On 24 September 2010 21:24, Brad Nicholson <bnichols@ca.afilias.info> wrote:
>> The pertinent difference between pg_stat_user_indexes and
>> pg_statio_user_indexes is the latter shows the number of blocks read from
>> disk or found in the cache.
>
> I have a minor, but very important correction involving this point. The
> pg_statio tables show you what blocks are found in the Postgres buffer
> cache, and what ones are not.
Right. Then, studying how the pg_statio table develops over time
would probably give a hint on my first question in my original post
... how to check the hypothesis that we're running out of memory.
That said, I've sent an email to our sysadmin asking him to consider
the pg_buffercache module suggested by Greg Smith.
Increasing the shared_buffers on the cost of OS caches would then have
one "selling point" ... better possibilities to monitor the memory
usage.