Josh Berkus wrote:
> Bruce, Sean,
>
> > Oh, yes, you are right. My hardware tuning guide mentions it.
> > Strange
> > it is called the transaction log file:
> >
> > http://www.ca.postgresql.org/docs/momjian/hw_performance/
> >
> > Unless you are seeing this more freqently than every minute, it
> > should
> > be fine.
>
> Actually, it's apparently a real problem, because the function never
> completes. Each cycle of "recycling transaction logs" takes longer
> and longer, and eventually locks up completely.
>
> What the function is doing is a succession of data cleanup procedures,
> updating the same table about 50 times. I will be very thankful for
> the day when I can commit within a procedure.
>
> Unfortunately, I am already at the maximum number of WAL files (64).
> What do I do now?
Wow, that is interesting. I thought one big transaction wouldn't lock
up the WAL records. I figured there would be a CHECKPOINT, and then the
WAL records could be recycled, even though the transaction is still
open.
Where do you see 64 as the maximum number of WAL segments. What is your
checkpoint_segments value? The actual number of files shouldn't be much
more than twice that value. What PostgreSQL version are you using?
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073
