On Tue, Sep 22, 2009 at 15:19, Alan McKay <alan.mckay@xxxxxxxxx> wrote: > On Tue, Sep 22, 2009 at 2:42 AM, Magnus Hagander <magnus@xxxxxxxxxxxx> wrote: >> That's not true at all. >> >> If you have many relations in your cluster that have at some point been >> touched, the starts collector can create a *significant* load on the I/o >> system. I've come across several cases where the only choice was to disable >> the collector completely, even given all the drawbacks from that. > > Thanks Magnus, I thought that other response sounded a bit fanciful :-) > > So is there any way to predict the load this will have? Or just try > it and hope for the best? :-) IIRC, the size of the statsfile will be: * Some header data (small) * For each database, not much data (IIRC about 10-15 32-bit values, so less than 100 bytes) * For each table, around 25 32-bit values, so somewhere around 100 bytes It's the table stuff that can increase the size, unless you have very many databases with just one or so tables in them. The table stats will also be written for system tables. This file will be written twice per second on 8.3 and earlier (on 8.4, only on demand). It will be written as a new file and then renamed into place, so there is also filesystem operations being created - which unfortunately are on your main data drive (unless, again, you're on 8.4 and moved it to tmpfs) -- Magnus Hagander Me: http://www.hagander.net/ Work: http://www.redpill-linpro.com/ -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
