[svcnemo-autobot]

Background

Partial workflow reruns inherit outputs from successful prerequisite jobs, which can retain an obsolete attempt-qualified runner label.

What changed

Pinned the routing helper to v1.9.2, passed runner_label_base through preflight, and appended github.run_attempt at each rerun-eligible Curator scheduling boundary.

Details

Legacy labels remain unchanged because the attempt suffix is added only for ephe-v2- labels.

flowchart LR
  R[Routing base] --> P[Preflight]
  P --> J[Runner consumer]
  A[Current attempt] --> J

Loading

Tested

• Parsed the changed workflow as YAML
• Verified all three runner consumers use the guarded expression
• Verified the preflight pin stayed unchanged
• Ran applicable YAML, whitespace, size, case, and private-key checks

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with no concrete correctness, security, or repository-rule violations identified.

Summary

• Pins the routing reusable workflow to v1.9.2 and passes its base runner label through preflight.
• Appends the current run-attempt suffix at each affected Curator runner consumer.
• Preserves legacy runner labels through an ephe-v2- prefix guard.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  Routing["_ephemeral_runner_routing.yml v1.9.2"] -->|"runner_label_base"| Preflight["Preflight"]
  Preflight -->|"runner_prefix"| Guard{"Starts with ephe-v2-?"}
  Attempt["github.run_attempt"] --> Guard
  Guard -->|"Yes: append -aN"| Consumers["Container build and GPU jobs"]
  Guard -->|"No: preserve label"| Consumers

Loading

[svcnvidia-nemo-ci]

/ok to test 3c0a3f9