Hi Jim, Thank you for your help. We are going to increase the max_fxm_pages according to the test I have been running through out the week. If we increase the max_fsm_pages, do we need to bump up the shared_buffers and the size of the shared memory segment of the Linux kernel(shmmax)? -----Original Message----- From: pgsql-general-owner@xxxxxxxxxxxxxx [mailto:pgsql-general-owner@xxxxxxxxxxxxxx] On Behalf Of Jim C. Nasby Sent: Monday, November 07, 2005 2:38 PM To: Carlos Oliva Cc: 'Tom Lane'; pgsql-general@xxxxxxxxxxxxxx Subject: Re: Setting max_fsm_pages On Sun, Nov 06, 2005 at 08:05:29PM -0500, Carlos Oliva wrote: > Thank you for your response Tom. > > Should I set the max_fsm_pages to the "total pages needed" obtained from a > full vacuum or from a analize vacuum? When I run a vacuum analyze (vacuumdb > -z -v), I get a smaller number of pages needed than when I run a full vacuum > with analyze (vacuumdb -f -z -v) There shouldn't be any difference because of analyze. But remember that as the tables change in size (as well as in the number of dead tuples), total pages needed can change. For example, if you run a vacuum immediately after a vacuum full on a system with no other activity, you'll get: INFO: free space map: 0 relations, 0 pages stored; 0 total pages needed That's because there's no dead space to be reclaimed. Your best bet is to do a vacuum verbose (vacuumdb -v) after the database has been running for a while using whatever vacuuming scheme you're going to use (such as pg_autovacuum). That will give you a pretty good estimate of how many pages you really need. Even that's not 100% reliable though, so you still need to include extra space as a safety margin. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@xxxxxxxxxxxxx Pervasive Software http://pervasive.com work: 512-231-6117 vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461 ---------------------------(end of broadcast)--------------------------- TIP 5: don't forget to increase your free space map settings ---------------------------(end of broadcast)--------------------------- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to majordomo@xxxxxxxxxxxxxx so that your message can get through to the mailing list cleanly
