#307 more fair scheduler for handling the build queue?
Closed: Fixed 3 years ago by praiskup. Opened 5 years ago by praiskup.

I'm waiting in queue for the second time today and many builders are wasted in builds of @realtime/krenel, where are probably enabled all available chroots :-) and where are e.g. run four builds in parallel (760763 760745 760706 760608). The point is that each chroot holds one VM for more than two hours.

  • could we somehow take the size of queue into account (and allow e.g. only 4 parallel builds for one user if the queue is large enough)

  • could we somehow count the time period dedicated to one user, and apply some intelligent quota?


Btw., weird situation here:
http://praiskup.fedorapeople.org/Screenshot_20180530_195213.png

Isn't there some problem with filtering webhook requests? It's wasteful to submit several builds for the same git-hash at the same time (if that's user-error).

@lkundrak do you think that is a bug in Copr?

No. We're using the "Custom" webhook instead of the GitHub one -- that one doesn't seem to work at all. Worse even it pretends to succeed returning a 200 status without any clues. My guess was that it's just not supposed to be used with "Custom build method" but I've never got around to actually reporting the issue. Perhaps @msuchy can tell what went wrong?

The unhappy consequence is that we now get triggered for a push into any branch but the build script has no idea why was it triggered and just proceeds to build master.

I'm disabling my builds now and going back to triggering the manually since my experiment seems to have fallen flat. Sorry for hogging the builders.

My guess was that it's just not supposed to be used with "Custom build method"

that's correct

The unhappy consequence is that we now get triggered for a push into any branch but the build script has no idea why was it triggered and just proceeds to build master.

The custom hook content can be parsed directly by your script, and you can detect what is happening. The bad thing is that there's no "API" to tell -- from the custom script -- that you want to skip the build..., (yet).

I'm disabling my builds now and going back to triggering the manually since my experiment seems to have fallen flat. Sorry for hogging the builders.

Stopping auto-rebuilds is not desirable; to not dive too much OT, can you ping me on irc so we can discuss how to better trigger the re-builds?

Metadata Update from @msuchy:
- Issue tagged with: RFE

5 years ago

Metadata Update from @praiskup:
- Issue marked as blocking: #817

4 years ago

Fair scheduler is deployed into production.

Metadata Update from @praiskup:
- Issue close_status updated to: Fixed
- Issue status updated to: Closed (was: Open)

3 years ago

Login to comment on this ticket.

Metadata