Re: repeatedly process termination in PG 8.2

On Thu, Sep 24, 2009 at 10:31 AM, Michael Molz
<michael(dot)molz(at)iwm-software(dot)de> wrote:
> one  of our systems on a PG 8.2 database crashes nearly daily; others are running fine. Host OS of this system is Windows, if the db crashes nothing is written in the OS event logs. The database log show every time the following messages:
>        2009-09-23 10:21:30 LOG:  server process (PID 1240) exited with exit code -1073741819
>        2009-09-23 10:21:30 LOG:  terminating any other active server processes
>        2009-09-23 10:21:30 WARNING:  terminating connection because of crash of another server process

Your symptoms point to one of two possibilities: 1) you have hardware
faults or 2) your DB on this one machine has a corruption that causes
postgres to try to read some absurd value of a pointer, causing a
panic error. However, since nothing is logged in the postgres log
complaining about bad pointers or something similar to that, I'd vote
you have a hardware fault.

One thing you can do to test your DB is to run vacuum full on it. If
this produce no errors, try reindexing your tables. This causes every
single row to be read which would throw an error if the table was
corrupted. Of course, on reindex you could just clear your problem if
an index itself was corrupted. If you do get an error during either
of these steps, you should reboot, then re-run the same command to see
if the error is in the exact same place every time. If not, then that
would be a very strong indicator of hardware fault. Run your machine
hardware diagnostics. If still nothing comes up, replace the machine.
:-(