Alvaro Herrera wrote:
> Stefan Kaltenbrunner wrote:
>> I'm still getting random failures from some of my buildfarm members
>> which is starting to get a bit irritating and annoying :-(
>>
>> some recent failures:
>>
>> http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=zebra&dt=2007-02-06%2015:25:04
>> http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=clownfish&dt=2007-02-06%2000:03:04
>> http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=clownfish&dt=2007-02-04%2003:03:09
>> http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=impala&dt=2007-02-02%2003:03:02
>>
>>
>> any ideas ?
>
> Since they are sporadic, my guess is that it's due to autovacuum not
> letting pgstat catch up. I'd try either setting
> autovacuum_vacuum_cost_delay to a low value, or extending the sleep
> period in the stats test, to give more time for pgstat to catch up with
> those messages.
hmm now that I look closer - all those members above are actually VMs of
some sort and the host is rather busy at times (multiple VMs competing
for CPU and IO) so that might be a factor here.
>
> Setting the cost_delay sounds a reasonable thing to do anyway, and in
> fact I already proposed it and nobody objected (AFAIR). Now we only
> have to agree on a reasonable value.
note sure on a reasonable value but we still have time to test different
values if needed for 8.3 - but I think we should really try to get rid
of those sporadic failures because they might lead to getting other
issues going unnoticed.
Stefan
