Re: Autovacuum seems to block database: WARNING worker took too long to start

Excerpts from Pablo Delgado Díaz-Pache's message of jue nov 18 08:57:16 -0300 2010:
> Well, we had the problem again. This is what we did
>
> 1) A "ps" to check what processes were running. Many "postmaster" processes
> in defunct state. A few postgres connections still working. A few "psql" (by
> shell script) queries hanged (they are scripts to monitor state of database.
> We use big-sister to do so).
>
> The normal state of the server is 1 postmaster pid and many "postgres"
> connections in "idle" state.
>
> This the result when we were having the problem. I've excluded irrelevant
> processes.
>
> *postgres   892  3889  0 09:07 ?        00:01:05 postgres: postgres
> international 10.19.0.51(49264) idle*
> *postgres   934  3889  0 12:00 ?        00:00:04 [postmaster] <defunct>*
> *postgres   935  3889  0 12:00 ?        00:00:04 [postmaster] <defunct>*

Hmm, these are processes that were forked from postmaster (PID 3889) and
died before they could become something useful -- but postmaster hasn't
reclaimed yet.

Can you see what process 3889 is doing?  strace it or get a stacktrace
with GDB.

--
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support