Re: Graceful handling of DB connectivity failures

From: Murray S. Kucherawy <msk_at_blackops.org>
Date: Fri, 18 Mar 2011 23:09:22 -0700 (PDT)

I tried it with your haproxy configuration, modified as described, and I
couldn't reproduce the behaviour you saw except when haproxy was down
completely. Across restarts, opendkim dealt with the situation properly
each time by reconnecting.

This is haproxy-1.4.13 and opendkim-2.3.0 on FreeBSD.

The "Connection timed out" is the most worrisome part of the report,
because it means opendkim had hung at that point somehow. If you can
reproduce that condition, a "gcore" followed by a "thread apply all bt"
from inside "gdb" would be really helpful.

-MSK
Received on Sat Mar 19 2011 - 06:09:40 PST