action #51716
openNo scheduling error generated for faulty PARALLEL_WITH config
0%
Description
Steps to reproduce¶
- Define test suite A with variable PARALLEL_WITH=B
- Define test suite B with variable PARALLEL_WITH=B
- Add both tests to same job group
- Try to call
/usr/bin/openqa-client isos post [all required vars DISTRI,VERSION,FLAVOR etc.] TEST=A,B
Observation¶
Nothing happens.
Expected¶
An error message that the scheduling failed due to cycle dependencies is generated.
note: In case you will try to call /usr/bin/openqa-client isos post [all required vars DISTRI,VERSION,FLAVOR etc.]
(without TEST
var; for the whole iso) you will get the expected error, e.g.:
{
error_messages => [
"There is a cycle in the dependencies of wicked_aggregate_ref at /usr/share/openqa/script/../lib/OpenQA/Schema/Result/ScheduledProducts.pm line 691.\n",
],
job_id => 1038,
}
Updated by okurz over 4 years ago
- Subject changed from No error on scheduling faulty MM config to Make scheduling errors more accessible. There is no error on scheduling faulty MM config
- Description updated (diff)
- Category changed from Regressions/Crashes to Feature requests
- Status changed from New to Workable
- Target version set to future
Merged in content from #65142
Updated by okurz over 4 years ago
- Has duplicate action #65142: Make scheduling errors more accessible added
Updated by mkittler over 4 years ago
- Has duplicate deleted (action #65142: Make scheduling errors more accessible)
Updated by mkittler over 4 years ago
- Related to action #65142: Make scheduling errors more accessible added
Updated by mkittler over 4 years ago
- Subject changed from Make scheduling errors more accessible. There is no error on scheduling faulty MM config to No scheduling error generated for faulty PARALLEL_WITH config
- Description updated (diff)
I moved the information copied over from #65142 out again because these are actually different issues. I keep the #65142 as related issue because both are about scheduling errors.
About the issue itself: I'm wondering whether it makes sense to assume PARALLEL_WITH
dependencies can be cyclic at all. I mean, these jobs are supposed to run together at the same time making a "job cluster" in any case: If job A
is supposed to run in parallel with job B
that means implicitly that job B
is supposed to run in parallel with A
. I see no problem in specifying that explicitly. (And by the way, if B
is additionally supposed to run in parallel with C
that also implies A
is supposed to run in parallel with C
.)