Pod is Dead. Container is up in K8s 1.13

8/30/2019

Has anyone faced a scenario where the pod gets evicted from one node(ie Node A) and then is immeditaley scheduled on another node(ie Node B). But the kubelet of Node A keeps complaining about not being able to delete the container of the pod. The container is up & is not killed.

Aug 30 20:29:36 staging-node-4 kubelet[2173]: I0830 20:29:36.358238    2173 kubelet_pods.go:1073] Killing unwanted pod "thanos-compactor-0"
Aug 30 20:29:36 staging-node-4 kubelet[2173]: I0830 20:29:36.362581    2173 kuberuntime_container.go:559] Killing container "docker://b22287cd406c3fe9eff4ff2df1792c6f84b5b92d001359f05ea73f8788715609" with 30 second grace period
Aug 30 20:29:36 staging-node-4 kubelet[2173]: E0830 20:29:36.363416    2173 kuberuntime_container.go:71] Can't make a ref to pod "thanos-compactor-0_thanos(23096418-c7b2-11e9-9be7-005056bc2883)", container thanos: selfLink was empty, can't make reference
Aug 30 20:29:38 staging-node-4 kubelet[2173]: I0830 20:29:38.362191    2173 kubelet_pods.go:1073] Killing unwanted pod "thanos-compactor-0"
Aug 30 20:29:38 staging-node-4 kubelet[2173]: I0830 20:29:38.366884    2173 kuberuntime_container.go:559] Killing container "docker://b22287cd406c3fe9eff4ff2df1792c6f84b5b92d001359f05ea73f8788715609" with 30 second grace period
Aug 30 20:29:38 staging-node-4 kubelet[2173]: E0830 20:29:38.367901    2173 kuberuntime_container.go:71] Can't make a ref to pod "thanos-compactor-0_thanos(23096418-c7b2-11e9-9be7-005056bc2883)", container thanos: selfLink was empty, can't make reference
Aug 30 20:29:38 staging-node-4 kubelet[2173]: I0830 20:29:38.997510    2173 setters.go:72] Using node IP: "172.17.21.146"
Aug 30 20:29:40 staging-node-4 kubelet[2173]: I0830 20:29:40.364388    2173 kubelet_pods.go:1073] Killing unwanted pod "thanos-compactor-0"
Aug 30 20:29:40 staging-node-4 kubelet[2173]: I0830 20:29:40.368226    2173 kuberuntime_container.go:559] Killing container "docker://b22287cd406c3fe9eff4ff2df1792c6f84b5b92d001359f05ea73f8788715609" with 30 second grace period
Aug 30 20:29:40 staging-node-4 kubelet[2173]: E0830 20:29:40.369235    2173 kuberuntime_container.go:71] Can't make a ref to pod "thanos-compactor-0_thanos(23096418-c7b2-11e9-9be7-005056bc2883)", container thanos: selfLink was empty, can't make reference

b22287cd is the docker containerid that is not deleted.

-- swetad90
docker
kubelet
kubernetes

1 Answer

8/30/2019

I experienced this issue a few weeks ago. Pods were evicted from a node and sometimes were 'stuck' in terminating state.

Check your pod state by running kubectl -n <NAMESPACE> get pod

Using a kubectl delete pod <PODNAME> --grace-period=0 --force --namespace <NAMESPACE> resolved the problem.

-- DavidPi
Source: StackOverflow