I ve got an odd problem pods fail to connect to kubeapi 10 4 Rancher Users #k3s

I've got an odd problem; pods fail to connect to k...

prehistoric-gpu-28362

09/11/2025, 9:14 AM

I've got an odd problem; pods fail to connect to kubeapi 10.43.0.1 on first attempt due to connection refused. I can connect from a pod I've created to test but, the pods that access the api seem to fail first time with

Copy code

dial tcp 10.43.0.1:443: connect: connection refused"

then work on restart/retry, any ideas?

prehistoric-gpu-28362

09/11/2025, 9:36 AM

My suspicion is (as we're on an old k3s) the firewall rule isn't getting put in place in time so first attempt fails, then works but short lived pods that don't retry just fail once

prehistoric-gpu-28362

09/11/2025, 12:19 PM

I've got a workaround which I suspect will be negated when we have time to upgrade k3s:

Copy code

spec:
          containers:
          - args:
            - >-
              /bin/sleep 2 && longhorn-manager -d recurring-job snapshot
              --manager-url <http://longhorn-backend:9500/v1>
            command:
            - /bin/sh
            - "-c"
            env:
            - name: POD_NAMESPACE

the

sleep

gives enough time for iptables to be updated before the pod tries to connect to kubeapi

2 Views

Open in Slack

Previous Next