The charm is trying to install docker-ce in auto mode

[Andrey Grebennikov]

Installing CDK with docker runtime.

Docker subordinate is included with no option, so docker_runtime is set to "auto".

When checking the nodes logs, they are trying to install docker-ce packages, however the additional repository has not been set up.

unit-dockerd-1: 18:48:22 DEBUG unit.dockerd/1.install docker.io was already not hold.
unit-dockerd-1: 18:48:22 INFO unit.dockerd/1.juju-log Marking ['docker-ce'] as unhold
unit-dockerd-1: 18:48:22 DEBUG unit.dockerd/1.install E: Can't select installed nor candidate version from package 'docker-ce' as it has neither of them
unit-dockerd-1: 18:48:22 DEBUG unit.dockerd/1.install E: No packages found
unit-dockerd-1: 18:48:22 INFO unit.dockerd/1.juju-log Marking ['docker-ce', 'nvidia-docker2', 'nvidia-container-runtime', 'nvidia-container-runtime-hook'] as unhold
unit-dockerd-1: 18:48:23 DEBUG unit.dockerd/1.install E: Can't select installed nor candidate version from package 'docker-ce' as it has neither of them
unit-dockerd-1: 18:48:23 DEBUG unit.dockerd/1.install E: Unable to locate package nvidia-docker2
unit-dockerd-1: 18:48:23 DEBUG unit.dockerd/1.install E: Unable to locate package nvidia-container-runtime
unit-dockerd-1: 18:48:23 DEBUG unit.dockerd/1.install E: Unable to locate package nvidia-container-runtime-hook
unit-dockerd-1: 18:48:23 DEBUG unit.dockerd/1.install E: No packages found

Looking at the code here https://github.com/charmed-kubernetes/charm-docker/blob/master/reactive/docker.py#L125

and trying to run the same lspci command in the debug mode - "out.decode('utf-8').lower().count('nvidia')" comes out as "0".

[George Kraft (cynerva)]

We need reproduction steps or full output from `juju status --format yaml` and `juju debug-log --replay` for this one.