In response to:
> We have a Debian guest running on a Debian Host (linux-image-2.6.26-1-
> vserver-amd64, util-vserver 0.30.216~r2772-6) for our Postgres 8.4
> database. After a period of time, Postgres got killed
>
> [27795665.838308] VM: killing process postgres
>
> This is reproducable, if we open a few connections and execute a large
> query on each of them, sequential. The guests "top" (VIRT_MEM) showed
> us, that after an executed query on a connection (process), the
> available memory decreases with the amount of postgres' shmem setting
> ( "shared_buffer_size", shared memory) until swap got filled. The
> settings for Postgres are ok, it never comes to that situation on a
> _real_ server with an equal configuration (available Ram, equal
> database content, equal configuration and so on). What could lead to
> that wrong behaviour?
I had the same experience with kernel 2.6.22.x-vs2.2.x and pgsql 8.1 - shmem is counted several times.
Best regards,
-- Lukasz CzarnowskiReceived on Thu Oct 29 00:00:36 2009