containerd isn't restarted when config changed

[NicolasT]

I have a node on which containerd was already installed and running, but with no specific configuration file. When running bootstrap.sh, this fails at some point because our fake registry config isn't taken into account.

It appears when we drop a (new) config file for containerd, we don't properly restart the service.

[TeddyAndrieux]

@NicolasT Which version did you try ?
I think it's fixed in 2.7.0+ (side effect of some rewritting in this commit 5e1a1a2 PR #3020 )

[NicolasT]

2.7.1

[alexandre-allard]

I can't reproduce this issue on 2.7.3.

I've installed containerd prior to launch bootstrap.sh and everything went well.
Looking at what's done in the formula, I can't see any issue, we're triggering a restart if the configuration is changed:

Configure registry IP in containerd conf:
  file.managed:
    - name: /etc/containerd/config.toml
    - makedirs: true
    - contents: |
        [plugins.cri.registry.mirrors."{{ repo.registry_endpoint }}"]
        endpoint = ["http://{{ registry_ip }}:{{ registry_port }}"]

        [debug]
        level = "{{ 'debug' if metalk8s.debug else 'info' }}"
    - watch_in:
        - service: Ensure containerd running

Ensure containerd running:
  service.running:
    - name: containerd
    - enable: True
    - init_delay: 2

[NicolasT]

I assume you installed and started containerd before bootstrap? If so, I guess we can close this as not reproducible 🤷‍♂️

[alexandre-allard]

Yes, that's what I did.

[alexandre-allard]

Closing this issue as it does not seem to be reproducible on latest versions.