[POLL] which database are you using?

[acalvo]

We've been trying zabbix for several months now, and we're pretty amazed with all the features it provides.
However, we've faced an internal discussion lately about which database engine will be the best for this application: mysql, postgresql and oracle.

Could you please post which database engine are you using and if you tunned it?

Thanks!

[qix]

MySQL

In my old environment I did tune it with mysqltuner. It was a master-master replicating setup.

Currently I'm running the stock version that comes with Debian Lenny since I don't yet have a lot of items/hosts. I did switch to InnoDB to be 'future-proof'

I don't have a lot of experience with PostgreSQL, but I might consider a switch seeing what is going on at Oracle with all the OSS projects.

[fgallese]

PostgreSQL Here

Hi,

We are using PostgreSQL.

~150 Hosts in passive mode, almost none host with active checks.
16.65 New values per second.
1473 Items.

Never tuned PostgreSQL.

Never had an issue with the database neither with performance.

Regards,
Fabio

[untergeek]

Oracle 11.2.0.2

Yes, we tuned it. We are still tuning it. It ran flawlessly with 11.2.0.1, but we've had some strange errors which crop up periodically with 11.2.0.2 which require a db restart. Our dba's are working on the problem.

One thing I will tell you, based on our experience: Do NOT use Sun hardware for the zabbix_server or for the frontend, unless it's x86 hardware. SPARC-based hardware has caused nothing but grief for our load because the single-thread speed was way too slow. Zabbix is compiled and written for Linux's single-thread speed on x86 hardware. It works, but it's slow and we had a constant backlog to our Oracle server. It might have been faster writing to MySQL, but not for Oracle.

We've since migrated to HP DL380s running RHEL5 for our Zabbix servers. The speed increase was noticeable.

[Rob]

MySQL on Ubuntu 10.04 using InnoDB. Tuned, but basically this just meant setting the innodb_buffer_pool_size to use a big chunk of RAM.

Server has 2x Dual core Xeons and 8GB. Monitoring 184 hosts, 14150 items, 9140 triggers, a total of 108 new values per second.

Server load average is 0.86 but this is only because of housekeeping runs which make it peak at around 2.5. In between housekeeping runs it averages 0.54. I'm pretty sure most of the load is disk IO rather than CPU.