[Linux-cluster] stuck processes on GFS partition?

[Matt Brookover]

We are getting processes stuck in device waits on one file system. 
These errors are logged in /var/log/messages:

Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: stuck in
gfs_releasepage()...
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: blkno =
12446334, bh->b_count = 9
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
bh->b_journal_head = !NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: gl = (4,
12477424)
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
bd_new_le.le_trans = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
bd_incore_le.le_trans = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: bd_frozen =
NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: bd_pinned =
0
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: bd_ail_tr =
NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: ip =
12477424/12477424
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: ip->i_count
= 1, ip->i_vnode = !NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[0] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[1] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[2] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[3] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[4] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[5] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[6] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[7] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[8] = NULL
Dec 12 10:04:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[9] = NULL
Dec 12 10:09:17 imagine su(pam_unix)[5104]: session closed for user root
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: stuck in
gfs_releasepage()...
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: blkno =
12446334, bh->b_count = 9
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
bh->b_journal_head = !NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: gl = (4,
12477424)
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
bd_new_le.le_trans = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
bd_incore_le.le_trans = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: bd_frozen =
NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: bd_pinned =
0
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: bd_ail_tr =
NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: ip =
12477424/12477424
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5: ip->i_count
= 1, ip->i_vnode = !NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[0] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[1] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[2] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[3] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[4] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[5] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[6] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[7] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[8] = NULL
Dec 12 10:14:02 imagine kernel: GFS: fsid=CSM_ACN:admin01.5:
ip->i_arch.i_cache[9] = NULL

The file system in question appears to work fine on the other nodes, I
unmounted it to be on the safe side.

This is redhat enterprise 3.6, kernel 2.4.21-37.ELsmp, GFS 6.0.2.27-0. 
GFS was built from the source.
There are 2 partitions in the admin pool, the second was added a week or
so ago.

I tried to unmount it, but the umount failed because of the processes
that are stuck in device waits.

Any ideas?

thank you

Matt
mbrookov at mines.edu


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20051212/d661677d/attachment.htm>