Skip to content

fix(runners): include all ten MI355X runners - #2895

Open
cquil11 wants to merge 1 commit into
mainfrom
codex/mi355x-ten-runner-capacity
Open

fix(runners): include all ten MI355X runners#2895
cquil11 wants to merge 1 commit into
mainfrom
codex/mi355x-ten-runner-capacity

Conversation

@cquil11

@cquil11 cquil11 commented Sep 8, 2026

Copy link
Copy Markdown
Collaborator

The MI355X cluster has ten online GitHub runners, but both runner inventories list only _00 through _08. Add mi355x-amds_09 to mi355x and cluster:mi355x-amds so repository configuration matches the live pool.

Validation: runner configuration schema passes; both lists exactly match the GitHub API inventory; 11 existing runner configuration tests pass. Nodes 31 and 37 passed GPU, container, and eight-link RDMA smoke checks, joined the ten-node compute partition, and received production Slurm job 42723. The deployed dashboard collector exclusions were separately removed and telemetry now reports ten nodes and 80 GPUs.


Note

Low Risk
Inventory-only YAML update with no auth, data, or runtime logic changes; misconfiguration risk is limited to job routing if the runner name were wrong.

Overview
Adds mi355x-amds_09 to the mi355x label list and cluster:mi355x-amds cluster inventory in configs/runners.yaml, bringing both entries from nine runners (_00_08) to ten so repo config matches the live GitHub Actions pool.

Reviewed by Cursor Bugbot for commit ceffec3. Bugbot is set up for automated code reviews on this repo. Configure here.

@claude claude Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good, straightforward config change.

What was reviewed: the two-line addition of mi355x-amds_09 to the mi355x and cluster:mi355x-amds label lists in configs/runners.yaml. Confirmed both lists now contain a consistent 00-09 range matching the existing mi325x/cluster:mi325x-amds pattern. Checked that runners/launch_mi355x-amds.sh exists and does not hardcode a runner count, and that no other config/script depends on the prior list length.

Extended reasoning...

Overview

The change is a purely additive, two-line edit to configs/runners.yaml, appending mi355x-amds_09 to the mi355x label array and the cluster:mi355x-amds label array. No scripts, schemas, docs, or other config files are touched.

Security risks

None. This is inventory metadata for a self-hosted GitHub Actions runner label scheme; it does not affect auth, secrets, or execution logic. Adding a runner name to a schedulable-label list only affects which jobs can be scheduled on that node, consistent with the ten already-onboarded mi355x-amds nodes described in the PR.

Level of scrutiny

Low. The change follows the exact naming/indentation pattern of the eight/nine existing entries in both lists, and both lists were kept in sync (a common source of bugs in this file if only one list is updated, which was checked and is not the case here). configs/runners.yaml is not listed under a specific CODEOWNERS entry (only amd-master.yaml and nvidia-master.yaml have named owners), so it falls under the default * @ InferenceX/core ownership with no special gating.

Other factors

No bug-hunter findings were reported for this PR, and the PR conversation timeline shows no outstanding third-party objections. I additionally confirmed runners/launch_mi355x-amds.sh exists and does not hardcode a node/runner count that this change would need to update, and found no other file (benchmark recipes, node_excludes.yaml, etc.) that depends on the length of these two label lists.

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

Labels

None yet

Projects

Status: No status

Development

Successfully merging this pull request may close these issues.

1 participant