Postgres causing high IOwait

[tenaciousC]

Hi,

We have isolated the Postgres db as the cause of high io wait on our server which is home to 314 mailboxes:

Code: Select all

top - 16:09:45 up 574 days, 23:28,  6 users,  load average: 12.05, 15.94, 22.97
Tasks: 659 total,   2 running, 657 sleeping,   0 stopped,   0 zombie
Cpu(s):  8.4%us,  6.0%sy,  0.0%ni, 59.6%id, 25.9%wa,  0.0%hi,  0.2%si,  0.0%st
Mem:   4145320k total,  3994588k used,   150732k free,    56304k buffers
Swap:  8193140k total,    20332k used,  8172808k free,  1633972k cached

The size of the database is 43Gb:

Code: Select all

du -sh /var/opt/scalix/mx/postgres/
43G   /var/opt/scalix/mx/postgres/

The nightly scalix reindex script seems to run and not finish.

Code: Select all

/opt/scalix-postgres/bin/sxpsql-reindex -f

We wonder about recreating the database using : /opt/scalix-postgres/bin/sxpsql-recreatedb

We would like to know if this will have any ill effects on Active Sync, as this db is used by the Active Sync server.

Any help greatly appreciated.

Cheers

C

[ls-al]

Recreating the DB is not an option unless you want that all your ActiveSync users have to recreate their AS configuration.
If your using a Postgres 8 database (RHEL5 for example) you will probably have to do a full vacuum. Auto-vacuum is not doing everything.

Code: Select all

/opt/scalix-postgres/bin/sxpsql-vacuum -F -f

This should be done during off-peak hours.

[tenaciousC]

Thanks for this.

After vacuum, db size has gone down to 3.4Gb and iowait has returned to very low. Most pleasing!


Cheers

C