File system deadlock -- postgresql related?

[Matthew Seaman]

Hi,

We've a FreeBSD-9.2-RELEASE-p3 server running postgresql-9.3.4 which is
giving us some problems.

I know of the performance problems with postgresql-9.3 referenced here:
http://postgresql.1045698.n5.nabble.com/Perfomance-degradation-9-3-vs-9-2-for-FreeBSD-td5800835.html
and also on this list.

However, we're seeing occasional system freezes for periods of maybe 5
-- 10 minutes, with many log messages like so appearing:

Apr 28 16:19:57 db-17a kernel:
g_vfs_done():mfid1p1[WRITE(offset=1058001420288, length=32768)]error = 11
Apr 28 16:19:57 db-17a kernel:
g_vfs_done():mfid1p1[WRITE(offset=1058001649664, length=32768)]error = 11
Apr 28 16:19:57 db-17a kernel:
g_vfs_done():mfid1p1[WRITE(offset=1058002010112, length=32768)]error = 11

All I can google about that problem suggests the 'error 11' is a
deadlock that occurs primarily when using disk encryption or other
setups where disk IO would be more complex than normal.  But that is
certainly not the case here --- it's a UFS2 filesystem on a hardware
RAID10 using 15k RPM SAS drives.

Any clues?  Am I correct to conclude that the mmap related performance
problem is somehow related to the g_vfs_done() log messages?

	Cheers,

	Matthew



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 1029 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20140428/9886a244/attachment.sig>