52, also eBay) for the 5770 video now in MacPro2. (This cheap adapter is only for smaller, single-link DVI monitors, like the second one on the MacPro.) Three USB adapters means that all four Thunderbolt 3 ports on MBP4 can be in use when dock-less on the road, etc. The additional laptop dock ($28.80, eBay) also came, but there was no time to check it out.

Tuesday, August 13, 2019

My on-the-road monitor has Displayport and DVI inputs, and the new OWCTCDOCK11PSG dock is HDMI. I didn't seem to have the necessary video cabling, so I ordered an HDMI⇆DVI cable, about $7. I checked the new dock out, the online reviews are scathing. I hooked it up in place of the downstairs Apple TV, and it seems to work OK. It even went to sleep and woke up again, so maybe it'll be all right. I did have to power-cycle it after the Ethernet and HDMI ports were cabled before they connected. Yes, the charging brick is a bit underpowered, but I think it'll be OK. It's missing the short cord that goes between the dock and the Mac, the seller says it had gotten destroyed. I ordered another one, about $14 via Amazon.

Friday, August 16, 2019

MacPro2 melted down again, and was unable to reboot successfully on its own, even after power-cycling. It crashed during a CMD-R recovery session disk repair, but it seems to have come back in working condition afterwards. It ran perfectly for only two weeks since last time.

Saturday, August 17, 2019

MacPro2 still wouldn't run, post-restore, so I went back to July 13 for another restoration, then turned it off to cool back to ambient.

Sunday, August 18, 2019

MacPro2 seems fine this morning. Sigh.

The HDMI⇆DVI cable came, I tried it out in the mobile rig. It all seems to work just fine. If it holds up, it looks like the cheap dock will be a success.

Friday, August 23, 2019

MacPro2 crashed. Seemed to come back OK, though.

Sunday, August 25, 2019

MacPro2 crashed again. Harder, this time. It doesn't seem to be coming back.

Monday, August 26, 2019

I installed a 250GB hard disk in MacPro2, and restored the OS to it, as I have long been distrustful of the SSD. Unfortunately this did not help, the machine was still far too flakey to use, and didn't behave even as well as with the SSD. I played memory swapping games for some time, but was never able to find a configuration that would stay up for very long, in operation.

In desperation, so that I could get some work done, I spun it up in Target Disk Mode, and mounted its disks on MBP4. That seems to work fairly well, but is not viable for the long-term and runs fairly warm.

I think I'm going to have to replace MacPro2. Ideally, with a 3,1 (or newer) that can run El Capitan natively. (I know I missed a shot at one on Craigslist fairly recently.) Now that the cheese graters all have been abandoned by the latest OS, prices should come down some more. Other than the flakiness, I've been very happy with it, I think getting another is still a reasonable thing to do; El Capitan has only been superseded for three years. I don't want to commit to using only the laptop, that leaves me too vulnerable.

Tuesday, September 10, 2019

T-Mobile is an abject failure. Doesn't work most places we spend time, and is crap on the road. Even with the cell phone amplifier.

I stopped by the shop and had a little discussion. One 'new' fact surfaced: The AT&T fallback is not blanket, it's negotiated on a tower-by-tower basis, and is only available where AT&T chooses to allow it. Clearly not available in Long Beach. Also, their coverage maps don't show all frequencies, just the new ones that my 4S can't use. So, I had zero coverage my last trip down there.

Saturday, September 21, 2019

I took another stab at getting OSX Snow Leopard (10.6) running in a VM on MBP4, but this time using VirtualBox (VB) 6.0.12. It worked! Unlike the other VM products I'm aware of, this one doesn't obey Apple's spurious restriction to disallow 10.6 Client (but not Server), and it booted the VM and ran the install from the DVD. Getting Migration Assistant to do its job, though, was a bit tricky, since VB doesn't seem to allow mounting host disk volumes to the VM. (It does allow mounting the optical media drive.) First you have to make sure the Network (VB Settings) is set to Bridging, allowing free bidirectional access to the local network. Then you have to mount the Time Machine image from one of the existing laptops' 10.6 backup partitions, and migrate from that.
mkdir /Volumes/TimeMachine
mount -t afp afp://user:password@MPServer.local/Backups /Volumes/TimeMachine
hdid /Volumes/TimeMachine/MBP.sparsebundle
Unlike the VMWare Fusion VM's, all of which I have pointed to the external Silverwood disk for storage, for this one I use the laptop's own SSD. That's because it's for supporting our own older Airport hardware, some of which can't be configured using anything newer. (Fusion is Silverwood's, and that's what I use it for.)

I then ran all the necessary Software Update cycles, and used this 10.6 VM to update the firmware on our Airport Extremes (2) to 7.8.1, the very latest. Worked slick.

Wednesday, September 25, 2019

Jill's been complaining that MBP2's trackpad clicker is becoming very non-responsive. This is a common complaint with these, and there's a pre-tensioning screw under the battery that you can fiddle with. Unfortunately it requires a tri-lobed driver, which I don't have. I was able to mess with it using a small flat-blade, I'm not sure whether I tightened it a skosh or not. I also shot some De-Ox-It under there, in the hopes that it'll wick into the switch contacts. She says it's better, but we'll see how it holds up over time. (The late 2008 MBP uses a unique trackpad, not that easy to come by.) I brought her the wireless trackpad (gotten with the Mac Mini) and paired it to MBP2, for now. In a pinch, she can still get things done that way.

