1.8.3 -> 1.8.4 Performance problems

[Peteris]

Hi,

After upgrading from Zabbix 1.8.3 to 1.8.4 we have run in to serious performance issues. Data queue up, and is not collected properly despite the fact that 1.8.4 should be faster.

I have added two graphs where you can see that data is collected very badly.

DB: Oracle.

Have any body had same issue? Is there any solution available to solve this?

[untergeek]

How many "new items per second" does Zabbix say you need?

How many DB Syncers do you have configured in your zabbix_server.conf?

What version of Oracle? How fast is the backend?

We are comfortably handling over 400 items per second (according to Zabbix) with 1.8.4 on RHEL with Oracle 11.2.0.2 running on a massive Sun Solaris backend box (128 cores, 128G of RAM).

Of course, Zabbix shares this database server with a LOT of other production databases, but you get the idea.

[Peteris]

About 305 items per second for 600 hosts.

DB Syncers = 32. How many you have or what is recommended - I tried to use from 16 to 64 didn't see any noticeable difference.

DB server is running Oracle 10, IBM AIX, 8G RAM, RAID 10 FC disks (SAN Array), not sure about cores, but it's only dedicated to Zabbix DB.

[Peteris]

We just downgraded to 1.8.3 because, 1.8.4 had one more problem with rights in front-end - user who had read/write permissions couldn't link template to host any more. Downgrade solved this problem...