issues with ext3 on lvm2 on md raid1 on sarge
Hello,
I am seeing issues while using an ext3 filesystem on top of lvm2 on top
of a md raid1 device.
My system is running sarge with:
kernel-image-2.6.8-2-386 2.6.8-16sarge1
lvm2 2.01.04-5
mdadm 1.9.0-4sarge1
bonnie++ 1.03a
jfsutils 1.1.7-1
e2fsprogs 1.37-2sarge1
The raid1 device is composed of two 250GB SATA drives on a Silicon Image
3112 based controller.
I created the raid1 device using the following.
localhost:~# mdadm -C /dev/md0 --level=1 --raid-devices=2 /dev/sda1
/dev/sdb1
After waiting for the array to build, I created the lvm volumes and some
filesystems.
localhost:~# pvcreate /dev/md0
localhost:~# vgcreate raid1 /dev/md0
localhost:~# lvcreate -L75G -n ext3filesystem raid1
localhost:~# lvcreate -L25G -n jfsfilesystem raid1
localhost:~# mkfs.ext3 /dev/raid1/ext3filesystem
localhost:~# mkfs.jfs /dev/raid1/jfsfilesystem
localhost:~# mkdir /mnt/ext3filesystem
localhost:~# mkdir /mnt/jfsfilesystem
localhost:~# mount /dev/raid1/ext3filesystem /mnt/ext3filesystem
localhost:~# mount /dev/raid1/jfsfilesystem /mnt/jfsfilesystem
I then decided to run some performance tests on each of the filesystems
using bonnie++.
localhost:~# bonnie++ -d /mnt/jfsfilesystem/ -u 1000
localhost:~# bonnie++ -d /mnt/ext3filesystem/ -u 1000
The jfs filesystem completed the test without any problems, but
everytime I run the test on the ext3 filesystem the test fails midway
through when the filesystem is remounted read only. The following
errors show up in dmesg
EXT3-fs error (device dm-2): ext3_free_blocks: bit already cleared for
block 278821
Aborting journal on device dm-2.
EXT3-fs error (device dm-2): ext3_free_blocks: bit already cleared for
block 278822
ext3_abort called.
EXT3-fs abort (device dm-2): ext3_journal_start: Detected aborted journal
Remounting filesystem read-only
ext3_reserve_inode_write: aborting transaction: Journal has aborted in
__ext3_journal_get_write_access<2>EXT3-fs error (device dm-2) in
ext3_reserve_inode_write: Journal has aborted
EXT3-fs error (device dm-2) in ext3_truncate: Journal has aborted
ext3_reserve_inode_write: aborting transaction: Journal has aborted in
__ext3_journal_get_write_access<2>EXT3-fs error (device dm-2) in
ext3_reserve_inode_write: Journal has aborted
EXT3-fs error (device dm-2) in ext3_orphan_del: Journal has aborted
ext3_reserve_inode_write: aborting transaction: Journal has aborted in
__ext3_journal_get_write_access<2>EXT3-fs error (device dm-2) in
ext3_reserve_inode_write: Journal has aborted
EXT3-fs error (device dm-2) in ext3_delete_inode: Journal has aborted
I also ran bonnie++ on the / filesystem which is ext3 but does not use
md raid or lvm.
localhost:~# bonnie++ -d /tmp/ -u 1000
This completed without any problems.
While searching about the problem I came across the following. Debian
bug #295657 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=295657 is
very similar to what I am experiencing. The bug was closed after the
problem was determined to be an issue with defective RAM. I also came
across this redhat bug
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=152162 which was
determined to have been introduced in the kernel around 2.6.10
To rule out defective RAM I ran memtest86+ 1.65 on the machine
overnight. It successfully completed 24 passes without any errors in
about 12.5 hours, so I think that my RAM is good. It does not appear to
be a hard drive issue since the tests complete fine on the jfs filesystem.
Are there any other hardware tests that I should try or is this likely
to be a software bug?
Mike
Reply to: