Agent Unreachable Problem

[mellis]

This is an ongoing problem that I have not been able to correct for several years. I am to the point that I want to walk away from Zabbix. It just pops up and I do not have a handle on why it does, and then out of the blue it goes away for a couple days or weeks then comes back. Currently it is showing up in one install.
I am running this Zabbix 4.2.5 as a VMWare 5.5 VM with 20GB Ram, and 500GB disk. 6 cores on a HP 360G6, 2.4Ghz


As you can see almost every Agent is unreachable



One thing I find odd is the latest data show a Up(1) which I thought would be ok?

My item is setup to poll every 5 mins

In addition, my Trigger is setup for 30mins
{Template App Zabbix Agent:agent.ping.nodata(30m)}=1


I setup a ICMP ping for these same workstations and it is not alerting at all.

These are my zabbix_server.conf values
StartPollers=96
StartPreprocessors=10
StartPollersunreachable=24
StartTrappers=20
StartHTTPPollers=2
StartTimers=16
StartEscalators=10
HousekeepingFrequency=2
HouseKeeperDelete=100000
CacheSize=84M
HistoryCacheSize=1536M
HistoryIndexCacheSize=256M
TrendCacheSize=128M
ValueCacheSize=128M
Timeout=30

[mellis]

Looking at the latest data screen the time stamps are about ~15min behind on the CPU Processor Load checks



I have reviewed the NTP setting and the Zabbix Server is getting it's time from the Windows Domain controllers and running date it does have the same date as the windows systems.

I have been looking in the logs, really not seeing anything that stands out, but not realty sure what to look for.

[mellis]

It looks like the windows time server is up and running and using the same source as the zabbix server, Offend I do see breaks in the graphs at the same time as the unreachable triggers start up. It really seems like the graphs break a min or two before the triggers kick off.
In the queue it backs up to over 500 items in the 10min column.

I did disable all the workstations that seem to be the problem and it will recover, adding them back in little at a time and the problem seems to return when I get a little over 200 host enable.

[Markku]

I see you are using passive agents, meaning that the Zabbix server will poll each of the agents to get the data. The recommended way is to use active agents (= ServerActive in agents and ”Zabbix agent (active)” in items). Is there a specific reason why you are not using active agent items?

Markku