Daniel brought home an older Dell laptop, an Inspiron 1520, running a lightweight Linux (MX Linux, Xfce). (A toy, basically, and a gift from a friend. Might be worth $50 on the open market, and pretty comparable to the Macbook Pro he's been using.) Naturally it doesn't have Apple's parental controls on it, but if Daniel thinks he can now surf in the middle of the night he's got a surprise in store. I can enable the MAC-address filtering on the wifi routers...

The Dell's wifi MAC address is 00:1E:4C:26:34:39, which I gleaned from the Mikrotik's DHCP Server report. I taught the cathey3 wifi (Airport Extreme) this address, and put it on a timer so that it's not available between 9PM and 6AM. Daniel doesn't know the key for cathey2, it's really hard to remember, and to type, so that's automatically pretty safe. (We don't need it anymore, it was a failed attempt at extending our in-house wifi later satisfied by cathey3. I'd remove cathey2 altogether, except that we're using the switch part of it for the TV stack.) The original cathey network is unprotected, but the mikrotik (host for cathey) has a basic access list mechanism, so I just told it (interface: local-ap) to never talk to the Dell—for the Dell it's cathey3 or nothing, and cathey3 is timed. Problem solved.

MBP3 (Daniel's) seems to be dead. It doesn't beep, and the activity LED only comes on dimly. Did we not even make it a year since I repaired it? It's not looking good.

Tuesday, October 1, 2019

I found the kernel crash logs for MacPro2, they're in /Library/Logs/DiagnosticReports/. There were a bunch, no particular pattern to them. All were either General Protection (13) or Page (14) faults, the error code for all was 0's. The fingered applications varied, indicating a general memory-type problem. Getting curious and knowing that Jill was away for the week, I took the memory cards from MacPro and put them in. We'll see if it can survive better using these. Maybe this is just bad memory? Most of it is not the official stuff with the giant heat sinks.

I ordered 8GB of the official heat-sinky RAM, $21 with shipping off eBay.

I ordered a replacement for MBP3, same early-2008 vintage. No HD or battery, but I can re-use our existing parts. $94, shipped, also from eBay.

Friday, October 4, 2019

MacPro2 ran for days, using Jill's memory cards, without a glitch. I ordered 8GB of official stuff, with the big heat sinks, for $20. With Jill back I put her machine back together, and using the original memory about the best that I can do is to go back to using my machine as a target disk.

Friday, October 11, 2019

I put in the 8GB of 'new' RAM and MacPro2 came up, at least. The RAM bank A (according to Temperature Monitor) was at 261°F, (!) but when I pulled it out it certainly didn't feel that hot. I swapped the two memory cards (switching A & B), and then all RAM read at 150°F, which better matches what I could feel with my hands. Anomaly?

On a roll, I also swapped the HD from MBP3 into the replacement. The 'new' one has a 2.6GHz CPU (vs the 2.4GHz of the dead one), and a glossy screen. Of concern was that not only was the HD missing, the HD cable was torn in half, and part was missing. Fortunately we can just move that from the dead machine too. Everything came up OK, even the DVD drive works.

Monday, November 11, 2019

I ordered an additional 16GB of used RAM for MacPro2, $33 via eBay. Assuming this works, this'll give both machines 24GB each, which is more than ample for our needs.

I took a shot at making a VirtualBox installation of Lion, so I could run DxO without rebooting, but I was unable to get VBox to map an unused SATA drive to be the disk for that OS. Using dtruss to investigate, this is the relevant error:

open("/dev/disk0", 0x1000000, 0x180)              = 7 0
fstat64(0x7, 0x7FFF59979218, 0x180)               = 0 0
ioctl(0x7, 0x40086419, 0x7FFF59979210)            = 0 0
ioctl(0x7, 0x40046418, 0x7FFF5997920C)            = 0 0
close(0x7)                                        = 0 0
open("/Users/jimc/VMLion.vmdk", 0x1000A02, 0x180) = 8 0
open("/dev/disk0", 0x1000002, 0x180)              = -1 Err#16
The open/ioctl/close sequence on the raw drive apparently causes the next open of the drive to fail:
VBoxManage: error: VMDK: could not open raw disk file '/dev/disk0'
VBoxManage: error: Error code VERR_RESOURCE_BUSY at .../Storage/VMDK.cpp(3426) in function int vmdkCreateRawImage(PVMDKIMAGE, const PVDISKRAW, uint64_t)
VBoxManage: error: Cannot create the raw disk VMDK: VERR_RESOURCE_BUSY
VBoxManage: error: The raw disk vmdk file was not created

Thursday, November 14, 2019

The joker was that disk0 was still partially mounted in some way, in spite of my issuing umount commands against it. This, however:
diskutil unmountDisk disk0
chmod 666 /dev/disk0
cured the problem. Thus I issued the following commands:
VBoxManage internalcommands createrawvmdk -filename /Users/jimc/VirtualBox\ VMs/VMLion.vmdk -rawdisk /dev/disk0
chmod 666 /Users/jimc/VirtualBox\ VMs/VMLion.vmdk
I then created a new VM, and told it to use this .vmdk disk control file. I mounted the optical drive to the VM, and inserted the Lion install disc. Then, Start.

Saturday, November 16, 2019

After much time lost to trying to get this VM created while running the normal live system, I finally punted and just rebooted and ran a Lion install native, giving the new drive minimal configuration. I then used Migration Assistant to take the photo-stripped Lion environment I'd made in .dmg files (base, plus shadow) from the external USB drive. This all took most of the day.

So far I've learned a lot, but nothing particularly useful.

