CODEWITHSUNDEEP
kubernetesopenshiftokd
mirekphd
mirekphd

Reputation: 6743

OKD 4.x: how to change status of *-deploy pods from Completed to Terminating or otherwise delete them (as in OCP/OKD 3.11)?

In OKD 4.x, when using DeploymentConfigs and their superset - Templates to deploy apps, after a Pod named <POD> has changed status to Running a auxiliary pod "deployer" named <POD>-deploy lingers on forever with the Completed status. This was not the case in OCP 3.11 and OKD 3.11, when these pods were auto-deleted. How to reproduce that functionality in OKD 4.x?

This is how it works in OCP 3.11 - notice what is happening at the end with the *-deploy pod:

NAME                                      READY     STATUS    RESTARTS   AGE
ml-admin-vsc-node2-1-mtjz5   1/1       Running   0         7s
ml-admin-vsc-node2-1-tdr4g   0/1       Terminating   1         8d
ml-admin-vsc-node2-1-deploy   0/1       Completed   0         13s
ml-admin-vsc-node2-1-deploy   0/1       Terminating   0         13s

The reasons to try to reproduce this OCP (Openshift) / OKD 3.11 behavior are many. The non-running pods share the best part of their name with the running ones, and thus need to be filtered out from pods lists when using CLI (oc=kubectl), count towards various limits, such as the 250 pods-per-node limit and the IPs addresses pool (because the IPs stay allocated to these non-running pods, which is possibly even a bug), and negatively affect cluster performance (e.g. by slowing down metadata processing in etcd).

Upvotes: 2

Views: 49

Answers (0)

Related Questions