-
Problem report
-
Resolution: Fixed
-
Major
-
5.0.2, 5.0.3
-
None
-
Zabbix 5.0.2 and 5.0.3 (tested with both)
Server, Proxy, Nginx, Frontend all inside Docker
Passive Agent on Macos.
-
Sprint 71 (Dec 2020)
-
1
In a somewhat larger Zabbix environment we encountered the problem that when updating from 4.x to 5.0.2, a huge amount of nodata() based triggers spawned new problems just after the successful upgrade of the environment. Now after a minor upgrade from 5.0.2 to 5.0.3 the same happened again, causing several hundrets of false positive alarms.
A deeper investigation in an isolated environment made it possible to reproduce this error behavior: Triggers on hosts monitored via a Zabbix proxy (only tested with ACTIVE proxy) based on the nodata() trigger function and being active (in problem state) will RESOLVE after stopping and starting the Zabbix Server. Apparently, this only happens if between stop and start of the Zabbix Server there is above around 1 minute of time gap. Those resolved triggers will then re-generate new problems after their nodata(xx) time is over, and this way generate new problems, new alerts etc.
Steps to reproduce:
Preparation:
- spawned Zabbix server, an active proxy, one (passive only) agent up.
- to the existing template "Template Module Zabbix agent", added a trigger "Agent nodata 5m" - "{Template Module Zabbix agent:agent.ping.nodata(5m)}=1" - see attachment "exported_template.xml"
- added the active Zabbix proxy to the configuration and make sure it fetches configuration
- added two hosts: macbook-direct and macbook-via-proxy, both using the template "Template Module Zabbix agent" - see attachment "exported_hosts.xml"
- wait until both hosts get "available"
Result:
Based on the above preparations, note the following procedure:
15:00 - both hosts "ZBX available", no problems present
15:10 - stopped Zabbix agent
15:15 - nodata (and originally existing zabbix-available) triggers spawn problems - see attachment "Screenshot01 - problems active.png"
15:38 - restart Zabbix server (downtime a bit over a minute)
15:39 - nodata-Trigger on the host being monitored via proxy recovers (while the one on the host NOT monitored via proxy stays active)
15:44 - nodata-Trigger on the host monitored via proxy re-triggers -> new problem (Latest data shows clearly, that there have been no values on the item being
responsible for the nodata trigger on the host monitored via proxy since 15:10) - see attachment "Screenshot02 - resolved and new problem.png"
Expected:
Problem on host monitored via proxy should not recover and also not re-spawn. It should just stay active as it was before the Zabbix server restart. See attachment "Screenshot03 - Latest Data.png" to prove there has not been any values on the Item in question during the whole test time frame.
- caused by
-
ZBXNEXT-1891 Implicit trigger dependency when monitored via proxy
- Closed