Re: FreeSpaceMap hashtable out of memory - Mailing list pgsql-hackers

From Tom Lane
Subject Re: FreeSpaceMap hashtable out of memory
Date
Msg-id 10658.1065050640@sss.pgh.pa.us
Whole thread Raw
In response to Re: FreeSpaceMap hashtable out of memory  ("Maksim Likharev" <mlikharev@aurigin.com>)
List pgsql-hackers
"Maksim Likharev" <mlikharev@aurigin.com> writes:
> It seems like 
> WARNING:  ShmemAlloc: out of memory
> ERROR:    FreeSpaceMap hashtable out of memory
> goes together, does it related to the size of Shared Memory

Yeah, the FSM hashtable is in shared memory, so your problem is that
you're running out of shared memory.  This is not necessarily the fault
of the FSM as such though; it could be that some other shared data
structure is growing bigger than it was expected to.

Thinking about it, I'm fairly certain that the FSM can't grow larger
than the bounds you set for it, and so the problem is presumably
elsewhere.  The most likely bet is that the lock table is getting larger
than expected.  There is a control knob for the estimated size of the
lock table (max_locks_per_transaction), so if that's where the problem
is, it's easy to fix.  You should try to find out if that's the issue
though.  When this happens, are there a very large number of entries in
the pg_locks view?
        regards, tom lane


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: Thoughts on maintaining 7.3
Next
From: "Jingren Zhou"
Date:
Subject: PREPARE/EXECUTE across backends?