Zabbix 2.0.3 false Agent unreachable alerts

[nicolasg]

Hi All,

Our current Zabbix setup is running the frontend in a different server from the backend (MySQL 5.5), after upgrading Zabbix 1.8.14 to 2.0.3 we are experiencing some serious issues with the new version :

Usually once every 10 days at a ramdom time Zabbix starts sending alerts that all Agents are unreachable for more than 5 minutes which is not true.

The trigger is fired and cleared right away on most of the cases but there are some times the alerts is ON for a couple of minutes. We can also see some degraded performance regarding the rendering of the Zabbix Screen in comparison to the old version...

Setting the notification with 1 minute delay didn't seem to help..

I have attached graphs from the Zabbix server internals which indicates some performance issues during the problem period but I have no clue what the problem is,,

Is anyone having the same experience that could help ?

Regards,
NicolasG.

[tchjts1]

Not having that issue, but how many hosts are you monitoring?
Maybe worth bumping up the StartPollersUnreachable value in your zabbix_server.conf

It only defaults to one poller. Mine is set to 2...

### Option: StartPollersUnreachable
# Number of pre-forked instances of pollers for unreachable hosts (including IPMI).
#
# Mandatory: no
# Range: 0-1000
# Default:
# StartPollersUnreachable=1
StartPollersUnreachable=2

[nicolasg]

thanks tchjts1, I have increase the value from 1 to 2. Need to wait now to see if the problem will appear again in a 2 weeks period.

Do you know what other values could also affect the issue ? We are also experiencing slow performance in the Screens and Zabbix graphs load pages..

[tchjts1]

No, sorry, I have no performance issues at the moment in graphs and screens. Since they implemented flicker-free, those have been drawing fairly fast for me.

You restarted your zabbix_server process after you adjusted that value, right?

Question on those spikes you see on the unreachable hosts... does that happen to coincide with any backup of the DB that happens during that time?

I noticed yesterday when we did a Mysqldump that I got a similar spike as you show.

[tchjts1]

We are also experiencing slow performance in the Screens and Zabbix graphs load pages..

You don't happen to be on RHEL v6.1 (Santiago) are you?

[nicolasg]

You restarted your zabbix_server process after you adjusted that value, right?

Yes

Question on those spikes you see on the unreachable hosts... does that happen to coincide with any backup of the DB that happens during that time?

I thought about that but as I wrote the issue occur at a random time, no relation to any backup or cronjob could be found...

You don't happen to be on RHEL v6.1 (Santiago) are you?

FrontEnd -> CentOS release 6.2

BackEnd (MySQL) -> CentOS release 5.8 (Final)