Re: autovacuum out of memory errors

tim truman <mitramnut(at)gmail(dot)com> wrote:

> [ memory usage map ]

There doesn't seem to be any memory context using an unusually large amount of RAM.

>  522f9128.1151 ERROR:  out of memory
>  522f9128.1151 DETAIL:  Failed on request of size 336150396.
>  522f9128.1151 CONTEXT:  automatic vacuum of table "client.public.product"
>
> SELECT version();
> PostgreSQL 8.4.11 on x86_64-pc-linux-gnu, compiled by GCC gcc-4.4.real (Debian 4.4.5-8) 4.4.5, 64-bit

http://www.postgresql.org/support/versioning/

> maintenance_work_mem         | 1GB
> max_connections              | 1000

> shared_buffers               | 11GB

> wal_buffers                  | 8MB

> work_mem                     | 8MB

> free -m
>              total       used       free     shared    buffers     cached
> Mem:         48396      48139        257          0        173      44771
> -/+ buffers/cache:       3194      45201
> Swap:          956         20        936

Clearly the free command was not run while the database service was
running.  Can we see what that looks like when it is running and
handling roughly the same workload that has been causing problems
(including any non-database processes which have been running on
the machine during failures)?

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company