Foo

[mdlbear]

The fileserver froze again last night -- this is getting old. I remain convinced that it's some kind of driver race condition: it's not happening with the near-identical drive on the recording box. It could also be due to lurking disk corruption causing it to hang on a read. In that case, a fsck might fix it. It might not.

In either case, I'm tired of fighting it. I'm dropping the fileserver back to a single IDE drive, and moving the (SATA) mirror to another machine. Hopefully by the time I need it, SATA support will be better.

Meanwhile, I have lots of other work to do, and I don't want to have to deal with random fileserver hangs on top of it all.

10:48 It's even having trouble staying up long enough to copy a partition to the new drive. So here's the plan:

1. reboot single user, fsck the mirror drive again, and do an rsync to it. It was recently fsck'ed, but could still have become corrupt if the system hung during a write. I'm pretty sure that's what happened last time.
2. Install the new drive, and install a rescue image on the small partition It's worthwhile trying to make this Lenny, if I can find a suitable installer.
3. Copy the saved partitions off the mirror. Use Ubuntu if Lenny doesn't stay up long enough.
4. Edit fstab and make sure it boots.
5. uninstall the SATA drives, and reboot.

If Lenny's kernel is still flaky with the SATA drives it may be necessary to use Ubuntu for the copy. That would be ok. If it does work I may just leave the machine at Lenny, although it would be simpler to upgrade the fully-configured system in place.

13:00 here's an alternative plan:

1. reboot single user, hoping to keep it up long enough to rsync to the mirror. This has worked many times in the past; it seems to be long writes that are most likely to kill it. Just to be sure, remount / and home read-only. (noatime would probably work, but why mess around?)
2. install the new drive and uninstall the main SATA drive. Keep the mirror.
3. install a rescue copy of Lenny.
4. mount the mirror read-only, and copy / and /home. Update grub and fstab, and reboot. Things should be usable at this point, but we still need the big partition. We still don't fully trust SATA, though.
5. move the mirror to Dorsai, which has been working well so far with a SATA drive, and mount it read-only. Use rsync to copy anything that we couldn't copy in the previous step. If it freezes, we should be able to reboot and keep going, without locking up the fileserver.