Ad Widget

Collapse

Zabbix Agent2 container in ECS deployment failing when switching to 7.0.22

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • jhboricua
    Senior Member
    • Dec 2021
    • 114

    #1

    Zabbix Agent2 container in ECS deployment failing when switching to 7.0.22

    We have been running our Zabbix server and frontends with containers in ECS since the 7.0 release came out. The ECS cluster has two tasks defined. One task deploys Zabbix Server, Agent2 and the snmptraps container (backend). The other task deploys two zabbix nginx frontend containers (frontend).

    We usually update during our quarterly patching cycle and it basically just means bumping the Zabbix version in the task to whatever is the current one during that time, and redeploying the services so that it pulls the new images. Currently we are on 7.0.21. Last week, we tried going to the 7.0.22 image and the deployment of the backend task that has the server, agent and snmp containers started flapping. Looking at the logs in ECS, the agent2 container is throwing an error and terminating, causing the task deployment to fail and retry. The error we see is:

    Code:
    zabbix_agent2 [7]: ERROR: Cannot register plugins: failed to get socket listener: failed to create plugin listener with socket path, /tmp/agent.plugin.sock, listen unix /tmp/agent.plugin.sock: bind: permission denied.
    I can't seem to find anything in release notes about this so I'm not sure what has changed. Tried setting an env var for ZBX_PLUGINSOCKET with a value of /var/run/zabbix_agent.plugin.sock in the ECS task definition for the container, thinking that would map to the agent2 config file variable PluginSocket but that doesn't seem to work, I'm still getting the error above and the path in the error didn't change. Anyone else running into this or have I run into a bug?
Working...