#47 RFE: allow tasks to fail on some arches for images/lives/appliances
Closed: Fixed 5 years ago Opened 6 years ago by ausil.

We have to submit livemedia and image tasks as a single task. but in the compose world we do not care if things fail on some arches. For instance FESCo has decided that 32 bit x86 is non release blocking. in order to have it truely not be release blocking when we submit a livemedia or image task we need a way to tell koji to not fail the build if i386 fails and only fail the build if x86_64 fails.

Theoretically Koji does not necessarily need to know which architectures are blocking. It might be enough to have an option to never fail the parent task even if children fail. The calling code (e.g. Pungi) would than deal with the result and check if all the required stuff is present.

It's not a clean solution by any measure, but it might be easier to implement and have something to test.

Apparently what I wrote above is actually issue #48.

https://fedorahosted.org/fesco/ticket/1469 is the fesco decision to make i686 non release blocking. there is no way we can not fail a rpm build on i686, however we need to be able to be able to have livemedia, image etc tasks that are recorded in the koji db as builds with a single parent task and children tasks for each arch to be able to fail on some of the arches. the list of failable arches should be recorded in the task info. so there should be no confusion for users as they will see in kojiweb that a arch was configured as failable if it is missing.

Secondary arches are non release blocking also so need to be able to fail also. as part of redefining secondary arches we plan to move them into the one koji, with the definition being where we ship output. There will need to be corresponding changes in pungi to account for having some arches being able to fail.

Why do you have to submit livemedia and image tasks as a single task? The correct thing to do here would be to run the non-blocking tasks separately, and let the calling code (pungi) decide whether a failure is critical or not. Letting certain subtasks fail breaks some low-level assumptions in Koji, and is also non-intuitive to users.

@mikem changed the status to Closed

5 years ago

Login to comment on this ticket.