Sunday, November 17, 2019

Failure. Not only was I never able to quite get Lion running, after many attempts, as it turns out VirtualBox cannot run more than 1 CPU for any OS X guest OS. (More is planned for some mythical future version.) Who knew? I guess I should have done more reading up front. I will not start this crap over with VMWare, which would require down-revving (another open-ended can of worms) in order to run on El Capitan, with no better guarantee of success than I had with VirtualBox...

But besides the inconvenience of rebooting to Lion to run DxO, the real problem was that running Lion against the same user directories as El Capitan caused it to mess up saved application state. However, if I keep the additional HD I put in for VM Lion, and just boot to it instead of the normal Lion/El Capitan disk, I should avoid that part of the problem. (The inconvenience remains, but is lessened considerably.) I would just buy DxO, but they won't sell me a version that will run on this machine... Near as I can tell, in their history they never would sell me a version that would run on any of my always-slightly-dated machines. They seem to operate only at the bleeding edge, those of us with trailing-edge budgets are always SOL.

Nothing, with computers, is ever as easy as it should be.

Monday, November 18, 2019

Jill's MacPro wouldn't come up, and it was blinking the 'bad RAM' light on the front. I investigated, and found that the two 4GB sticks in it were NFG. These were the last two in service of the free batch I got from Ciena during their housecleaning, and the only two in the machine that didn't have the correct big heat sinks. These have all failed in service, clearly the additional heat sinks are important when used in a Mac Pro. That still leaves her with 16GB in the machine, so we'll see if I need to replace it or not.

Wednesday, November 20, 2019

The additional 16GB of RAM for MacPro2 came today and I installed it, giving it 24GB of RAM total. It seems to work. There's now ample room for VM's, but it doesn't look like I'm going to need them. I did set up and try 10.6 in VirtualBox, which seemed to work fine.

Tuesday, March 3, 2020

A MacPro 5,1 (firmware upgraded from 4,1) turned up on eBay, and I bit. $516, shipped. This is almost exactly what I am running now, 8× 3GHz cores and both SSD and HD, except that it can run High Sierra, which means I can run the Silverwood VMware and get entirely off the laptop while working at home. I can upgrade this to 12× 3.4GHz (2× X5690) cores if I should want to, though this is both difficult and not particularly cheap.

Sunday, March 8, 2020

Yesterday a quad 3GHz Macpro 1,1 running El Capitan and loaded with 32GB of RAM, with 23" Apple monitor (via ATI 5770), turned up on Craigslist; I got it for $200. (It has WiFi, but not Bluetooth, and two optical drives.) The plan is to raid it for some RAM, and maybe a drive sled, then set it up and give it to Steve. (This was easily $4,500 when new, perhaps more if you price everything using original prices: $3300[CPU] + $200[HD] + $1200[HD] + $100[OD] + $1600[Video] + $1000[Monitor] = $7,500 plus RAM, for which I was unable to find original prices.) I decided that the 2-drive (small SSD, bigger HD) configuration, in combination with an onboard Time Machine backup to a third drive is safest and best, so I'm going to pursue that. (You must not take the OS update it wants to give you, or you brick the machine. If he should happen to do that, using the proposed three-drive configuration he could alt-boot to the HD, and then restore the SSD from the TM backup. That should get everything back running with a minimum of fuss.)

I took half the RAM out and put it in MacPro2. I took the 4 2GB RAM sticks out of MacPro2 and put them in MPServer, and took the 2GB (2 half-good 2GB sticks) from MPServer and put them in this new machine. The final RAM yields are: 32GB (up from 24) in MacPro2, 8GB (up from 2) in MPServer, and 18GB (down from 32) in Steve's.

I ordered a drive sled adapter (SSD) for $16, and I'm looking for a gently used SSD of modest size and price. One of Silverwood's cast-off 2TB drives should do for the Time Machine backup.

Monday, March 9, 2020

I managed to upgrade the firmware in Steve's MacPro to 2,1 today, with extreme difficulty. For whatever reason it refused to actually do the job cleanly, everything worked 'normally' except that it didn't actually burn firmware. I flailed around quite a bit, including downgrading the video card (borrowing from MPServer) and the OS (borrowing the VMLion disk from MacPro2), none of which worked. I was able to build the modified RamDisk under El Capitan, but under Lion it was unable to download what it needed from the network. Under ElCap I used dd to copy the disk image it had built to a file on the Lion disk. From Lion I mounted that image, which looked like Apple's original firmware updater, and ran some modified instructions that told how to do the procedure somewhat manually. After rebooting that one it finally upgraded the firmware. I do not know exactly what steps were necessary, versus the extraneous flailing I did. Regardless, in the end it worked. This computer could now accept an 8-core CPU array, if anybody ever cared to do so.

During this I noticed that the RAM bus speed was 553 MHz, not 667. This was due to the presence of the half-good 2GB sticks, which I promptly removed and put back in MPServer. 16GB is enough, and gaining bus speed on a general purpose computer is far more significant than the loss of 2GB of RAM.

I ordered 3 expansion slot covers, to plug the mouse doorways on the machine; $10. I'm picking up a 250GB SSD ($20) today, too. When I'm done with it this machine will essentially be exactly what I'm using (MacPro2) today, happily, except with only 4 cores, not 8. (This should not matter to Steve, and barely does to me; only photo processing sessions remotely tax the machine, and that is a rare occurrance.)

