QA (public) » openQA Project (public)

Observation¶

Since recently t/04-scheduler.t is very unstable.
https://app.circleci.com/pipelines/github/os-autoinst/openQA/4269/workflows/ea59d97b-61bd-4e86-b593-3401b6678a7e/jobs/40930/steps shows

not ok 11 - job grab (WORKER_CLASS mismatch)
Subtest: job grab (WORKER_CLASS mismatch)
[DEBUG] +=----------------=+
[DEBUG] -> Scheduling new jobs.
[DEBUG]      Free workers: 1/1
[DEBUG]      Scheduled jobs: 2
[DEBUG] Scheduler took 0.05042s to perform operations and allocated 0 jobs
    ok 1 - job not grabbed due to default WORKER_CLASS
    ok 2 - no workers/jobs allocated
    not ok 3 - t_seen has not changed

    #   Failed test 't_seen has not changed'
    #   at t/04-scheduler.t line 306.
    #          got: '2020-09-18T19:42:26'
    #     expected: '2020-09-18T19:42:25'
    1..3
    # Looks like you failed 1 test of 3.

I can reproduce the same locally

Steps to reproduce¶

Probably reproducable locally with

make test TESTS=t/04-scheduler.t

Suggestions¶

Bisect where the regression comes from and fix that to be stable locally and within CI.