User-Agent: Microsoft-MacOutlook/14.2.3.120616
Date: Mon, 19 Aug 2013 12:37:16 +0300
Subject: Re: postgres crashes on insert in 40 different threads
From: Dzmitry <dzmitry.nikitsin@gmail.com>
To: Albe Laurenz <laurenz.albe@wien.gv.at>,
 =?ISO-8859-1?B?U3TpcGhhbmU=?= Schildknecht
 <stephane.schildknecht@postgresql.fr>,
 "pgsql-admin@postgresql.org" <pgsql-admin@postgresql.org>
Message-ID: <CE37C19E.116F8%dzmitry.nikitsin@gmail.com>
Thread-Topic: [ADMIN] postgres crashes on insert in 40 different threads
In-Reply-To: 
 <A737B7A37273E048B164557ADEF4A58B17C06C2D@ntex2010a.host.magwien.gv.at>
Mime-version: 1.0
Content-type: text/plain;
	charset="windows-1254"
Content-transfer-encoding: quoted-printable
Precedence: bulk
Sender: pgsql-admin-owner@postgresql.org

Do you mean postgres log file(in postgres.conf)

log_filename =3D 'postgresql-%Y-%m-%d_%H%M%S.log'
log_min_messages =3D warning

Or /var/log/messages ? Because I haven't this file :(


Thanks,
  Dzmitry


On 8/19/13 12:26 PM, "Albe Laurenz" <laurenz.albe@wien.gv.at> wrote:

>Dzmitry wrote:
>> On 8/19/13 11:36 AM, "St=E9phane Schildknecht"
>><stephane.schildknecht@postgresql.fr> wrote:
>>> Le 19/08/2013 10:07, Dzmitry a =E9crit :
>>>>  I have postgres server running on ubuntu 12,Intel Xeon 8 CPUs 29 GB
>>>>RAM.
>>>> With following settings:
>>>> max_connections =3D 550
>>>> shared_buffers =3D 12GB
>>>> temp_buffers =3D 8MB
>>>> max_prepared_transactions =3D 0
>>>> work_mem =3D 50MB
>>>> maintenance_work_mem =3D 1GB
>>>> fsync =3D on
>>>> wal_buffers =3D 16MB
>>>> commit_delay =3D 50
>>>> commit_siblings =3D 7
>>>> checkpoint_segments =3D 32
>>>> checkpoint_completion_target =3D 0.9
>>>> effective_cache_size =3D 22GB
>>>> autovacuum =3D on
>>>> autovacuum_vacuum_threshold =3D 1800
>>>> autovacuum_analyze_threshold =3D 900
>>>>
>>>> I am doing a lot of writes to DB in 40 different threads =AD so every
>>>>thread
>>>> check if record exists =AD if not =3D> insert record, if exists =3D> u=
pdate
>>>>record.
>>>> During this update, my disk IO almost always =AD 100% and sometimes it
>>>>crash my
>>>> DB with following message:
>>>>
>>>> 2013-08-19 03:18:00 UTC LOG:  checkpointer process (PID 28354) was
>>>>terminated by signal 9: Killed
>[...]
>>>>
>>>> My DB size is not very big =AD 169GB.
>>>>
>>>> Anyone know how can I get rid of DB crash  ?
>
>>> The fact that the checkpointer was killed -9 let me think the
>>>OOMKiller has
>>> detected you were out of memory.
>>>
>>> Could that be the case?
>>>
>>> 12GB of shared_buffers on a 29Gb box is too high. You should try to
>>>lower that
>>> value to 6GB, for instance.
>>> 550*50MB, that is 27GB of RAM that PostgreSQL could try to adress.
>>>
>>> I can imagine your system is swapping a lot, and you exhaust swap
>>>memory before crash.
>
>> I don't think it's the case. I am using newrelic for monitoring my DB
>> servers(I have one master and 2 slaves - all use the same
>>configuration) -
>> memory is not going above 12.5GB, so I have a good reserve, also I don't
>> see any swapping there :(
>
>You can check by examining /var/log/messages to see if the OOM
>killer is at fault, which is highly likely.
>
>The OOM killer uses heuristics, so it does the wrong thing occasionally.
>
>The documentation is helpful:
>http://www.postgresql.org/docs/9.2/static/kernel-resources.html#LINUX-MEMO
>RY-OVERCOMMIT
>
>Yours,
>Laurenz Albe


--=20
Sent via pgsql-admin mailing list (pgsql-admin@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-admin