proxy stops working

[OvGU]

After updating three proxies from 3.4.12 to 4.0.0.2, one of them stops collecting data after about 2 hours. After restarting the service it works fine again for 2 hours.
To be more precisely: icmp is still working but snmp- and agents-checks are not.

[OvGU]

To be more precisely: icmp is still working but snmp- or agent-checks are not.

[ingus.vilnis]

Hi,
What is in the zabbix_proxy.log file? Seeing anything suspicious there?

[OvGU]

I made a debug 5 log, but I can't see something suspicious, except that hosts checks disappears. The proxy-service is still running, but only ipmi- and icmp-checks are in the log.
One strange thing is the following:
3443:20181022:062500.882 End of process_value()
3443:20181022:062500.882 In process_value()
appears often in the log after the "crash".

[ingus.vilnis]

Debug 5 shows too much data to have a good oversight. Level 3 should be enough to spot majority of the problems.

[OvGU]

In den Level 3 Log is nothing strange at that time the checks (Agent Ping) stops. I forgot to say, that snmp-traps are still working, when that happened.

[OvGU]

I made a complete new proxy, but the same behaviour. for some clients I get the message: Zabbix agent on XXX ist unreachable for 5 minutes but zabbix_get for that host ist still working.

[ingus.vilnis]

Have a look at the proxy performance graphs (in Monitoring -> Graphs, select your proxy). Make sure the proxy host is first of all added to hosts and it is monitored by itself, never by Zabbix server. Particularly look at Zabbix data gathering process busy graph.

[OvGU]

The relevant proxy was monitrored by another proxy. The gathering process busy graph looks fine. The average of most of the processes are below 10 %. Except the pinger process. It has 61%, But I increased the StartPingers.

[ingus.vilnis]

In that case the graphs you saw were irrelevant to the problematic proxy. They were from the other proxy. You need to properly collect them from the problem proxy, observe the data and only then any further comments can be made about those graphs.

[OvGU]

Sorry, I changed the monitoring to the proxy by it self. All values decreased execpt busy poller.

[ingus.vilnis]

Pollers how busy on avg now?

[OvGU]

The higest scores are:
poller: 17%
pinger: 12%
unreachable: 4%

[ingus.vilnis]

And still 2h after the restart only this particular proxy stops collecting passive items?

[OvGU]

I moved all affected clients to other proxies. It was repeatable, when I monitoring proxy2 by proxy1.
The funny thing is, I have an old proxy1, this was updated and there are problems. So I installed a new one. Now monitoring proxy2 by proxy1-old works.
Step by step I will move the affected host back to proxy1.