Pipeline fails using k8s/PVs/PVCs/Helm?


I ran % helm install --name cc <clone: stable/concourse>. This is running on a self hosted k8s 1.14 cluster with a master and two workers. I have an nfs server available providing three PVs.

Pods come up. I can log in and create a pipeline. The pipeline has a git-resource (trigger) that is working. It also has a simple repo with a python app and a pytest script.

"git-resource trigger" ------------> "pytest job"

Persistence Disabled:
Only postgres is using a PVC. The job runs tests successfully.

Persistence Enabled:
The workers use PVCs. When the latter job runs, it fails with “waiting for docker to come up…”, This is repeated several times and then terminates.

What is causing this? Is there a fix?