QA (public) » openQA Project (public) » openQA Tests (public) » qe-yam

After reviewing the solution proposed in draft: https://gitlab.suse.de/qe-yam/openqa-job-groups/-/merge_requests/163/diffs
we came to the conclusion that verifying the image existence would add complextity and difficulty maintaining the code of the CI. Additionally, using variable syntax like {{my_var}} should be avoided (in slack discussion: https://suse.slack.com/archives/C02CLSHS3SB/p1714467732188989 "A better choice would be ${{ ... }}. The problem is that { is special in YAML because it creates a flow mapping (it's basically just like in jSON). With putting $ at the start you avoid having to quote it". )

Proposed solutions :
1 Using a similar way as in the above draft, in order to replace a variable only for the image build, according to the latest leap build, each time the CI runs for the particular job group. Simplifying the bash script as:

get_latest_build() {
    local latest_build
    local leap_version
    mapfile -t -n 2 latest_build < <(curl -L -sS "openqa.opensuse.org/group_overview/50.json" \
         | jq -r '([.build_results[] | select(.version=="15.6") | .build] | sort | reverse)[]')
    if [[ "${#latest_builds[@]}" -lt 1 ]]; then
        warn "Unable to find latest Leap builds."
        exit 1
    fi
}

This would require a scheduled pipeline that will run the particular job once every week/biweekly, because we cannot guarantee that there will be a merge request editing the relevant job group often enough.

2 Another approach: Instead of adding a job in the CI jobs that run while merging or committing code, we can create just a scheduled pipeline that will call a script to edit directly the master branch files that include leap image dependent jobs and will merge the code itself. See bellow useful links:
https://stackoverflow.com/questions/66941947/gitalb-ci-cd-update-a-file-on-gitlab-repository-after-modification-on-gitlab-run
https://stackoverflow.com/questions/65816737/gitlab-ci-scheduled-pipeline-not-running-on-default-branch
https://docs.gitlab.com/ee/ci/pipelines/schedules.html

This solution is cleaner as it will not affect pipelines while we create our own merge requests and we will avoid adding variables in the leap image names in the job group yaml files. The master branch will contain for example:
HDD_1_URL: 'https://openqa.opensuse.org/tests/3976134/asset/hdd/opensuse-15.6-x86_64-623.2-textmode@64bit.qcow2'
HDD_1: opensuse-15.6-x86_64-623.2-textmode@64bit.qcow2
Instead of
HDD_1_URL: 'https://openqa.opensuse.org/tests/3976134/asset/hdd/opensuse-15.6-x86_64-${{latest_leap_build}}-textmode@64bit.qcow2'
HDD_1: opensuse-15.6-x86_64-${{latest_leap_build}}-textmode@64bit.qcow2

The build number will be updated every time the scheduled pipeline runs to the latest leap build number.

3 The fastest and simplest solution would be to add a cron job that would edit the osd test suite data base directly, in order to keep using test reference in our job group , but the possible complications are unknown.

Back to new after discussing it today in the call.