Re: NOTICE messages during table drop

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: kurt_miller(at)sfgh(dot)org
Cc: pgsql-general(at)postgresql(dot)org
Subject: Re: NOTICE messages during table drop
Date: 2000-06-29 16:09:10
Message-ID: 721.962294950@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

"kurt miller" <miller_kurt_e(at)hotmail(dot)com> writes:
> Found these messages in the log this morning.
> Can anyone explain why?

> NOTICE: RegisterSharedInvalid: SI buffer overflow
> NOTICE: InvalidateSharedInvalid: cache state reset

Probably this indicates that you had another backend somewhere that
had been sitting in an open transaction for a long time and therefore
was not reading its SI "mail" about system table changes. Eventually
the SI message buffer overflows and the above notices result.

7.0 recovers cleanly from an SI overflow, so the notices should be
pretty harmless (and in a release or two they'll probably be removed,
or at least downgraded to DEBUG level so they don't appear by default).
But in prior releases this notice was often a harbinger of impending
doom :-(, because the cache-reset code didn't really work reliably.

If you see a *lot* of these during normal operations, you might have
reason to be concerned about the performance lost due to all the cache
flushes --- everybody pays for one backend's slowness when this happens.
In that case it'd be worth figuring out why your clients are leaving
backends sitting idle for long periods within open transaction blocks,
and trying to avoid that. But an occasional SI overrun is normal and
nothing to worry about ... at least not in 7.x.

regards, tom lane

In response to

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Bob Parkinson 2000-06-29 16:24:52 pg_dump/load quoting.
Previous Message Tom Lane 2000-06-29 15:03:11 Re: disk backups