This message was deleted Rancher Users #rke

Join Slack

This message was deleted.

# rke

adamant-kite-43734

09/27/2023, 11:07 PM

This message was deleted.

quaint-microphone-1347

09/27/2023, 11:12 PM

i've seen this happen in larger clusters with kubelet, and the only path i've had is to delete the node and rejoin it. I would do this with a control plane node as well, I guess -- but this is a test cluster and its running as a single node.

high-army-50818

09/28/2023, 12:36 AM

Kubelet/controller-manager being rm’d isn’t interesting and a local issue. However, a supported path for restoring a rancher launched cluster from an etcd snap without the rancher ui is interesting

microscopic-diamond-94749

09/28/2023, 10:34 AM

can you run

rke up

again @quaint-microphone-1347 ?

quaint-microphone-1347

09/29/2023, 5:51 PM

the rke/rancher

local

cluster is healthy -- its a rancher launched cluster which is effectively offline with no available control plane nodes to accept commands from rancher itself. (but I have an etcd snap)

quaint-microphone-1347

09/29/2023, 10:01 PM

current workaround is using

pip install runlike

to reconstruct some of the

docker run

command to launch kube-controller-manager -- but it doesnt parse the entrypoint correctly, which needed a bit of help. was able to get the

kube-controller-manager

up and running, which let the cluster actually start things and heal itself

Open in Slack

Previous Next