systemctl restart configure-agent-env.service fails on master and worker node

[Narinder Gupta]

Once the magnum can create the cluster. sometimes master and worker nodes are in an unhealthy state. Log into the any master or worker node and check the status.

systemctl status configure-agent-env.service service is in failed state
systemctl restart configure-agent-env.service service is in failed state because mkdir /etc/kubernetes fails states /etc/kubernetes already exist.

When nodes are in a healthy state.
ssh into any node.

systemctl restart configure-agent-env.service fails because mkdir /etc/kubernetes fails states /etc/kubernetes already exist.

edit file configure-agent-env.sh
change mkdir /etc/kubernetes/ to mkdir -p /etc/kubernetes/ and service can be restarted.

this issue causes customers to be unable to list the cluster as healthy even though the kubectl command gets the container.

[Felipe Reyes (freyes)]

After updating the file fcct-config.yaml it’s necessary to update the “compiled” version in user_data.json with the following command:

podman run --rm -v $(pwd)/fcct-config.yaml:/config.fcc quay.io/coreos/fcct:release --pretty --strict /config.fcc > ./user_data.json

I'm testing the change in a lab environment, will follow up with the results.

[Felipe Reyes (freyes)]

I tested successfully the package ( 10.0.0+git2021031516.2f217810-0ubuntu1.1+ppa3 ) available at https://launchpad.net/~freyes/+archive/ubuntu/focal-ussuri .

Narinder, could you give it a try in your testing environment?.

Thanks,

testing log: https://pastebin.ubuntu.com/p/GWT5bssQXs/