action #8082
closedcalculate overall test result in os-autoinst or worker?
0%
Description
os-autoinst doesn't know the overall status of a run anymore, it only uploads individual module results the openqa and that one calculates the result when the job is set to done.
When uploading hard disk images however one needs to know the overall status as images shouldn't be uploaded if the test completed but failed
Updated by coolo about 9 years ago
I don't want to duplicate the logic - but how about querying openQA for the result before uploading?
Updated by coolo almost 9 years ago
I added the job_result to the update_status call, so the worker knows the overall state:
https://github.com/os-autoinst/openQA/pull/499
Updated by okurz about 8 years ago
- Status changed from New to Resolved
IMHO this is solved. The whole communication protocol was reworked and the status reporting is fine. Also, images to be published are only uploaded for passed jobs.
Updated by coolo about 8 years ago
says who? openQA worker will happily upload images for failed jobs because it doesn't know if the job failed
Updated by okurz about 8 years ago
- Status changed from Resolved to In Progress
Why don't we have images as assets for failed jobs in the webUI then? Because they are discarded after uploading? If this is the case then it would be a good starting point to preserve qcow images of failed jobs for investigation purposes on demand.
Updated by oholecek about 8 years ago
os-autoinst will not mark the image for upload in case of fatal failure. But will mark it ready to upload in case of failed important test. As of os-autoinst#PR597
Updated by coolo about 8 years ago
we don't want to upload them as the STARTS_AFTER tests should only run on working images. And we don't upload failed images for investigation as we rollback failing job modules - make any hints disappear.
Updated by okurz about 8 years ago
coolo wrote:
we don't want to upload them as the STARTS_AFTER tests should only run on working images. And we don't upload failed images for investigation as we rollback failing job modules - make any hints disappear.
I don't see a general problem with that. My idea was to provide some kind of variable, like we have with KEEPHDDS, which can be used to make all modules fatal, not trigger START_AFTER tests and publish the image maybe even to a special "investigation" space pool with it's own quota where we can store these.
Updated by coolo about 7 years ago
- Priority changed from Normal to Low
- Target version set to Ready
this became less relevant as we stopped overwriting disk images on upload, but we still want to avoid uploading failed images. It just doesn't sound too important at this point
Updated by szarate about 7 years ago
- Status changed from In Progress to Feedback
Nobody is working on this
Updated by okurz over 5 years ago
- Category changed from 140 to Feature requests
Updated by okurz over 5 years ago
- Status changed from New to Resolved
so I guess we can state it was done with https://github.com/os-autoinst/os-autoinst/pull/597 . Also there is another, dedicated ticket for the feature request to upload images on demand even from failed jobs (but also not trigger downstream jobs on them)