agent.ping.nodata(300) false alarms, v1.8.2

[fascinatedcow]

Hi,

I believe this issue is similar to this one:

http://www.zabbix.com/forum/showthread.php?t=14831&highlight=nodata

But I'm not sure if it's exactly the same so I thought I'd start a new thread.

I have a single node and a single proxy currently in operation, both version 1.8.2.

What I am seeing is every so often a bunch of servers trigger on the above expression. Usually about 5 at a time. It seems to happen completely randomly. As advised by Richlv, I have added the following item to my node:

zabbix[proxy,proxy00,lastaccess]

along with the trigger:

{node:zabbix[proxy,proxy00,lastaccess].fuzzytime(120)}=0

I would expect that if my proxy was not delivering data on time that I would see this trigger fire 3 minutes before any agent.ping.nodata(300) triggers. I am not, however seeing this. The last occurrence of a false alarm was 3 hosts paging down at 15:18, then clearing within a minute. The last occurrence of the proxy being out of touch for over 2 minutes prior to that was at 15:06, and it cleared at 15:07.

Checking latest data for one of the hosts in question shows no missing agent.ping item values (which are all 1).

Does anyone have any idea what's going on here and how to fix it? Is it worth changing the DataSenderFrequency variable to something lower than the default of 1? Will it be more efficient to keep local data on the proxy for, say 10 seconds, or 30 even, and then send it in a batch to the node?

Any input greatly appreciated.


Matt

Running sqlite on proxy btw.

[davidpe]

I'm experiencing the same issue as well with only hosts being monitored via a proxy. I'm also using sqlite on the proxy and am wondering if the performance of sqlite is contributing to it and if a migration to mysql is in order.

[fascinatedcow]

Hi davidpe,

It could be your sqlite but that wasn't the answer for me. My answer was here: http://www.zabbix.com/forum/showthre...9422#post69422.

Do you see the same behaviour when you monitor from your node, or only from the proxy? Perhaps you need to change the number of pollers you have started on your proxy.

To see if your proxy is getting behind, use the internal item lastseen (I think that's what it's called).

Also, to make sqlite faster you can put the db on a filesystem that's mounted with the noatime option.


Matt