Re: Win32 Backend Cash - pre-existing shared memory block is still in use

Excerpts from Jeremy Palmer's message of mar ago 17 22:59:08 -0400 2010:
>
> I'm getting infrequent backend crashes on a windows instance of PostgreSQL. The error I get is in the log below. It seems to relate to the share memory each time. Does anyone have any ideas what the problem is here, or what additional things I can do to get more information out next time the backend crashes?
>
> I'm running PostgreSQL 8.4.4 on Windows server 2003 Standard.

The first entries are unrelated.

> 2010-08-17 16:57:22 NZSTLOG: unexpected EOF on client connection

This is a memory dump and could be unrelated (or maybe not).

> TopMemoryContext: 268428304 total in 26 blocks; 5528 free (22 chunks); 268422776 used
> Local Buffer Lookup Table: 2088960 total in 8 blocks; 501416 free (22 chunks); 1587544 used
[snip some more]

The real problem is here:

> 2010-08-17 19:19:32 NZSTLOG: server process (PID 6684) exited with exit code 128
> 2010-08-17 19:19:32 NZSTLOG: terminating any other active server processes

The rest is just noise about other processes being killed by postgres to
reset memory to a known-good state.

What you need to investigate is what caused that process to exit with
code 128.

PS: it'd be a good idea to append some whitespace to log_line_prefix, to
ensure that timezone NZST is separate from the "LOG" tag.

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