Postgres overload with Zabbix 1.4

[Zachar Balazs]

Hello!

I have Zabbix 1.4 with postgresql-server-8.1.3-12.

I am monitoring about 30 server (about 600 item) with zabbix... Everything works fine as long as I rebooted the server whereon the zabbix_server and the postgresql are run...

After the reboot, the items in the zabbix queue were going up, and the postgresql's CPU usage become 99%... And the zabbix is monitored just several items (about 25) from the 600 item, and it was randomly...

When I disable all host in zabbix, the postgresql's CPU usage become 0,3%... But when I enable just one host with 10 item the database cpu usage become 99% again, and the zabbix is get value from only 3 item... In the queue there are the another 7 but with very old monitoring time (two days before, etc...). It looks like the items are freeze into the queue...

I attached the server log when I start the zabbix_server with one host enabled...


Has anybody Idea?

Sorry for my poor english, i hope you understand it...

regards,
Balázs

[Alexei]

Did you run vacuum? PostgreSQL is a highly inefficient database when it comes to a very frequent updates of a single table. Periodic (auto-)vacuuming may help you.

[Zachar Balazs]

Thanks, its works!

Thank you Alexei, now its works fine again!

I always use mysql before, and I didn't know this...

Thanks a lot!

Zabbix is the King of monitoring...

Regards,
Balazs

[NOB]

Hi,

but in src/libs/zbxdbhigh/db.c there is a DBVacuum
function which is enabled, if PostgreSQL is used.
Is the list of tables not complete ?

I am not using it (commented because of high DB load when used), but use
(PostgreSQL server-side) auto-vacuum instead.
That's why I might not have noticed it in the beta versions 1.3.x.

Regards,

Norbert.