Good morning. During a maintainance, I've changed (20 hours ago) the reserved
block count in my servers' data partition to 0. In some hours, I've got a number
of errors (5) in ext3 in different servers that are not possible by simple luck.
The errors are:
=========================================================
EXT3-fs error (device md3): ext3_new_block: Allocating block in system zone - blocks from 8192000, length 1
Aborting journal on device md3.
Remounting filesystem read-only
EXT3-fs error (device md3): ext3_free_blocks: Freeing blocks in system zones - Block = 8192000, count = 1
EXT3-fs error (device md3) in ext3_free_blocks_sb: Journal has aborted
__journal_remove_journal_head: freeing b_committed_data
__journal_remove_journal_head: freeing b_committed_data
...
=========================================================
or
=========================================================
EXT3-fs error (device md3): ext3_free_blocks_sb: bit already cleared for block 22479363
Aborting journal on device md3.
Remounting filesystem read-only
EXT3-fs error (device md3): ext3_free_blocks_sb: bit already cleared for block 22479364
EXT3-fs error (device md3): ext3_free_blocks_sb: bit already cleared for block 22479365
EXT3-fs error (device md3): ext3_free_blocks_sb: bit already cleared for block 22479367
EXT3-fs error (device md3) in ext3_free_blocks_sb: Journal has aborted
EXT3-fs error (device md3) in ext3_free_blocks_sb: Journal has aborted
EXT3-fs error (device md3) in ext3_free_blocks_sb: Journal has aborted
...
=========================================================
I have some servers with a single SAS disc and others with a software RAID10
volume over 4 discs. So far, only the RAID10 volumes are showing errors.
Coincidence?
This is Debian etch, and 2.6.24.2. The command line we've run was:
tune2fs -r 0 partition
while MySQL was running. The e2fsprogs version is 1.39+1.40-WIP-2006.11.14+dfsg-2etch1.
Do you know of any problem with this setup? I've reviewed the e2fsprogs
changelog searching for something like this but it seems rather related to
RAID10+ext3 interaction.
We have lots of servers waiting for crashing, I suspect. While I was writing
this mail another one crashed. Probably I can provide superblock copies for
analysis or any other information.
Best regards,
Ender.
--
I once farted on the set of Blue Lagoon.
-- Brooke Shields (South Park).
--
Responsable de sistemas
tuenti.com