Marcin <migor@op.pl> writes:
> Yesterday morning one of our PostgreSQL engines stopped unexpectedly,
> [ with SIGSEGV at ]
> (gdb) bt
> #0 0x081b4e79 in GetRawDatabaseInfo ()
> #1 0x081b4865 in InitPostgres ()
> #2 0x08154741 in PostgresMain ()
This suggests corrupted data in pg_database, probably a bad page header
or a bad item offset.
> The server is running for nearly two years now, with PostgreSQL for all
> the time, and as it's the first problem I encountered, I don't suspect
> it's a hardware fault.
I do ;-). It would be interesting though to examine pg_database with
pg_filedump or a similar tool. See
http://sources.redhat.com/rhdb/utilities.html
for pg_filedump. Also, you might want to update to something newer than
7.4.1 --- I see at least one potentially relevant bug fix in the 7.4
branch CVS history.
regards, tom lane