k8s cluster info:
Bare metal install on Debian 11.5 using CRI-O runtime and MetalLB for ingress. The cluster has a Rancher dashboard and Longhorn storage, all working without issue.
I’ve got a fresh install on a bare metal k8s cluster. I’m attempting to configure it to use a distributed dispatcher, however the predis service times out despite redis being available and accessible via CLI form all pods in the namespace.
If I only enable SESSION_DRIVER the pod will start, however this is error is found in the librenms.log file
[2022-10-24T09:36:22.960966-06:00] production.ERROR: Operation timed out [tcp://redis:tcp://172.16.240.253:6379] {"exception":"[object] (Predis\\Connection\\ConnectionException(code: 110): Operation timed out [tcp://redis:tcp://172.16.240.253:6379] at /opt/librenms/vendor/predis/predis/src/Connection/AbstractConnection.php:155)"}
If I enable both CACHE_DRIVER and SESSION_DRIVER, this error is encountered which causes a CrashedLoopBackOff event and the pod never comes fully online.
This error happens after Database seeding completes successfully, during the Clear cache.
In AbstractConnection.php line 155: Operation timed out [tcp://redis:tcp://172.16.240.253:6379]
I can’t seem to solve why Predis is timing out when trying to reach the redis service, which is online and available on the IP listed in the error logs. Command line tests using redis-cli from any node work without issue.
Note: If I leave both SESSION_DRIVER and CACHE_DRIVER undefined, LibreNMS loads with the missing poller and python wrapper error, which has been discussed in other help threads already submitted.