Thread: FSM and VM file

akp geek wrote on 02.03.2010 22:11:
> Hi all -
>                There are lot of FSM and VM files getting generated in
> the base directory. Do we need these files and I don't know the reason
> why these files are getting generated. I read the documentation, but not
> able to follow well. I will read it again. But do we need to keep these
> files or there is any process can we run to clean these. Can you please
> help?
> Regards

My understanding is that the FSM files are the "Free Space Map" files.

But you should never ever under no circumstance delete files manually from the data directory!

Thomas

On 02/03/10 21:47, akp geek wrote:
> I am doing a replication of one of the database and the size of the slave
> database is growing exponentially . Right now the size of master db is 849M
> and the slave is 7GB.

> my master is 8.3 and slave is 8.4

I'm guessing your slave isn't being vacuumed. Or, perhaps you have a
long-running transaction on the slave that is preventing dead rows from
being cleaned up.

Two useful commands:

   vacuum verbose mytable;

This will show you how many rows/pages can/should be cleaned up.

   SELECT pg_size_pretty( pg_total_relation_size('mytable') );

This will show you the size of "mytable" (formatted nicely).

--
   Richard Huxton
   Archonet Ltd

On 03/03/10 13:32, akp geek wrote:
> Thank you all for the suggestions. I did a vacuum and the size has gone down
> drastically. But still it is not the same size as my master. I am looking
> into it

Check your autovacuum settings and you should be able to keep things
stable at least.

You might need to cluster tables / restart the replication to get the
best case. Vacuuming needs to be a continual process.

--
   Richard Huxton
   Archonet Ltd

On Wed, 2010-03-03 at 08:32 -0500, akp geek wrote:
> Thank you all for the suggestions. I did a vacuum and the size has
> gone down drastically. But still it is not the same size as my master.
> I am looking into it

Plain vacuum does not file system size of relations. It just marks dead
spaces are writable. If you haven't vacuumed for a long time, it
probably means that you have lots of space to be reused.
--
Devrim GÜNDÜZ
PostgreSQL Danışmanı/Consultant, Red Hat Certified Engineer
devrim~gunduz.org, devrim~PostgreSQL.org, devrim.gunduz~linux.org.tr
http://www.gunduz.org  Twitter: http://twitter.com/devrimgunduz

On 05/03/10 15:14, akp geek wrote:
> Hi All -
>
>                  I have turned on the auto vacuum on the slave and for some
> reason the db is getting bloated up. The master size is only 1G and the
> slave is at 9.2GB now. I did cluster on couple of tables also. did any one
> run into this situation? Can you please help?

OK - so autovacuum is now running on the slave, but your database keeps
growing anyway.

So - either:

1. You have a long-running transaction blocking the vacuum.
2. Your autovacuum isn't running often enough.

The first step with #1 is to check for idle transactions. Look in the
system view: pg_stat_activity
   SELECT * FROM pg_stat_activity;
Is there anything there that started a long time ago (probably marked
IDLE IN TRANSACTION)?


For #2, I would start with seeing what tables are affected, then check
your configuration settings. It might be that autovacuum_naptime is too
large for your workload, or (auto)vacuum_cost_delay is too high.

http://www.postgresql.org/docs/8.4/static/routine-vacuuming.html#AUTOVACUUM

http://developer.postgresql.org/pgdocs/postgres/runtime-config-autovacuum.html#RUNTIME-CONFIG-AUTOVACUUM


--
   Richard Huxton
   Archonet Ltd