Hello.
I have a couple JMX interfaces defined on one of the hosts.
They handle one application each on this host.
It works fine. The data is contiuous, Graphs are smooth. That is until one application dies or is switched off.
In that case java jmx gateway logs timeout exceptions (thats obvious) but all monitoring data for other still running jvm's on that host gets chopped. Monitoring data can be interrupted for a couple minutes, sometimes hours. Graphs have gaps in them.
After unconfiguring jmx interface which is not working at the moment everything goes back to normal and monitoring is working fine.
In the same time other hosts which are monitored in the same way - through jmx - are not affected by this.
In other words: When one jmx interface on host dies then all other jmx monitoring data are incomplete on that host.
Other hosts are unaffected by this.
We have tried to add some pollers (from 5 to 20 and then up to 100)) with no change in behaviour.
The problem is: How to configure zabbix to be able continuously monitor jmx interfaces and dont have to unconfigure unused ones to have continuous monitoring data?
I have read a bit on the forum and found out that it could be configured in a way to have one host per jvm in zabbix. That might work. It was a case regarding templates but it should work in my case.
But is there a different way to handle this?
Regards
Łukasz Sczygiel.
I have a couple JMX interfaces defined on one of the hosts.
They handle one application each on this host.
It works fine. The data is contiuous, Graphs are smooth. That is until one application dies or is switched off.
In that case java jmx gateway logs timeout exceptions (thats obvious) but all monitoring data for other still running jvm's on that host gets chopped. Monitoring data can be interrupted for a couple minutes, sometimes hours. Graphs have gaps in them.
After unconfiguring jmx interface which is not working at the moment everything goes back to normal and monitoring is working fine.
In the same time other hosts which are monitored in the same way - through jmx - are not affected by this.
In other words: When one jmx interface on host dies then all other jmx monitoring data are incomplete on that host.
Other hosts are unaffected by this.
We have tried to add some pollers (from 5 to 20 and then up to 100)) with no change in behaviour.
The problem is: How to configure zabbix to be able continuously monitor jmx interfaces and dont have to unconfigure unused ones to have continuous monitoring data?
I have read a bit on the forum and found out that it could be configured in a way to have one host per jvm in zabbix. That might work. It was a case regarding templates but it should work in my case.
But is there a different way to handle this?
Regards
Łukasz Sczygiel.
Comment