Unambiguous lookup of "preferred machine" for same amount of multiple machine types
|Target version:||Current Sprint|
As a user of /tests/overview I want to have consistent display of test names to not confuse a changing name with a differing machine type
- AC1: On reload of the same URL the display of test names should not differ on a random base
- investigate the reason for the seemingly "random behavior", probably when there are as many tests of one machine type as for the other
- make it unambiguous in any way, e.g. just always select the "first machine type" if that is also not random
original observation by ktsamis:
Display of hostname jumps randomly between scheduled tests and switches the test in another row
qam-minimal@s390x-zVM-hsi-l3 and qam-minimal-full scheduled
and on reload (maybe a couple reloads)
qam-minimal and qam-minimal-full@s390x-zVM scheduled
and that causes the test with the hostname to be moved into a new row.
I am attaching screenshots of both cases.
#1 Updated by okurz about 3 years ago
- Subject changed from Display of hostname jumps randomly between scheduled tests and switches the test in another row to Unambiguous lookup of "preferred machine" for same amount of multiple machine types
- Description updated (diff)
- Category set to Feature requests
- Target version set to Milestone 6
#5 Updated by ktsamis about 1 year ago
It this still happening? Can you provide a more recent example? Because the error is not easy to spot in the source code.
I have no idea. I opened this 2 years ago and I haven't done much openqa since then. I don't have any instance to check this. I can't provide more info on this.
https://openqa.suse.de/tests/overview?distri=sle&version=15&build=%3A10940%3Asystemd&groupid=159 - reload the page to see the Server-DVD-Incidents-Install flavor jump
- Assignee set to mkittler
That one I can reproduce locally. Sometimes
qam-incidentinstall@zkvm is in its own row and sometimes shown together with
qam-incidentinstall. Which one is the behavior we want? I'd say it should be shown together to use the space more efficiently.
- Status changed from New to In Progress
- Target version changed from Ready to Current Sprint
When looking at the code I understand now why the machines are sometimes rendered differently. There is no general way to be always in favor of one of the variants (which are that s390 gets its own row or that it is shown in the same row as the other archs). But at least one of the variants is now used consistently.
I'm wondering what the purpose of this "preferred machine" behavior is (https://github.com/os-autoinst/openQA/pull/224 has not much info on it). At least here is seems arbitrary that we sometimes append the machine and sometimes not - even though this arbitrary decision now done in a consistent way with my PR. The most space efficient way would be removing that "preferred machine" feature completely and just show all archs for the same test in the same row.