Skip to content

Delete ate.scheduler.eligible_workers from resume path #1802

Closed

Description

@JeffLuoo

Decision

Delete ate.scheduler.eligible_workers. A log at the rejection answers the question better and costs nothing on the success path.

Why

It is expensive on the resume path. The metric builds a second slice of matching workers that scheduling never reads, walks it again, and repeats the room check for each entry. That check parses resource quantity strings up to three times for each worker. The metric roughly doubles the per-worker arithmetic of a placement.

A metric cannot answer the question it was built for. #564 wanted to tell a full fleet apart from an empty constraint intersection, such as a cordoned node behind a node requirement. That needs the selector, the node requirement and the actor. substrate.yaml forbids those as metric labels and says to put them on logs and spans.

Activity

  1. added this to the M3 milestone on Sep 22, 2026
  2. added
    kind/designDesign discussion, investigation, or research required before implementation
    on Sep 22, 2026
  3. JeffLuoo commented on Sep 22, 2026

    @JeffLuoo
    CollaboratorAuthor
  4. changed the title [-]Remove ate.scheduler.eligible_workers from the resume path[/-] [+]Delete ate.scheduler.eligible_workers and diagnose scheduling rejections from logs[/+] on Sep 22, 2026
  5. changed the title [-]Delete ate.scheduler.eligible_workers and diagnose scheduling rejections from logs[/-] [+]Delete ate.scheduler.eligible_workers from resume path[/+] on Sep 22, 2026
  6. added a commit that references this issue on Sep 22, 2026
    9e503c4
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Labels

area/observabilitykind/designDesign discussion, investigation, or research required before implementation

Type

Projects

No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions