[OPEN-ILS-GENERAL] EG 2.0 network errors

[John Morris]

Hi, having some wierd problems with random network errors.

Cut over to 2.0.10 Monday morning and things appeared to go well, but
then the random glitches started hitting.  Would occasionally get one of
these on 1.6.0.8 but rare enough folks just worked around it, now it is
really hampering getting things done.  Googled on the messages and found
some help in the mailing list archives for a problem that seems to match
ours in that the errors were the same and only seemed to happen in
virtual machines.  So I manually applied these patches to OpenSRF:

719b5344d5cc5a85bb506af0432ef0c563800fff

ab63faa85e349020774e0b3297a84370281d573f

Here is the link to the thread that seemed to apply:

http://list.georgialibraries.org/pipermail/open-ils-general/2011-October/005639.html

Today things are better but we are still getting hit pretty hard.  We
have massive overkill in our hardware provisioning, system load never
hits 1.0, 0.30 is more typical, even when a transaction that is about to
fail is in progress.  Have two physical hosts with a six core Opteron
and 16GB of RAM in each.  Each is currently inhabited by a single
libvirt/KVM based VM, one the front end and one running Postgresql.
Each VM has more than enough ram to prevent swapping and the DB should
be essentially running from ram, and anyway the discs are SSDs. There is
enough computing grunt here to easily handle the work.  The real and
virtual OS on all is Debian 6.0 with the 2.6.32 kernel backported to get
needed hardware support in the real machines.

So has anyone else seem this problem and what else can I try?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://libmail.georgialibraries.org/pipermail/open-ils-general/attachments/20111215/2b2362bc/attachment.pgp>