Load limiting #603
Labels
No labels
Area/build-packaging
Area/cli
Area/evaluator
Area/fetching
Area/flakes
Area/language
Area/profiles
Area/protocol
Area/releng
Area/remote-builds
Area/repl
Area/store
bug
crash 💥
Cross Compilation
devx
docs
Downstream Dependents
E/easy
E/hard
E/help wanted
E/reproducible
E/requires rearchitecture
imported
Needs Langver
OS/Linux
OS/macOS
performance
regression
release-blocker
RFD
stability
Status
blocked
Status
invalid
Status
postponed
Status
wontfix
testing
testing/flakey
ux
No milestone
No project
No assignees
3 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference: lix-project/lix#603
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Is your feature request related to a problem? Please describe.
It's currently not possible to configure a load limit for builds. For a long time the stdenv limited load to the same as cores, but this default restriction was removed.
Three attempts have been made to solve this in CppNix, some appearing to be ready, but they got stuck in review.
https://github.com/NixOS/nix/issues/7091
https://github.com/NixOS/nix/pull/11143#issuecomment-2385642400
I don't have a clear description of what I want, I'm not sure what a good solution is.
I think the above linked PR demonstrates a good solution assuming stdenv gets support added.
In terms of build scheduling, if this is made responsibility of lix (i am not convinced it should necessarily be, given we have enough on our plate as it is) rather than something like Hydra (which, to be clear, has its own giant tech debt pile), it would require protocol changes to communicate the builder load and likely some serious redesign of the way that build-remote (and build-hook) works.
Arguably, I think this is yet another hack and https://github.com/NixOS/nixpkgs/pull/314888 is the real solution here.
Unfortunately, we are stuck on upstreams not being mature enough so we could really benefit from a jobserver. If somefew could push the state of the art over there and bring forth a good protocol, we could get lovely performance benefits.