Tuesday, March 10, 2020

The MacPro 4/5,1 showed up yesterday. Today I fired it up, and all seems well. It has Firmware, which is the latest stuff available. It's missing drive sleds #1 and #4, and the monitor's firewire cable (lacking an adapter) can't connect anywhere; it could also use two more 8GB RAM sticks (one on each CPU) for maximum performance. (There are three RAM channels on these CPU's, but four slots each; as delivered only two each were occupied.) I ordered all of these items on eBay, spending just shy of $100. None of this, however, should keep it from being deployed, today.

A very nice compendium of Mac Pro upgrades is here.

I managed to damage the OS while trying to re-map the user accounts, so most of the day was spent re-loading fresh High Sierras onto both drives.

Wednesday, March 11, 2020

A little experimentation showed that the low-end Nvidia GT 120 (from the 5,1) works in the 2,1 but does not show anything until the OS is fully up. (Apparently it has only EFI-64 firmware in it, and not the requisite EFI-32 firmware. Or, preferably, both.) This would not make it a good choice for a lone machine that has to support itself. The ATI HD 5770 (from the 2,1) works nicely in the 5,1 of course, and works with the boot drive selector. The 5770 looks like it is very much the superior card, but both machines need to support the boot screen, so I can't off the lower-performing GT 120 on Steve. The GT 120 runs Unigen's Valley demo (quality set on Low) at a dreadful 3 FPS frame rate on the 30" monitor. Steve is more likely to game than I, so keep the GT 120 I shall.

Of continuing concern is the tremendous delay the 5,1 has coming out of sleep. (That's what prompted the video card experiment in the first place, and putting bootable OS's on both drives.) My existing 2,1 is notably fast at this, and part of what drove my decision to upgrade it rather than move over entirely to the laptop instead. OTOH, if I'm using the 5,1 all day it won't be going to sleep that much. (Unlike today's usage pattern, where the 2,1 is mostly only for personal use.) Still, this'll be quite the irritant if it can't be alleviated.

I got MacPro3 installed in place of MacPro2, and all software and files transferred over. After all this I find that the Nehalem CPU's in this thing, Intel X5570's, are not supported by VMWare Fusion 11, the whole point of the exercise; I guess I will have to upgrade from an 8- to a 12-core system. (At least I haven't seemed to have lost anything, except money if this doesn't eventually work out.) The CPU's in MacPro3 are newer than what is in MacPro2, as the CPUs' 'new' hyperthreading feature results in 16 CPU usage graphs (in an 8-core system) in the Activity Monitor, though not new enough for VMWare. So, earlier than I had intended, I ordered 2× Intel Westmere-EP X5690 CPU's; $148 from eBay.


I drove out and picked up the $20 250GB SSD, and slipped it into a SATA connector in Steve's machine. (Because it's so small and light I don't need a drive sled for temporary use, so long as I can lie the machine on its side.) I used the machine's target disk mode and Daniel's laptop in recovery mode to 'restore' Jill's machine onto the SSD, which was the easiest way to get the patched but clean El Capitan OS onto it. (Three-machine hat trick?) This operation brought a goodly number of applications with it. I still have to do some tidying up of accounts, etc., but this can wait until I have the full planned complement of drives and sleds.

Thursday, March 12, 2020

So far I've noticed two things about MacPro3 that aren't right:
  1. The wakup time from sleep is incredibly slow, on the order of 35 seconds or more—by far the longest of any of our machines;
  2. The CTRL-UP key sequence does not bring up the Mission Control screen, unlike on the laptop. Both machines are running High Sierra.
A little investigation showed that the latter is curable just by re-binding the MC keys in the System Preferences dialog, to match the settings of the laptop; problem solved.

Further investigation of the former shows, in the logs (reformatted):

$ pmset -g log | fgrep "Delays to Wake"
2020-03-12 08:41:22 -0700 Kernel Client Acks  	Delays to Wake notifications: [NVDATesla driver is slow(msg: SetState to 1)(370 ms)]
                                                                              [en0 driver is slow(msg: DidChangeState to 1)(421 ms)] 
                                                                              [en1 driver is slow(msg: DidChangeState to 1)(421 ms)] 
                                                                              [PRT4 driver is slow(msg: SetState to 2)(738 ms)] 
                                                                              [PRT0 driver is slow(msg: SetState to 2)(1211 ms)] 
                                                                              [PRT2 driver is slow(msg: SetState to 2)(1211 ms)] 
                                                                              [PRT5 driver is slow(msg: SetState to 2)(1211 ms)] 
                                                                              [IOSCSIPeripheralDeviceType05 driver is slow(msg: SetState to 4)(1254 ms)] 
                                                                              [PRT3 driver is slow(msg: SetState to 2)(11776 ms)]           
2020-03-12 08:41:37 -0700 Kernel Client Acks  	Delays to Wake notifications: [AppleHDADriver driver is slow(msg: SetState to 1)(1481 ms)]           
Summing the reported delays adds up to a suspiciously large amount of time—roughly 16 seconds in the PRTN drivers alone, whatever they are.

And now for something completely different: A little bit of research shows that DxO appears to support the MacPro3 machine, albeit at the very trailing edge of their current system requirements. So, I sprung for it: $199 for the Elite edition of PhotoLab 3. (I've been using a pirated down-rev copy for years because they simply would not sell me a copy that actually ran on the hardware I had.) After downloading it and applying the activation code it appears to run. Unfortunately, most of my camera/lens combinations are on the old side, and are not directly supported. I put in some requests, though I doubt anything will happen. (Many of these combinations were supported by older versions of their software.)

