Setting max_fsm_pages

"Carlos Oliva" <carlos@pbsinet.com> writes:

Should I set the max_fsm_pages to the value reported (vacuum verbose) as
pages stored or the value reported as total pages needed?

Total pages needed ... if not more.

regards, tom lane

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)

-----Original Message-----
From: pgsql-general-owner@postgresql.org
[mailto:pgsql-general-owner@postgresql.org] On Behalf Of Tom Lane
Sent: Saturday, November 05, 2005 10:23 PM
To: Carlos Oliva
Cc: pgsql-general@postgresql.org
Subject: Re: [GENERAL] Setting max_fsm_pages

"Carlos Oliva" <carlos@pbsinet.com> writes:

Should I set the max_fsm_pages to the value reported (vacuum verbose) as
pages stored or the value reported as total pages needed?

Total pages needed ... if not more.

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match

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@pervasive.com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461

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@postgresql.org
[mailto:pgsql-general-owner@postgresql.org] On Behalf Of Jim C. Nasby
Sent: Monday, November 07, 2005 2:38 PM
To: Carlos Oliva
Cc: 'Tom Lane'; pgsql-general@postgresql.org
Subject: Re: [GENERAL] 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@pervasive.com
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

You might have to bump up shmmax, but fsm is completely unrelated to
shared_buffers.

On Mon, Nov 07, 2005 at 02:55:41PM -0500, Carlos Oliva wrote:

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@postgresql.org
[mailto:pgsql-general-owner@postgresql.org] On Behalf Of Jim C. Nasby
Sent: Monday, November 07, 2005 2:38 PM
To: Carlos Oliva
Cc: 'Tom Lane'; pgsql-general@postgresql.org
Subject: Re: [GENERAL] 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@pervasive.com
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@postgresql.org so that your
message can get through to the mailing list cleanly

--
Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461

As we're talking about fsm, I've got a question I've been asking myself for
some time :
If a cluster is restarted, is the fsm forgotten ? (so does it mean one should
run a vacuum as soon as the database is restarted ?)

Le Dimanche 06 Novembre 2005 03:44, Carlos Oliva a écrit :

Marc Cousin <cousinmarc@free.fr> writes:

If a cluster is restarted, is the fsm forgotten ?

Given a normal postmaster shutdown, no. In a crash-restart situation,
yes.

regards, tom lane

On Tue, Nov 08, 2005 at 03:36:40PM -0500, Tom Lane wrote:

Marc Cousin <cousinmarc@free.fr> writes:

If a cluster is restarted, is the fsm forgotten ?

Given a normal postmaster shutdown, no. In a crash-restart situation,
yes.

Does that include restarts due to things like failed asserts and kill
-9'ing a backend?
--
Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461