Tom Lane wrote:
> Should we redesign the stats signaling logic to work around this,
> or just hope we can nag kernel people into fixing it?
>
Even if there was something to be done in kernel space, how many years
from now would it be before it made this problem go away for the
majority of near future 9.0 users? We've been seeing a fairly regular
stream of "pgstat wait timeout" reports come in. The one I reported was
from recent hardware and a very mainstream Linux setup. I'm not real
optimistic that this one can get punted toward the OS and get anything
done about it in time to head off problems in the field.
This particularly pathologic case with jaguar is great because it's made
it possible to nail down how to report the problem. I don't think it's
possible to make a strong conclusion about how to resolve this just from
that data though. What we probably need is for your additional logging
code to catch this again on some systems that are not so obviously
broken, to get a better idea what a normal (rather than extreme)
manifestation looks like. How much skew is showing up, whether those do
in fact correspond with the wait timeouts, that sort of thing.
--
Greg Smith 2ndQuadrant US Baltimore, MD
PostgreSQL Training, Services and Support
greg@2ndQuadrant.com www.2ndQuadrant.us