From: | Merlin Moncure <mmoncure(at)gmail(dot)com> |
---|---|
To: | Hans Braxmeier <hans(dot)braxmeier(at)outlook(dot)com> |
Cc: | "pgsql-performance(at)postgresql(dot)org" <pgsql-performance(at)postgresql(dot)org> |
Subject: | Re: Postgres uses too much RAM |
Date: | 2017-05-08 23:12:12 |
Message-ID: | CAHyXU0wq-hh-CWWFsA6g3c0OBLw2h3pxmwWsDe+GbGH6mhaqcA@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On Mon, May 8, 2017 at 3:23 PM, Hans Braxmeier
<hans(dot)braxmeier(at)outlook(dot)com> wrote:
> Hello,
>
>
> on our old server (120 GB RAM) PostgreSQL 9.4.5 was using less than 10 GB of
> ram. On our new server (same system) Postgres 9.4.11 is using up to 40 GB
> Ram. Especially each idle process is consuming 2.4 GB: postgres 30764 8.3
> 2.4 3358400 3215920 ? Ss 21:58 0:24 postgres: testuser testdb [local] idle
>
>
> Summing up PG needs currently 72.14GB (also the slab_cache was increasing
> from 20 GB to 40 GB. ). For monitoring we are using munin. Is this a bug of
> 9.4.11 or what could be wrong?
can you paste unredacted snippet from, say, 'top'? A common
measuring error is to assume that shared memory usage is specific
cumulative to each process rather than from a shared pool. It's hard
to say either way from your info above.
If you do have extremely high resident memory usage, culprits might be:
*) bona fide memory leak (although this is rare)
*) bloat in the cache context (relcache, plancache, etc), especially
if you have huge numbers of tables. workaround is to recycle
processes occasionally and/or use pgbouncer
*) 3rd party package attached to the postgres process (say, pl/java).
merlin
From | Date | Subject | |
---|---|---|---|
Next Message | Scott Marlowe | 2017-05-08 23:35:38 | Re: Speed differences between two servers |
Previous Message | Scott Marlowe | 2017-05-08 23:06:05 | Re: Speed differences between two servers |