Re: Scaling XLog insertion (was Re: Moving more work outside WALInsertLock)

On Mon, Mar 5, 2012 at 8:50 AM, Heikki Linnakangas
<heikki(dot)linnakangas(at)enterprisedb(dot)com> wrote:
>
> That particular issue would be very hard to hit in practice, so I don't know
> if this could explain the recovery failures that Jeff saw. I got the test
> script running (thanks for that Jeff!), but unfortunately have not seen any
> failures yet (aside from the issue with crossing xlogid boundary), with
> either this or the older version of the patch.
>
> Attached is a new version of the patch.

I've run patch v10 for 14109 cycles of crash and recovery, and there
were 8 assertion failures at "xlog.c", Line: 2106 during the
end-of-recovery checkpoint.

How many cycles have you run? Assuming the crashes follow a simple
binomial distribution with the frequency I see, you would have to run
for ~1230 cycles for a 50% chance of experiencing at least one, or for
~8120 cycles for a 99% chance of experiencing at least one.

I think Fujii's method if provoking this problem is more efficient
than mine, although I haven't tried it myself.

Dual Core AMD Opteron(tm) Processor 275
2.6.32.36-0.5-default #1 SMP 2011-04-14 10:12:31 +0200 x86_64 x86_64
x86_64 GNU/Linux

Cheers,

Jeff