Skip to content

Rust canary on every capacity #31

Closed

Description

@sodre

Parent

#18

What to build

The live proof of #18 in a consuming repo, as #12 was for the first build. On our runners it shows:

  • a Rust build on ci-fargate-spot, and the same build on ci-spot at ci-cpu-8 / x64;
  • a job on arm64, on EC2 and on Fargate;
  • a swap repaired, with no stranded job;
  • a Fargate Spot reclaim producing its check run on the right commit;
  • a job deferred at a vCPU budget;
  • a task stopped by the sweeper.

Acceptance criteria

  • Each item above is observed, with a link to the run or log line on this ticket.
  • The runbook's "Measured numbers" has boot and build times for each capacity the canary used.
  • Any new trap found goes in the traps table.

Activity

  1. sodre commented on Nov 14, 2026

    @sodre
    MemberAuthor

    Status from the live proofs so far (integration deployed):

    Still for a person: a real zae-zcm Rust job on ci-fargate-spot and on ci-spot ci-cpu-8, and noticing a real reclaim or 2 h stop when one happens.

  2. sodre commented on Nov 14, 2026

    @sodre
    MemberAuthor

    Correction to my earlier checklist: the Rust item is done. The same consumer Rust commit (one test + eight cargo mutants jobs) ran on ci-fargate-spot (ci-cpu-8, and again at ci-cpu-2) and on ci-spot at ci-cpu-8 / x64, all green with identical mutation results; numbers per shape go into the runbook's "Measured numbers" in #42.

    Status now:

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions