Re: pg_dump being killed by oom killer - Mailing list pgsql-admin

From Paul Warren
Subject Re: pg_dump being killed by oom killer
Date
Msg-id CALn4Y7GfsM+TUkyaYu3kx8vo2RUSEVzEoXWmG0zrOm4euucCPA@mail.gmail.com
Whole thread Raw
In response to Re: pg_dump being killed by oom killer  (JotaComm <jota.comm@gmail.com>)
List pgsql-admin
Yesterday I turned off the oom killer on the psychical server rather then on the VM the vm is running on promox which is a container so oom killer wasn't a option to turn off on the vm. When I did this it caused the server to run out of memory, I rebooted server and pg dump now doesn't get killed by oom killer but it now runs then uses most fo the ram up and then hangs on terminal and i ended up killing it with loads of pg_dump command not found errors




On Wed, Oct 30, 2013 at 12:31 PM, JotaComm <jota.comm@gmail.com> wrote:
Hello,


2013/10/29 Paul Warren <pw@researchresearch.com>
root@www1:/# ulimit -m
unlimited

And parameters: overcommit_memory and overcommit_ratio.


On Tue, Oct 29, 2013 at 1:06 PM, Sergey Klochkov <klochkov@iqbuzz.ru> wrote:
Hello Paul,

If so, what does ulimit -m show?


29.10.2013 17:04, Paul Warren пишет:
Hey,

I think it might not be a issue with pg_dump / db etc but more to do
with proxmox and it's container limits, this is a VM running on proxmox
and could be the cause of the issue. Will look into that first and come
back if it's not.


thanks
Paul


On Tue, Oct 29, 2013 at 1:01 PM, Sergey Klochkov <klochkov@iqbuzz.ru
<mailto:klochkov@iqbuzz.ru>> wrote:

    Hello Paul,

    Recently I've run into similar issue, see my message to this mailing
    list sent on October 1st, 2013. It happens if there is a large
    enough amount of rows in pg_largeobject of the relevant database.
    pg_dump loads metadata (owner, access rights, etc) of each large
    object deparately until it is out of memory. How many large objects
    are in your DB?

    Hope this helps.

    29.10.2013 16:17, Paul Warren пишет:

        Hi,

        I'm to the lists so bare with me please.

        I have a vm running debian 6 with linux kernel 2.6.32-11-pve,
        when i run
        pg_dump -Fc -U postgres  db name > db name.dump

        after so long the process is killed, when I look in the
        /var/log/kern.log i see the following entry
        Oct 29 12:07:14 www1 kernel: OOM killed process 0 (pg_dump)
        vm:5269904kB, rss:5210224kB, swap:0kB


        The vm has 16 cpu's and is using around 10%, when I run pg_dump
        it goes
        up by 3 / 5 % so not using alot of cpu usage. IT has 30gb of ram
        and 1gb
        of swap, the system uses around 2mb of swap and 15gb of ram, when
        pg_dump runs the swap doesn't move but the ram goes up by 3 / 4 gb
        before the process is killed.

        The machine had only 20gb of ram so I've increased the ram usage and
        still get the same issue. if I run pg_dump on any other db it works
        fine, if we run a pg_dump script by table on the failing database it
        works fine.

        The backup is going directly to a nas device which has 1.2 tb
        free. The
        vm has aroudn 40gb free as well.

        Any ideas what this could be? it seems to be a memory issue but
        I can't
        seem to see what / why it's happening.

        oh the DB is around 1.5 gb / 2gb in size


        Thanks
        Paul



    --
    Best regards, Klochkov Sergey
    www.iqbuzz.ru <http://www.iqbuzz.ru>




--
С уважением, Клочков Сергей
www.iqbuzz.ru



Best regards

pgsql-admin by date:

Previous
From: Raghu Ram
Date:
Subject: Re: Connecting to Oracle from Postgresql 9.2 on Windows
Next
From: Rune Gullstrom
Date:
Subject: Connecting to Oracle from Postgresql 9.2 on Windows