ACCOZZAGLIA

This morning finds me doing what I’ve been trying to accomplish now for over two weeks: the revival and stabilization of my pesky, annoying server. The server, the slowest Apple Macintosh G4 ever made (n.b., for you geek types, a G4/350 PCI Yikes!, aka, “G3 in drag”), is trying to run Panther Server.

About three Saturdays ago (or 16 days, depending on how you count), I shut down and restarted the server to perform some sceheduled maintenance. Upon restart, though, the WindowServer and loginwindow started going completely mad. At the time, I’d been running 10.3.6 for about a week and change with only a couple of minor annoyances. This was only six weeks after a previous OS reinstall after running semi-smoothly for about eight months, give or take. Then again, all restarts with my system have been a crapshoot always, so.

But for each time I tried to isolate the problem, it only got more confusing. Even setting aside the existing OS and reinstalling a clean install didn’t help. In fact, half the time, I couldn’t keep the system from restarting itself from a CD install boot.

Eventually, I managed a reinstall, and during early last week, devised a system strategy whereby the /Users and /Applications directories at root level were ditto-ed to different volumes and symlinked (while logged in as single-user root). I did this, not only to mimic traditional UNIX/Linux installations, but in the event of system reinstalls (hi there), I could reinstall the system without losing user and application data.

I finished the ditto -rsrcFork process for these remotely through an ssh session — with my laptop, at a restaurant equipped with wi-fi, located in the suburbs, all while sitting with shimmerydeath and grazing on red, white and green-coloured tortilla-chip nachos. I felt mad accomplishment at being able to do this to a computer (my “POS computer”, as I’ve come to regard it) that sits being a robust firewall — remotely. No, I’m not sysadmin material, but it seemed to impress chitah, who has assumed the accidental rôle of a *NIX mentor for me. He assures me that I’m not dumb, despite all reason that I’ve called myself such for many, many years.

So, fast forward to the past weekend. I reinstalled the system with 10.3, then tried a combo upgrade to 10.3.7. It wouldn’t restart, rife with kernel panics galore every last time. So, all my memory — four 256MB SDRAM modules — were tested with a command-line util called memtest, aptly enough. Turns out that two of the modules, including the OEM 256MB that shipped with the system five years ago, tested with errors. Not many, but more than zero is enough for me. The OEM module, however, was less plagued than a third-party generic module (which was positively nightmarish with scrolling pages of memory address errors), but I had to remove the two of them and let the system operate on the two remaining modules that tested cleanly.

So, then I tried another install with a borrowed CD set for 10.3.4. This booted up okay, but upon upgrading to 10.3.5 (the last known stable update the server used), using the delta updater (versus the combo), the system started having dyld (link) errors, crashing the WindowServer and other functions and preventing a full boot process.

Fortunately, upon the last reinstall of Panther Server, I made a backup of the system on a separate partition, making restoration less nightmarish than full CD reinstalls. Ugh. It’s what I had to do last night to get my system launched.

As we speak, it’s updating with the 10.3.5 combo, but I expect it not to work. Why? Because the system install yesterday happened with all but one memory module removed. The remaining was the original OEM module, which later proved somewhat faulty with memtest. Which makes me think that the read/write from CD install to disk, which must go through the memory, is probably corrupt from the get-go.

Which means, if my hypothesis is correct, then this update in session, from 10.3.4 to 10.3.5, as I write this journal entry, will fail, and I will have to reinstall the OS again with the two remaining clean memory modules in place. And if the hypothesis, carried out to its natural ends is correct, then that installation will be stable, and all updates will prove to be smooth and unpained as these last several days, weeks, whatever have been.

I absolutely detest having to lose time on this. It reminds me of a certain lemon of a car (a 1984 VW Rabbit GTI in white) I once drove in the very early 1990s, one which taught me everything I know about working on cars (CV joints, alternators, water pumps, transmissions, electrical problems, changing tires — giving me all the tomboy cred I’m ever going to need, let me just say). I’m grateful that the last car I owned, in five years and 125,000km of steady operation, never once broke down.

There is a music library waaaay overdue to be placed on eBay, and I kind of need to cover my bills for January. Which aren’t much (roughly $200, if you can believe that), but when you’re as impoverished as I am, every cent counts. Plus, I’m now looking into therapy to work through the causations behind my depression, so even with sliding-scale, amounts like $10-25 a session is still a tremendous amount of money for me.

As it is, I’ll close out 2004 with an pre-AGI of $500. I don’t want 2005 to be another 2004 or 2003 (when the AGI was $2,937), financially speaking. But that means I have to handle the trauma — the actual, professionally diagnosed PTSD, mind — behind me in the workplace (namely, dealing with interviews and HR-types) before I can even hope to make a cent working for somebody.

//end ramble

//begin addendum update

The system updated to 10.3.5, and it restarted cleanly enough that I have a desktop. I don’t get it. At all. Time to go through the latest CrashReporter logs, I suppose, and see if any new errors were spawned upon this last reinstall. ::sigh::

//end addendum update and coming out of the closet with no holds barred geekness