For DxO's sake I put the GT 120 away and stole the HD 5770 video card from MacPro2, putting the flashed HD 4870 card back in it. This gets me 20–30 FPS frame rate (vs 2–3) on the 30" monitor, using Unigen's Valley demo as a benchmark.

I ordered a second SATA optical drive for MacPro, a used SATA DVD superdrive for about $12. Having two drives makes copying discs easier, should the need occur. The one already in MacPro3 claims to be Blu-ray capable, but duping those is not something I'm likely to do, so a cheaper DVD drive is just fine.

Friday, March 13, 2020

I tried running a trial version of VMware Fusion 10, which took some time and effort (including downgrading the hardware requirements [from #16 to #14] on the VM images, using VMware 11), but it turns out the Nehalem CPU's are also too old for Fusion 10. Fusion 8 is the newest thing that will run on MacPro3 unless it gets its CPU transplant, and I'm not going to pursue this temporary downgrade path any further.

The PCI cover plates and the SSD mounting adapter came today. I installed them in Steve's machine, and re-mapped the user accounts to be suitable for him. The machine can be option-booted to either drive, or CMD-R booted to the recovery partition on the SSD. (There is no recovery partition on the HDD, no doubt due to whatever the PO did to get El Capitan onto this system.) All the machine needs now is a drive for Time Machine.

Saturday, March 14, 2020

The 2TB backup disks are nearly full, they've been in service since 2014, and doing a flawless job. I need to add MacPro3 to the backup system, but there isn't enough room left, nor do I want to (yet) dump MacPro2's backups.

I checked online prices, and it looks like 3TB drives are the new bigger-than-2TB lowest-cost option, $40–50 each, though the price-per-gigabyte sweet spot is probably the 4TB drives at around $70 each. Still a little bit expensive considering I'd need to buy four new drives.

Which means I need to look into trimming down the existing backups somehow.

Sunday, March 15, 2020

Waking of MacPro3 is still very slow, taking about 15 seconds to get to where I can enter the password, and 30 more seconds to refresh the screen and allow activity.
$ pmset -g log | fgrep "Delays to Wake"
2020-03-15 06:15:35 -0700 Kernel Client Acks  	Delays to Wake notifications: [en1 driver is slow(msg: DidChangeState to 1)(420 ms)]
                                                                              [en0 driver is slow(msg: DidChangeState to 1)(420 ms)]
                                                                              [PRT4 driver is slow(msg: SetState to 2)(745 ms)]
                                                                              [PRT5 driver is slow(msg: SetState to 2)(1208 ms)]
                                                                              [PRT2 driver is slow(msg: SetState to 2)(1211 ms)]
                                                                              [PRT0 driver is slow(msg: SetState to 2)(1212 ms)]
                                                                              [PRT3 driver is slow(msg: SetState to 2)(11831 ms)]           
2020-03-15 06:15:59 -0700 Kernel Client Acks  	Delays to Wake notifications: [AppleHDAHDMI_DPDriver driver is slow(msg: SetState to 1)(402 ms)]
                                                                              [AppleHDADriver driver is slow(msg: SetState to 1)(1479 ms)]           

The backup system is nearly full, and I don't think I'm quite ready to spring for 4TB drives. 3TB drives just aren't that much bigger than 2TB, so I'm going to look into trimming the backup sets.

First up is the fact that MPServer's own local backup directory reads as 9TB used. On a 2TB RAID that's completely ludicrous. Because there's nothing critical (as in: can't be re-made from scratch) on MPServer's own disk, Step 1 is to turn off TM on it, entirely delete its backup, and start over. Trying a simplistic:

sudo rm -rf /Volumes/Backups/Backups.db
did not work, there were lots of errors. I then used "tmutil" to remove the backup sets one at a time:
cd /Volumes/Backups
for bs in Backups.backupdb/MPServer/*; do
    sudo tmutil delete $bs
rm -rf Backups.backupdb
This took a fair amount of time and almost worked, but the "Latest" link choked things, and at that point nothing I did could get rid of those last empty (?) directories. After a lot of flailing around, including booting the recovery system and booting into single-user, and trying various fsck, diskutil, and chgflags commands, all to no avail, I finally decided to just rename this crummy thing and ignore it thereafter. After the rename, though, I could delete it. Clearly something was stuck in its craw, and the renaming cleared it up.

Deleting the server's own backup made free space go from about 20GB to about 140GB. Trying to turn on the backup again, though, claimed to need even more space than that. WTF? I waded through the home directories, and found many leftovers, files which were in use (and backed up) on other, primary, machines. These got deleted, and after that and a bit more tidying up the backup only claimed to need 60GB. So, the net recovery was about 60GB. [In fact it only consumed about 34GB once it had finished.]

Not enough. The G5 Quad, which used to be Jill's primary machine (and is now Daniel's game machine; she's now using MacPro) has backups going back to mid-2014, which we do not need. Its backup set consumed some 232GB, and removing all of 2014–2017 and compacting:

cd /Volumes/Backups
hdiutil attach G5Q*.sparsebundle
cd /Volumes/Time*
for bs in Backups.backupdb/G5Q/201[4-7]*; do
    sudo tmutil delete $bs
cd -
hdiutil detach ../Time*
hdiutil compact G5Q*.sparsebundle
reclaimed 93GB. (Because of the hard links it's difficult to predict how much space each backup takes.) Now we're talking. I did the same thing to MBP2, and got back 57GB there.

With some 356GB of free space now claimed on the Backups RAID I enabled backups on MacPro3, the point of all this activity.

