Closed shashankv02 closed 3 years ago
@shashankv02 sorry for the long delay - yes, this sounds like it could be a bug.
That said, if you're not actually performing an upgrade from host-local to Calico IPAM, you can just remove this init container entirely, since it's not needed otherwise.
@caseydavenport I'm not performing an direct upgrade but I did uninstall the previous host-local installation(deleted all calico pods, ran kubeadm reset) and then tried the calico-ipam based installation on new k8s installation. But looks like the old /var/lib/cni directory and iptables rules from previous installation are not deleted automatically. After removing that directory and flushing the iptables manually, I didn't see this error.
Is there a guide to cleanly uninstalling calico before attempting to re-install it?
Is there a guide to cleanly uninstalling calico before attempting to re-install it?
No, I don't think such a document exists today. Seems like a useful thing to have.
When using calico v3.9.5 on k8s v1.17.2 cluster, the calico-node pod is stuck in crashloopbackoff. The upgrade-ipam container in the pod is crashing with panic.
Suspecting this error log - https://github.com/projectcalico/cni-plugin/blob/master/pkg/upgrade/migrate.go#L114
Interfaces
calico-node manifest:
Your Environment