On 10/23/07, Martin Strejc <martin.strejc@super-web.cz> wrote:
> Hi!
>
> I've a special kind of problem with PostgreSQL 8.2.5 and I haven't
> found anything about that on your web or searchig by Google.
>
> The database slows down when more over many (cca 100 connections)
> are incoming. The problem begun the same time as the VACUUM foroze on
> pg_toast.pg_toast without any reason or message in log. Actually it got
> after that troubles with reindex and VACUUM ANALYZE.
> Of course the shmmax, shared mem, max connections and some other
> parameter in postgresql.conf are set to get the best preformance of the
> server.
This is all a little vague. What were the sequence of events, and
what were the actual error messages you got? What do you mean by
froze? how long was it not returning, and what did you do to unfreeze
it, if anything?
what are your postgresql.conf settings, specifically the ones you've changed.
What does the output of vacuum verbose on this database say?
Specifically the last 5 or 10 lines.
The more info you can give us about your problem, the better we can help.
> Is it possible to delete any of pg_toast or pg_temp tables? How to
> do that if possible?
Not a good idea. pg_temp tables will usually not be an issue, unless
you're running out of free disk space. pg_toast is where your
overlarge text fields go.