[BUG] epicli upgrade of cluster created by Epiphany v0.5 may fail on [kubernetes_master : Patch deployment: coredns] task

[to-bar]

Describe the bug

The issue appeared on AWS after upgrading Docker and is non-deterministic.

Error message:

INFO cli.engine.ansible.AnsibleCommand - TASK [kubernetes_master : Patch deployment: coredns]
ERROR cli.engine.ansible.AnsibleCommand - fatal: [ec2-34-244-180-113.eu-west-1.compute.amazonaws.com]: FAILED! => {"changed": true, "cmd": ["kubectl", "patch", "deployment", "coredns", "--namespace", "kube-system", "--type", "strategic", "--patch", "{\"spec\": {\"template\": {\"spec\": {\"containers\": [{\"name\": \"coredns\", \"resources\": {\"requests\": {\"ephemeral-storage\": \"2Mi\"}}, \"volumeMounts\": [{\"name\": \"hosts-volume\", \"mountPath\": \"/etc/hosts\", \"readOnly\": true}]}], \"volumes\": [{\"name\": \"hosts-volume\", \"hostPath\": {\"path\": \"/etc/hosts\", \"type\": \"File\"}}]}}}}"], "delta": "0:00:00.091792", "end": "2020-12-03 09:50:52.285496", "msg": "non-zero return code", "rc": 1, "start": "2020-12-03 09:50:52.193704", "stderr": "Error from server (NotFound): the server could not find the requested resource", "stderr_lines": ["Error from server (NotFound): the server could not find the requested resource"], "stdout": "", "stdout_lines": []}

How to reproduce
Steps to reproduce the behavior:

1. Create cluster with Epiphany v0.5
2. Run epicli upgrade using code from a branch with upgraded Docker to v19.03.14.

Expected behavior
Upgrade should succeed.

Environment

• Cloud provider: AWS
• OS: Ubuntu, RHEL

Additional context
Issue not seen when using Calico.

[przemyslavic]

Upgrade to develop tested automatically in the pipeline.
✔️ AWS/Azure x RHEL/Ubuntu x flannel/calico/canal