action #111161
closedcoordination #103323: [epic] BCI testing
BCI: Combine podman and docker jobs into single job
0%
Description
Currently, we are testing images with podman and docker using different jobs:
https://openqa.suse.de/tests/overview?distri=sle&version=15-SP3&build=_15-SP3_26.9_minimal&groupid=409
which sums up to ~40 jobs per image...
At the beginning there was only base image, and not many updates. Today we have 20+ different images, so we could find a way to not spam openQA with so many jobs.
Since we need to boot different Host versions to validate each container image, we can't group those jobs. However, we could group the jobs using the same qcow.
e.g. bci_on_SLES_15-SP3_host_podman and bci_on_SLES_15-SP3_host_docker could be just bci_on_SLES_15-SP3_host and we run docker and podman tests in the same job. If we see that there are problems when using podman and docker at the same time, we can also use snapshots in the jobs, but running docker and podman together should be supported.
This way, we can tell which engines to run from the container-release-bot. For example, init and pcp images are not supported with docker and we are just running the jobs and skipping the test, it would be more efficient to not run docker tests at all.
Updated by jlausuch over 2 years ago
- Related to action #112754: Combine Kubernetes tests for BCI images into single job added
Updated by jlausuch over 2 years ago
- Subject changed from BCI: Reduce number of jobs triggered per container image by grouping jobs together to BCI: Combine podman and docker jobs into single job
- Description updated (diff)
Updated by pdostal over 2 years ago
In my opinion we should not run neither install Podman and Docker together.
Updated by jlausuch over 2 years ago
pdostal wrote:
In my opinion we should not run neither install Podman and Docker together.
In container image tests, we are already doing this in the autoyast profiles that create the HDDs:
https://github.com/os-autoinst/os-autoinst-distri-opensuse/blob/master/data/containers/autoyast_containers.xml.ep#L112
and there was never a problem.
Updated by jlausuch over 2 years ago
- Status changed from Workable to In Progress
Updated by mdati over 2 years ago
- Status changed from In Progress to Feedback
After studied the needed changes, by means of a simplified mockup version, the current evidence is that it would imply to modify code parts in many modules (mainly main_containers.pm, bci_prepare.pm, bci_repo.pm, bci_test.pm, host_configuration.pm) basically because of inter-dependencies with the CONTAINER_RUNTIME parameter.
A double check on the status is needed.
Updated by jlausuch over 2 years ago
- Status changed from Feedback to In Progress
- Assignee changed from mdati to jlausuch
I'm taking this over.
Updated by jlausuch over 2 years ago
- Status changed from In Progress to Resolved