Yesterday when I entered the server room at 1600 I thought I'd be out by
1800, with the intention of adding a Fujitsu MAM3367MP (15krpm 36G) to
a raid1 array and converting it to raid5.
I did the following:
- halt machine (dell Poweredge 1400 dual 1Ghz), add third drive
- reboot with root=/dev/sda1 instead of root=/dev/md0 to have 2 free
drives out of 3 to build a degraded raid5 array,
- create /dev/md0 (small raid1 boot partition), /dev/md1 (raid5 root),
/dev/md2 (raid5 swap) with /dev/sdb (former /dev/sda mirror) and
/dev/sdc (new identical drive),
- switch to runlevel 1, copy / (/dev/sda1) to /mnt (/dev/md1)
- adjust /mnt/etc/fstab to mount /dev/md1 as root and /dev/md2 as swap
- reboot with root=/dev/md1
- all is well, cat /proc/mdstat shows /dev/md[012] are working fine in
degraded mode,
- /dev/sda is partitioned identically to /dev/sd[bc]
- final steps:
- "mdadm /dev/md0 --add /dev/sda1" goes OK
- "mdadm /dev/md1 --add /dev/sda2" starts OK, says 15 minutes to go,
when at 25% rebuild suddenly it stops with "md_do_sync() caught
signal, exiting"
Now the machine is hung, I can switch virtual consoles but can't login.
No disk activity. I hit the reset button and on reboot /dev/md1 won't
start. I tried the multiple disk failure recovery from the howto without
success: no filesystem identified on the partition.
To make sure I retried all these steps three times with always the same
result: the array resync stops at 25%, the machine hangs and the array
won't come back on-line.
Is there a known issue with 2.4.19-smp and raid5? Did I hit a bug or is
the problem somewhere in my hardware.
The exact same procedure on a uniprocessor 2.4.19 with 3 IDE drives
worked fine.
I can provide more details if needed. I'd really like to understand what
happened there.
In the end I exited the server room around 0130 sunday morning after
recovering the machine from backup.
Cheers,
--
PHEDRE: Malheureuse ! Voilà comme tu m'as perdue.
Au jour que je fuyais c'est toi qui m'as rendue.
(Phèdre, J-B Racine, acte 4, scène 6)
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
