Обсуждение: Vacuum process idle but hogging memory 8.2.4

Hello list,

System is running linux kernel 2.6.18 with postgres 8.2.4 and 1GB ram.

I'm having a 50GB database with the biggest table taking about 30 GB
and has about 200 million rows.

I'm already started to redesign the database to avoid the hugh number
of rows in this big table but I'm still curious why autovacuum hogs
over 200MB when it is not running?

Is it the shared_buffers?


Thanks,
Henke

shared_buffers = 128MB
work_mem = 10MB
maintenance_work_mem = 64MB

vacuum_cost_delay = 0            # 0-1000 milliseconds
vacuum_cost_limit = 200        # 0-10000 credits

effective_cache_size = 256MB

autovacuum_vacuum_cost_delay = 50
autovacuum_vacuum_cost_limit = 150

Henrik <henke@mac.se> writes:
> I'm already started to redesign the database to avoid the hugh number
> of rows in this big table but I'm still curious why autovacuum hogs
> over 200MB when it is not running?

On what do you base that assertion?

> Is it the shared_buffers?

Well, 128M in shared buffers plus 64M maintenance_work_mem would go
a long way towards explaining a 200M process address space, but it's
hardly "hogging" the shared buffers.

            regards, tom lane