Zabbix server blocks communications?

[Gotxi]

[SOLVED] (see last post)

Ok... this is so weird...

I deployed zabbix on one of my clients to monitor their windows servers and i created 10 hosts and assigned them the windows os template and installed agents on every host.

But every 5 minutes or so, i have triggers saying that agent on some machines are unreachable for 5 minutes (agent.ping)

Then i try to manually ping those machines from terminal when that happens, it misses 100% of the packets. The only way i can ping those machines again is to stop zabbix-server service and then the machines reply ping packets again.

While this happens, i can ping those machines from every other machine on the network with 0% packet loss, so it is a zabbix server issue only.

If i leave that problem as it is and wait 10-15 minutes, the triggers go OK and then i can manually ping those machines again until 5-10 minutes when they fail again.


The zabbix server is on an ubuntu 14.04 LTS virtual machine with full apt-get update and upgrade done, the machine doesnt seem resource intensive even when this problem happens:

This is the process TOP record:
http://pastebin.com/mk35UkxU

This is zabbix.conf without comments:
http://pastebin.com/BRF59KVL


If you need more info tell me what you need.

Anyone know what happens here?

[evertonco]

Have you tried disable firewall of windows hosts?

[Gotxi]

No, but that would be always 0% ping, not "sometimes". Im pretty sure is not a firewall issue.

And when zabbix-server is stopped every ping works again, so nop

[ingus.vilnis]

Hi,

Check that everything is fine with the performance of your Zabbix server.

https://www.zabbix.com/forum/showthread.php?t=47781

Check also Zabbix server log file if you see some issues there.

Best Regards,
Ingus

[Gotxi]

I have finally managed to solve it. It was not a Zabbix issue.

My Zabbix is a vmware VM and the problem was that it dropped a huge percent of received packets, even with other vm's in the same host, same vswitch, without firewalls or antivirus.

Of course i tried different vNIC drivers and the percent was different, but the problemd persisted.

Zabbix was installed on top of an ubuntu 14.04 and the solution was to increment RX buffer with ethtool cmd tool like this:

ethtool -G eth0 rx 3000

The default was 256, i tried some increments and saw that the drop packet counter were less as i incremented the RX buffer. The value im working on right now is 3000.

There is still a little % of dropped packets, but it is not affecting Zabbix anymore, and also, linux kernel now handles differently the dropped packets counter so im fine right now.