Monday, March 16, 2020

It looks like in a 4/5,1 Mac Pro you can slip a bare drive into Bay 1, having it lie on the PCI card guide; no need to wait for the sled to show up in the mail before I can get started. I put the spare 2TB drive in, and began a bulk copy from the portable Silverwood drive. It took all day.

...And, the RAM, drive sleds, firewire converter cable, and DVD drive all came today.

Tuesday, March 17, 2020

I installed the items from yesterday. Everything looks like it works. The hardest part was finding four more screws for the DVD superdrive. I found that the RAM was mis-installed, according to the charts. This could have resulted in slightly lower performance. (If all 8 memory module slots are not filled, two particular slots should be left empty; they were not.)

I broke down and ordered a Mac Pro 4,1 1-CPU card that's already got a 2.66GHz 6-core X5650 in it; $87.12 including shipping. This gives me something (that will run VMware) to use while I upgrade the dual CPU card with X5690's, and gives me something to fall back on if the upgrade process fails or completion is delayed, or in case of later hardware failure. I don't really want to have to fall back to MacPro2, I'm already spoiled and I'm not even close to finishing the upgrade.

Friday, March 20, 2020

I drove to the Park and picked up one of the surplus 2TB drives. Once home I put it in Steve's cheese grater, aimed Time Machine at it, and let it begin making backups. This machine is now ready to go.

It's possible that next year they'd have enough surplus 4TB drives that I could upgrade my backup server.

Saturday, March 21, 2020

The 'downgrade' (single 6-core) CPU card came, and I put it in MacPro3, along with 24GB of the RAM from the original (dual) CPU card. It's not even that much of a downgrade from the current (×2 X5570 SLBFX 2.9 GHz quad-core); though it's got two less CPU cores and they're 12% slower, it has hyperthreading and the RAM is 25% faster (though there's less total RAM available). It booted up fine, and VMware 11 now runs. (The upgraded dual CPU card should eat the downgrade's lunch, in the end, with twice the RAM, twice the cores, and, most significantly, a 29% faster CPU clock.)

Monday, March 23, 2020

The X5690 CPUs came today. I read up on the swap again, and watched a couple of videos. It might be wise for me to first buy a pair of slow processors (cheap) to practice on.

The problem with MPServer's not being able to go to sleep turned out to be the old Temperature Monitor package. I quit that and MPServer's sleep problems are gone. Because of this, and because it kept having problems losing connection to sensors and re-creating its windows, I also quit the newer Hardware Monitor (the equivalent tool for High Sierra) that I'd been running on MacPro3. If I'm ever curious about heat I can just run it on demand.

Tuesday, March 24, 2020

I bought an X5650 on eBay for $6, shipped. For practice! This is the same 'slow' (2.66 GHz) CPU that's currently in the 'downgrade' CPU tray.

For maximum confidence in this upgrade process I should test all three new CPU's before I start de-lidding. (I'd want to be able to return any faulty processors, and I couldn't if I'd already cut into it.) Supposedly the single-CPU tray uses lidded processors, so I'd need to use it as a test mule for all three new CPU's, before I even get started upgrading. This, of course, puts my current hardware at some risk. Still, fortune favors the bold...

Upgrade plan:

The problem with MacPro3's waking up in a timely fashion seems to be due to the (High Sierra) Hardware Monitor, which I was using to monitor temperatures. Looks like this is a bad-news tool all around, and should only be run when specifically looking for problems.

Friday, March 27, 2020

No, MacPro3 took 66 seconds to awaken this morning. WTF???

For the record, MacPro3 currently has 24GB in it and I've got it pretty well loaded up with processes. In use it generally feels snappy, and is largely fast enough at everything I've tried. According to the Activity Monitor it is compressing memory, and has used some swap space—both to the tune of about ½GB; the eventual target of 48GB RAM should be plenty, and there's always the easy less-performant 64GB option from there if I should need it. (Or even more, with significant expense for larger-than-8GB memory modules, all of which [at that level] must match in size.)

The laptop (MBP4) would periodically go into high-fan activity, especially when running Windows in a VM, but MacPro3 has so far never even breathed the slightest bit hard while working. The tower upgrade has been a lot of work and a fair amount of expense, but so far I'm liking the move off the laptop.

Monday, March 30, 2020

The $6 practice CPU showed up today. All I need is free time, and for the heat sink goop to arrive in order to begin surgery.

