fesco

#3519 Do we need an ability to set resource limits on builds?
Closed: Insufficient data by decathorpe. Opened by salimma.

Closed: Insufficient data
Reopen Issue

e.g.
* limit of concurrent builds per packager / SIG
* disallow concurrent builds of the same NEVRA

Context as of posting a lot of builds are blocked because 11 duplicated JDKs are being built on all the s390x builders

❯ koji list-tasks --arch s390x | grep buildArch | grep -v FREE | grep java | sed -e 's|.*buildArch|buildArch|g' | sort | uniq -c
      1 buildArch (java-latest-openjdk-portable-25.0.1.0.8-1.rolling.fc43.src.rpm, s390x)
      2 buildArch (java-latest-openjdk-portable-25.0.1.0.8-2.rolling.el10_2.src.rpm, s390x)
      4 buildArch (java-latest-openjdk-portable-25.0.1.0.8-2.rolling.el8.src.rpm, s390x)
      2 buildArch (java-latest-openjdk-portable-25.0.1.0.8-2.rolling.el9.src.rpm, s390x)
      2 buildArch (java-latest-openjdk-portable-25.0.1.0.8-2.rolling.fc43.src.rpm, s390x)

Note that almost all/all of these were scratch builds. I canceled ones that had already failed on other arches.

This tends to happen with distgit PR scratch builds since existing builds aren't cancelled when new commits are pushed, and they can be multiplied with combinations of Fedora CI, Zuul, and/or Packit Distgit CI being enabled for some packages.

yeah, but for that reason, ci builds on s390x are sent to a channel that only has a subset of builders.
These were not ci builds.

Is there anything that we should do here (other than maybe ask people not to submit lots of long-running scratch builds, and if they really need to do that, use --background)?

We discussed this during today's meeting (log starting at ~18:28 UTC).

We acknowledge that resource starvation is an issue for contributors, but that there is currently no mechanism available to limit what resources are available to users. In extreme cases, release engineering can cancel tasks / builds that starve other contributors of resources, but this is more of a workaround than a solution. In general, we would ask users who submit larger numbers of test builds to submit them with lowered scheduling priority (i.e. the --background arguments for koji build / fedpkg build) to alleviate this problem.

@salimma will file an RFE for koji to discuss possible improvements.

Metadata Update from @decathorpe:
- Issue close_status updated to: Insufficient data
- Issue status updated to: Closed (was: Open)
Metadata
None
None
None
None
None
Static archive. pagure.io was sunset in 2026; this is a read-only snapshot.
Red Hat Logo

Fedora is sponsored by Red Hat. Learn more about the relationship between Red Hat and Fedora »

© Red Hat, Inc. and others.