Fresh microk8s 1.19.0 stalled, never becomes ready

Has anyone else experienced this issue, where you install a fresh microk8s on a system, but e.g. “microk8s kubectl get nodes” always reports the node as being NotReady? This is on Ubuntu 20.04, with no previous installations of microk8s or anything.

Installing 1.18.0 first, then refreshing that to 1.19.0 seems to work as expected, so there’s something seemingly broken with 1.19.0.

Hi,

1.19 introduced significant changes. Most notably are

  • the migration from flannel to calico.
  • Use of dqlite instead of etcd.

If you’re coming from 1.18, refreshing to 1.19 will continue to honor the previously running components. That could be the reason why it “works”.

It does take some time to get calico up and running. Usually longer than flannel.

Going back to why it may be reporting NotReadymaybe you have multiple interfaces preventing calico from detecting the right one. Could it be similar to this? https://github.com/ubuntu/microk8s/issues/1554

Feel free to log issue in github and attach the inspect tarball.

Thanks

I see. That explains that.

I don’t have any interfaces with “lo” in their names other than the loopback-interface itself, so no, I don’t think that issue applies. I did try the steps there anyway and they didn’t work. That said, that issue did make me think that this is also related to networking, so I did, indeed, open a new issue there.