Our application is broken down quite well. We have two main writing processes
writing to two separate sets of tables. No crossing over, nothign to prohibit the
vacuuming in the nature which you describe.
My longest transaction on the tables in question are typically quite short until
of course they begin to bloat.
On Wednesday 21 June 2006 11:08, Csaba Nagy wrote:
> > So, it appears my autovacuum is just NOT working... I must have screwed something up, but I cannot see what.
>
> Is it possible that you have long running transactions ? If yes, VACUUM
> is simply not efficient, as it won't eliminate the dead space
> accumulated during the long running transaction. In that case VACUUM
> FULL won't help you either as it also can't eliminate dead space still
> visible by old transactions, but from what you say I guess you really
> stop everything before doing VACUUM FULL so you might as well stopped
> the culprit transaction too... that's why the VACUUM FULL worked (if my
> assumption is correct).
>
> To check if this is the case, look for "idle in transaction" in your
> process listing (ps auxww|grep "idle in transaction"). If you got one
> (or more) of that, you found your problem. If not, hopefully others will
> help you :-)
>
> Cheers,
> Csaba.
>
>
>
>