Zabbix 2.0 queue: 18095 in more than 10 minutes

[gandalf]

As subject, I have 18095 events queued more than 10 minutes.
I'm using only SNMP discovery with low level discovery but seems to be very unreliable.

Any suggestions? Actually, I'm unable to graph network traffic and many other things due to an overloaded queue.

With Zabbix 1.8 i had no issue.

[mike@vtg]

Gandalf, have you done any database tuning since upgrading? I had some significant queue backlogs until I spent time tuning my installation of MySQL using the various tips in the forums. I suspect that I had more problems in 1.8.x than I knew of and they appeared after going to 2.0.

The performance graphs may reveal much if you have the zabbix server monitored by the built-in zabbix server template.

Specifically my problems were related to disk IO performance and were most noticed when the housekeeper process runs. I changed housekeeper to run at 24 hours as opposed to 1 hour, and also added threads to poller. The most important change that helped was allocating much more RAM to InnoDB (I used 75% of the 8GB in my server).

[gandalf]

Actually i'm using a VPS with 1GB RAM and zabbix 2.0
The same number of server was monitored with zabbix 1.8 but with half ram with no issue at all.

I'm not running any housekeeper at the moment.
Increasing pollers count will give me tons of concurrency issue, like table lock and so on.

Also, from dashboard, latest 20 issue are all the same, I see the same line replicated 20 times.

[marcherren]

Gandalf,

Are you using SNMP v2 or SNMP v3? I face a similar problem when using v3

See here

So far using v2c it works fine.

Best regards,

Marc

[gandalf]

i'm using snmp v2

[johnw230873]

Is it just one host causing the issue? In the past I would look for the oldest queue record and disable that host to see if the queue would drop. If it did then I knew it was a configuration with that host.