Re: long running transactions - Mailing list pgsql-performance

From Tobias Brox
Subject Re: long running transactions
Date
Msg-id 20061010184354.GA24098@oppetid.no
Whole thread Raw
In response to Re: long running transactions  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-performance
[Tom Lane - Tue at 02:26:53PM -0400]
> > autovacuum_vacuum_cost_delay = 500
> > autovacuum_vacuum_cost_limit = 200
>
> Well, that's going to cause it to sleep half a second after every dozen
> or so page I/Os.  I think you'd be well advised to reduce the delay.

Modified it to 20/250, and it definitively helped.  Sorry for the
list verbosity; I should have been able to resolve this myself already
some 2-3 emails ago :-)  I wanted a "soft" introduction of autovac in
production, and assumed that it was better to begin with too much sleep
than too little!  Well, well.

pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: long running transactions
Next
From: "Jim C. Nasby"
Date:
Subject: Re: Simple join optimized badly?