"Greg Sabino Mullane" <greg@turnstep.com> writes:
>> Increase the statement_timeout value used in the prepared_xacts
>> regression test. We have seen some buildfarm failures that seem
>> to be due to this limit being unexpectedly exceeded when the machine
>> is under load.
> Just as a data point, I thought I saw fennec in the list of machines
> exhibiting this problem, and I can safely vouch that it extremely unlikely
> the machine was under load at the time.
Yeah, I'm not totally sold that this is a "fix" either --- although
there's at least one prior case of an intermittent buildfarm failure
that went away when we increased the timeout sufficiently.
I think a 2-line change is an appropriate amount of effort for now.
If we keep seeing the failure then we'll know we need to look closer.
regards, tom lane