QA (public) » openQA Project (public)

https://openqa.suse.de/tests/3320973 is a more severe case of a job that had been scheduled automatically (so not a user restarting it manually). At the time the job had a chance to run the asset was already removed (osd:/var/log/openqa states "[2019-09-04T06:35:23.0481 CEST] [info] Removing asset hdd/sle-12-SP5-ppc64le-Build0303-Server-DVD@ppc64le-gnome-encrypted.qcow2 (assigned to groups: 142)")

mkittler and me tried to come up with ideas as mkittler was already working in a "neighboring code area". Our idea was in the "job_restart" API method to check for the asset existance which we can do simply with methods that the job ResultSet object already has plus utility functions, e.g. lib/OpenQA/Utils.pm:locate_asset like die 'Missing asset ' . $self->name . "\n" unless locate_asset($self->asset->type, $self->asset->name, mustexist => 1) in the low-level function and in the high-level API function catch the exception and return {error => $msg} which is already handled by javascript to show a flash message when someone tries to trigger the job.

What I thought about initial triggering, not cloning, is that we could set the jobs to cancel by checking for asset existance even before asking any workers to pick them up.