Wednesday, April 1, 2020

I took another stab at putting Snow Leopard (10.6) on VMware. This involves faking a 10.6 Server installation, so that VMware will accept it. Instructions (from https://tinyapps.org/docs/snow_leopard_guest_os.html), titled "Installing Snow Leopard into VMWare Fusion 8 on El Capitan", reproduced here to protect against bit rot:
  1. Insert Snow Leopard retail DVD (OS X 10.6.3) > Disk Utility > highlight the DVD drive (not the "Mac OS X Install DVD" volume) > File > New Image > Image from "Untitled" > select read/write format (though, according to VMWare, any format should work) > assign desired filename (e.g., SL.dmg) and location (e.g., Desktop) > Save
  2. Create a shadow file:
    hdiutil attach ~/Desktop/SL.dmg -shadow
  3. Copy SystemVersion.plist to ServerVersion.plist:
    cp /Volumes/Mac\ OS\ X\ Install\ DVD/System/Library/CoreServices/SystemVersion.plist /Volumes/Mac\ OS\ X\ Install\ DVD/System/Library/CoreServices/ServerVersion.plist
  4. Eject the volume:
    hdiutil eject /Volumes/Mac\ OS\ X\ Install\ DVD/
  5. Write the changes to a new image (e.g., SLserver.dmg):
    hdiutil convert -format UDRO -o ~/Desktop/SLserver.dmg ~/Desktop/SL.dmg -shadow
  6. VMWare Fusion 8 > File > New... > "Create a custom virtual machine" (dragging in the DMG misidentifies it as OS X 10.5) > Continue > Apple OS X > Mac OS X Server 10.6 > Continue > Create a new virtual disk > Continue > Customize Settings > assign filename (e.g., SnowLeopard) > Save > CD/DVD (SATA) > Choose a disc or disc image... > select SLserver.dmg > adjust CPU, RAM, hard disk size as desired while you're in settings > close 10.6 window
  7. Downgrade the VMware virtual CPU to one the early 10.6 software can accept. Right click SnowLeopard in list of virtual machines > Show in Finder > right click SnowLeopard.vmwarevm > Show Package Contents > add the following line to SnowLeopard.vmx (to mask the CPU since new CPUs are not supported by versions prior to 10.6.7):
    cpuid.1.eax = "0000:0000:0000:0001:0000:0110:1010:0101"
  8. Start SnowLeopard virtual machine > click desired language and press Return > Utilities > Terminal
  9. Make necessary directories and file:
    mkdir -p /Volumes/Macintosh\ HD/System/Library/CoreServices/ && touch /Volumes/Macintosh\ HD/System/Library/CoreServices/ServerVersion.plist
  10. Quit Terminal, install Snow Leopard, and boot to Desktop. Don't forget to ask for Rosetta and/or QuickTime if you want these.
  11. Before installing VMWare Tools, move ServerVersion.plist somewhere (e.g., Desktop):
    sudo mv /System/Library/CoreServices/ServerVersion.plist ~/Desktop/ServerVersion.plist
    (Otherwise you will receive a "VMWare Tools can't be installed on this disk. VMWare Tools requires Mac OS X version 10.5 or higher." error.)
  12. Click Virtual Machine > Install VMWare Tools > Install > double click Install VMWare Tools > Continue > Continue > Install > enter password > OK > Continue Installation > before clicking Restart, restore ServerVersion.plist:
    sudo mv ~/Desktop/ServerVersion.plist /System/Library/CoreServices/ServerVersion.plist
  13. Click Apple menu > Software Update... > install the single update that appears (iLife Support 9.0.4)
  14. Checking for updates again returns: "Your software is up to date. Software Update doesn't have any new software for your computer at this time." Move ServerVersion.plist again to resolve:
    sudo mv /System/Library/CoreServices/ServerVersion.plist ~/Desktop/ServerVersion.plist
  15. Checking for updates again now returns:
    1. Remote Desktop Client Update 3.5.4
    2. iTunes 10.6.3 (skip it - we'll install 11.4 after the OS X 10.6.8 combo update)
    3. Mac OS X Update Combined 10.6.8 version 1.1
    4. Airport Utility 5.6.1
  16. After installing the updates, but just before rebooting, restore ServerVersion.plist:
    sudo mv ~/Desktop/ServerVersion.plist /System/Library/CoreServices/ServerVersion.plist
  17. After rebooting, update ServerVersion.plist:
    sudo cp /System/Library/CoreServices/SystemVersion.plist /System/Library/CoreServices/ServerVersion.plist
  18. Check for and install final updates:
    1. Java for Mac OS X 10.6 Update 17
    2. Apple Software Installer Update 1.0
    3. Safari 5.1.10 for Snow Leopard
    4. Migration Assistant Update for Mac OS X Snow Leopard v1.1
    5. iTunes 11.4 for OS X 10.6
    6. Security Update 2013-004 (Snow Leopard)
    7. Mac App Store Update for OS X Snow Leopard
  19. Shut down the virtual machine.
  20. Remove:
    cpuid.1.eax = "0000:0000:0000:0001:0000:0110:1010:0101"
    from the vmx file.
  21. For sound:
    1. Download and install EnsoniqAudioPCI_v1.0.3_Common_Installer.pkg inside the Snow Leopard guest
    2. Shut down the VM
    3. Virtual Machines > Settings... > Add Device... > Sound Card > Add...
    4. Remove this line from SnowLeopard.vmx:
      sound.virtualDev = "hdaudio"
But, it didn't work. Step 8 never completed booting, and eventually showed a barred circle icon. The instructions are for VMware Fusion 8, not the VMware Fusion 11 that I am running, but that really shouldn't matter.

Thursday, April 2, 2020

More Googling suggested that the problem was the original image I'd made from the 10.6 boot DVD. By not carefully reading the instructions, I'd copied the image from the volume and not from the device; collective wisdom says that the partition and driver information thus missing is vital. Starting over, but using the correct source and following yesterday's instructions, more carefully, it installed and seemed to be happy. (I also used UDTO format, for CD/DVD masters, but that might not be important.)

It is vital that you never reboot the VM (the Guest OS) with the ServerVersion.plist not in place, else VMware will refuse to run it. I don't think there's a recovery plan if you should screw this up.

The 10.6.8 combo update attempt via Software Update fails every time, and takes a long time to do so. Dr. Google says that directly downloading the combo update .dmg works much better, so I did this. Yes, that did the trick. (You attach the .dmg as VMware's CD image and run it from there.)

I downloaded the Xcode 3.2.6 .dmg file from Apple and installed it in the 10.6 VM.

I seem to have a fully-functional, fully equipped 10.6 virtual system, for using Rosetta if I should ever need to. Finally!

