Problems with Zabbix Graphs

[lhaig]

Hi,

We have a zabbix system that monitors about 170 hosts with about 30 items per host. and between 15 to 30 triggers per host.

We graph a large amount of data on the system and have noticed that some of the graphs if not all have gaps in the data sometimes large amounts of data is missing.

What highlighted the issue is that we are monitoring a netscreen ssg140 and are trying to track down why a line is failing every now and again.

When we tried to lookup the throughput graphs we noticed that the graph only populated every now and again.

Our Zabbix server has the following specification.

Xen VM
6GB RAM
2CPU's
1NIC (1GB)
3 HDD's
10Gb Root
100GB logs
500GB MYSQL DB

Our DB is really large

I was wondering if we are expecting to much from our zabbix server, and whether we need to split off the tasks to more than one.

Do you guys have any suggestions?

Thanks

Lance

[tchjts1]

1. What version of Zabbix Server are you using?
2. What version of Zabbix agents are you using?
3. Are you using "Zabbix Agent" or "Zabbix Agent (Active)" for your items?
4. What are the update intervals for the items you see gaps in?
5. How many pollers does Zabbix start? (zabbix_server.conf)

I've seen this issue in my environments with 1.6.x and older versions of Zabbix. Changing over to using Zabbix agent (Active) solved the gaps issue for me.

For the active agent to work, you must have the Hostname= field populated in your zabbix_agentd.conf file on each monitored server, and whatever name you have in there must exactly match the "name" of the host as you have it entered in Zabbix fronted, including case.

[MrKen]

A quick fix for gaps in graphs is to increase the number of StartPollers and StartTrappers in your zabbix_server.conf

While that fix may solve the problem in the short-term, I believe the correct answer is that it is time to do some mysql tuning. As your database grows your mysql configuration needs to change to accomodate that growth.

You may find some benefit in moving your database to its own server. Also, moving Zabbix to a physical server rather than a virtual machine will also help.

Here's a link to a very helpful forum post called 'Performance Tuning Mysql':

http://www.zabbix.com/forum/showthread.php?t=12407

MrKen

[lhaig]

Thanks for the tips Guys,

After posting this I sat down with a colleague and noticed our queue had over 5000 items waiting more than 1 minute. Looking at our checks we had far to many 30 60 80 and 90 second checks. We have changed many of these and it has improved things. But the queue is still around 1000.

I will get the exact zabbix versions from the zabbix admin later today and post them.

I think that it is actually our Db that is the problem as it is rather large.