Zabbix performance problems

[emz]

Hello,

I'm sorry to bother you but I'm now desperate to find a solution for this problem because I've tried everything and nothing helped.

So, here is the situation: I have one newly installed VMWare virtual server - it has 4 virtual CPUs and 6 GB RAM. It is deployed on RAID 5 disk array.

Zabbix Graphs:

http://postimg.org/gallery/blul41o0/

The trapper and poller processes are constantly working at 100%. It is stable for only few minutes after a restart of the Zabbix server service is performed and after that it goes to 100% again.

Here is my configuration files:
MySQL my.cnf - http://pastebin.com/VXks67Fe
Zabbix Server (v2.2.2) config file - http://pastebin.com/b6w7yPjP

Please, I need help on this.

[tchjts1]

Several things you can do.

Edit your zabbix_server.conf file and increase the values for:
StartPollers= <-- Try increasing that one by 50
StartTrappers= <-- Try increasing that one by 15 or 20
Timeout= <-- Try setting that one to 10

There are also 3 cache parameters that can be set in that file. If you are still using the default values, I would set each one to 128M

Note that for all of the above settings, you either have to enter a new line with the comment # sign removed, or adjust the existing values, but remove the comment # sign.

Restart your Zabbix server process. Let it run for an hour or so then recheck your internal Zabbix graphs and make any further adjustments.

So, are you saying that you run Zabbix application AND DB on the same VM, which only has 6GB of RAM? If that is the case, you can pretty much expect marginal performance out of that. Ideally you would run each on it's own VM and you would give your DB server a good amount of RAM to work with.

I am on VM's and my DB has 16G.

Lastly, looking at your my.cnf file, IMO you should have invoked the innodb engine with the file_per_table setting when you did your initial install. I am not really a DB guy, so before you change any of your my.cnf settings, check with someone who is in the know of such things. I don't know what the outcome would be to try and change that in mid stream.

(Edit) Just looked at your zabbix_server.conf file. Your trappers and pollers are already set at a number that I would think sufficient, and your Timeout= value is fine. Your cache settings also look reasonable, but maybe try bumping them up slightly... maybe 50M per setting.

[aib]

Will be nice to see the Queue - Overview to understand, which kind of Items is the slowest one.

Then you can try to improve the amount of pollers for that kind of items.

[emz]

Still the same

I've tried to increase again the cache settings in zabbix_server.conf. Unfortunately, nothing helped.
Most of the Zabbix pollers are 100% busy, the graphs couldn't be drawn and in the same time I see that the memory is 90% free - like the system doesn't use the memory that's allocated for it:

http://s2.postimg.org/v3ktp42w9/screenshot_10_116_70_243_2014_06_13_09_21_52.png

So, the new events - for example for a restarted server, or for too busy hard disks on the monitored servers seem to appear frequently, but most of the graphs are empty.

Queue - Overview:

http://s27.postimg.org/3uig40jtv/screenshot_10_116_70_243_2014_06_13_09_16_06.png

Any additional help is highly appreciated!

[tchjts1]

When you increased those values, did you remove the # sign that begins the line? If not, those new values are ignored and the default values are still used.