Saturday, April 4, 2020

The heat sink paste came yesterday. There were no obstacles remaining in the way of the planned CPU upgrade surgery, so I began. One problem I had is that the captive nut for the outlier (fifth) heat sink screw on the single-CPU board broke off the PCB when I tried to put back the heat sink the first time. I'm fairly careful, it must have been weak somehow. It didn't look like it was actually required for mechanical strength in the area, nor was it holding anything against the heat sink for thermal purposes, so I mixed up some 5-minute epoxy and re-secured it to the board. Because it wasn't really necessary I did not tighten it until well after it had set up, after I'd done all the pre-surgery CPU testing.

I can also confirm, definitively, that the single-CPU Mac Pro 4,1 processor trays use normal CPUs, with integrated heat spreaders (a.k.a. protective lids). The single-CPU heat sink, though, is 50% larger than the heat sinks on the dual-CPU card. Supposedly this is the reason for using lidless CPU's on the dual: there simply wasn't room for two of the correct (larger) sinks and so the protective CPU lids had to go in order to improve the efficacy of the smaller sinks.

The first step was to test all of the new CPUs before surgery, which had to be done in the single-CPU card. (You can't return a defective CPU after you remove the lid, so you have to test them unmodified!) All passed, then I restored the card to its original state.

(At this point one could consider upgrading the single-CPU card to its ultimate extreme: the 4.4 GHz X5698, but that's not a good idea for several reasons:
  1. You can't find one, they were only ever made in limited quantities;
  2. Even if you could find one, they are extremely expensive, like $1,500! (Versus $20,000 new; what a deal.)
  3. They are only 2-core, and are intended for applications where single-process throughput is paramount, and/or for bragging rights. I don't have anything like that.
  4. They don't do any clock throttling, even when idle, and so run fairly hot all the time. These are for the cost-is-no-object crowd.
  5. Apparently it has already been tried, and does not work.
basically this is just a dumb idea all around.)

Next was to prove that the dual-CPU card would function with only one CPU in it. I removed CPU B, the forward one, and found it to be, as expected, an X5570 (SLBFX) device. Yes, the fans eventually spun up with one CPU missing, but otherwise the system was usable.

Now it starts getting harder.

The first lid removal, of the $6 X5650, went smoothly at first, but it turned bad. I used a razor blade to cut through the glue around the edge, which went well. I then worked four razor blades in around the edges, to be a heat shield and to provide some spring pressure to cause the lid to pop off when the heat was applied. I used some wide-opening needlenose vise grips to gently grip the body of the CPU at the edges, similarly to how you would hold the CPU by hand, so that I could freehand everything (while I was outside) for better heat control. I then aimed the lid towards the ground and applied heat with a propane torch until the lid popped off. There were two problems:

  1. As I heated the lid I heard a 'CRACK' sound, but nothing seemed amiss and so I continued. Later, after it was all over, I found that the edge of the CPU's PCB had chunked away under the vise grip's teeth, which was very Not Good. This is a multilayer PCB, and I could see where the chunk cut into the layers. The expansion from the heat, and possibly localized weakening because of the heat, might have caused the failure.

  2. When the lid popped loose the shock jumped the CPU out of the vise grips, and the whole hot mess hit the ground. Molten solder sloshed all over the innards of the CPU, and seems to have removed one of the auxiliary surface-mount devices. There's an empty site, anyway. I looked, in the solder and on the ground, but I never found the device. There's an outside chance that it was never there, but I suppose I'll never know. There was also a chewed-looking corner of the big chip's die, where the thing had hit the ground face-first, hard. Also very Not Good.
I cleaned up the mess as well as I could, using a razor blade to remove the solder from the die, and trial-installed the CPU anyway. You must re-use the black plastic spacer/guide from the original CPU, making sure to insert it the correct way. Next, test it.

Of course it did not work; there goes $6.

Moving on, I took up one of the X5690's and de-lidded it. Learning from the MCF of the first one I used a small wooden screw clamp to grip the CPU's PCB by the edges, to avoid point stresses. I wrapped the jaws in aluminum foil to protect the wood from the heat. I then de-lidded the CPU using the torch. When the solder lets loose the noise and cascade to the ground of the (now) hot spring-loaded lid, molten metal, and hot razor blades is quite disturbing. (Watch your feet!) The lid came off cleanly, and cleanup of the solder remaining on the die was easy. (I used one of the razor blades, once it cooled a bit.) I installed the CPU and fired up the system.

It worked! All 6 cores showed up, and all 3 RAM modules. (There's no point in putting CPU B's RAM in yet as this is NUMA, and without the second CPU the second set of RAM modules isn't connected to anything.) The system identified at the expected speed: 3.46 GHz. With a missing second CPU, eventually the fans went on high alert, but otherwise the system seemed OK.

I then de-lidded the second CPU, using exactly the same technique as before. It also went smoothly, nothing abby-normal to report. I installed CPU B, moved the RAM back to the dual-CPU card, and fired it up.

It still worked! All 12 cores (24 hyperthreads) showed up, and all 6 (48GB) RAM modules. The system identified at the expected speeds: 3.46 GHz (CPU) and 1,333 MHz (RAM). It took a lot longer to 'bong' at boot due to the large amount of RAM, which was a bit alarming at first. This system is now as fast and capable as it's ever going to get.

It turns out that buying the $6 practice CPU was a very good idea. Making identifiable mistakes on that one was the key to a successful operation at the projected cost.

Return to Site Home