Christopher Cashell wrote:
> [1] Upon rebooting, the autovacuum utility kind of harfed. Here is
> the log entries that it made:
>
> ---
> [2004-06-30 08:44:53 AM] Failed connection to database template1
> with error: FATAL: the database system is starting up
> .
> [2004-06-30 08:44:53 AM] Failed connection to database template1
> with error: FATAL: the database system is starting up
> .
> [2004-06-30 08:44:53 AM] Error: Cannot connect to template1,
> exiting.
> ---
>
> When I first saw this, the problem seemed to still be there, and I
> saw "FATAL: the database system is starting up" running across my
> brain in bright red letters.
>
> I took a minute to breath, and luckily realized that at least one
> program (that starts automatically upon reboot) seemed to be
> accessing the database just fine. At that, I felt a little hope
> return.
>
> So, I shut down Postgres, then restarted Postgres, then restarted
> the autovacuum utility, and everything worked just peachy. I'm
> guessing that perhaps the autovacuum tool was trying to connect to
> Postgres while it was replaying the transaction log from not
> having been shut down cleanly, and that's why it choked, but I
> don't know that for sure.
This is a known issue with the autovacuum contrib module. This problem
goes away when pg_autovacuum is integrated into the backend. I
submitted a patch that makes pg_autovacuum a backend process, but so
far, I have yet to get any response from anyone.
Matthew