cephadm-dashboard-e2e: replace docker with podman

[nizamial09]

• cleanup the docker installation.

[nizamial09]

i haven't tested this in the jenkins yet by updating the job configuration yet. Only done it in my local VM (which worked fine). Will wait for an approval before testing it out in the jenkins machines.

[nizamial09]

replacing docker with podman.

[dmick]

This is good as far as it goes, but I see two problems:

1. there are build hosts that have installed docker in the standard way from the standard repos, and rely on it, so "is docker present, if so remove it" is going to break those hosts

2. the broken hosts will stay broken until they happen to be selected for a dashboard run

so this isn't sufficient, and I don't have time to make it better this week

[nizamial09]

1. there are build hosts that have installed docker in the standard way from the standard repos, and rely on it, so "is docker present, if so remove it" is going to break those hosts

@dmick i've modified the condition to check whether docker is installed through the docker official repo (docker-ce) and then only remove it. Maybe its more safer than before. Again I tried this out in my local VM which succesfully worked.

[djgalloway]

Can I suggest instead you use https://github.com/ceph/ceph-build/blob/main/scripts/setup_container_runtime.sh

[nizamial09]

yup, sure. I could give it a try. thanks

[nizamial09]

sorry for the delay, but updated the PR now.

[djgalloway]

I mentioned this in DM but we might be able to do a manual one-time removal of docker and its ilk from the Jenkins builders and not worry about it here. Only the ceph-rook-e2e job still wants docker. I pinged @rkachach asking if we can retire the job as it looks abandoned.

[nizamial09]

I mentioned this in DM but we might be able to do a manual one-time removal of docker and its ilk from the Jenkins

so I removed the docker uninstall command from this PR since we just need it one time.

Also I used to test the updated config with jenkins-job update command that pushes the config temporarily into the jenkins. but it doesn't work now and fails with 403 even though i have authenticated..

  File "/usr/lib/python3.14/site-packages/requests/models.py", line 1026, in raise_for_status
    raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 403 Client Error: Forbidden for url: https://jenkins.ceph.com/job/ceph-dashboard-cephadm-e2e/config.xml

so maybe we need to merge and test this?

[djgalloway]

All kcli commands should run inside a sg libvirt -c 'kcli ...

[nizamial09]

this would require changes in https://github.com/ceph/ceph/blob/main/src/pybind/mgr/dashboard/ci/cephadm/start-cluster.sh which require a bit more testing. should i handle it separately?

I am thinking maybe a wraparound would be better.. but it needs to be added for both scripts.

asdf

[djgalloway]

Sure. Can't make things worse.