Re: Possible bug with shared memory buffers

I believe I am seeing a similar problem in 7.2b4 under Linux.
Large and complex queries work fine after the postmaster is
first started up.  The same queries fail days/weeks later.

Typically psql unexpectedly disconnects while running the
queries.  Restarting psql will not clear the problem, but
stopping/restarting the postmaster will.

Mark Rae wrote:


>
> Case 2: This is with 7.2b4 on Linux
> I created one database as before and everything works as expected,
> however I then created a second database without dropping the original,
> and it also seemed to be working as expected.
> However as it grew in size, it slowed down significantly (20-30 times slower),
> as though it could not get enough memory for the indexes (The largest of
> which can be 0.5G-1G in size).
> Although this time the backend process appeared to have the whole 512M
> 'swapped in'. Again restarting postgres cleared the problem.
>
>
> Unfortunately the problem does not appear to be reliably repeatable,
> so I am just guessing that the problem might be related to the SHM buffers
> being 'leaked' or becoming permanently assigned to a specific
> database, even when they are no longer needed.


--
P. J. "Josh" Rovero                                 Sonalysts, Inc.
Email: rovero@sonalysts.com    www.sonalysts.com    215 Parkway North
Work: (860)326-3671 or 442-4355                     Waterford CT 06385
***********************************************************************