Re: Shared memory usage in PostgreSQL 9.1

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Christoph Zwerschke <cito(at)online(dot)de>
Cc: pgsql-general(at)postgresql(dot)org
Subject: Re: Shared memory usage in PostgreSQL 9.1
Date: 2011-12-04 21:23:52
Message-ID: 17913.1323033832@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

Christoph Zwerschke <cito(at)online(dot)de> writes:
> Am 03.12.2011 20:31, schrieb Christoph Zwerschke:
>> Then, the corrected sum is 449627320 Bytes, which is only about 2MB less
>> than was requested. This remaining discrepancy can probably be explained
>> by additional overhead for a PostgreSQL 9.1 64bit server vs. a
>> PostgreSQL 8.3 32bit server for which the table was valid.

> And this additional overhead obviously is created per max_connections,
> not per shared_buffers. While the docs suggest there should be 19kB per
> connection, we measured about 45kB per connection. This explains the
> about 2MB difference when max_connections is 100.

I suspect most of the difference from 8.3 to 9.1 has to do with the
additional shared memory eaten by the predicate lock manager (for SSI).
That table really ought to get updated to include a factor for
max_pred_locks_per_transaction. (And I wonder why
max_locks_per_transaction and max_pred_locks_per_transaction aren't
documented as part of the "memory consumption" GUC group?)

regards, tom lane

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Andreas Karlsson 2011-12-05 00:16:54 Re: BUG #6325: Useless Index updates
Previous Message hubert depesz lubaczewski 2011-12-04 20:07:21 Re: BUG #6325: Useless Index updates