Skip to content

SCHED-962: Set finite default SuspendTime for Slurm power-save - #2586

Merged
Uburro merged 1 commit into
soperator-release-3.0from
SCHED-1848/0
Jun 8, 2026
Merged

Uburro merged 1 commit into
soperator-release-3.0from
SCHED-1848/0

Conversation

@Uburro

@Uburro Uburro commented Jun 8, 2026 •

Copy link
Copy Markdown
Collaborator

Problem

SuspendTime=-1 maps to Slurm INFINITE, which disables power-save when partitions do not define their own finite SuspendTime. As a result, scontrol power down <node> cannot trigger SuspendProgram for ephemeral nodes.

Solution

Changed the default SuspendTime to 1000000000, a large finite value that keeps Slurm power-save enabled while effectively avoiding automatic idle power down in normal operation. Updated the CRD documentation and added a render test for this value.

Testing

Manually on the live cluster
Ran:

go test ./internal/render/common

Release Notes

Changed the default Slurm SuspendTime to a large finite value so ephemeral node power-save commands can trigger SuspendProgram reliably. No migration is required; users can still override spec.slurmConfig.suspendTime explicitly.

@Uburro
Uburro merged commit 41f6a2f into soperator-release-3.0 Jun 8, 2026
40 of 61 checks passed
@Uburro
Uburro deleted the SCHED-1848/0 branch June 8, 2026 18:16

This branch was previously deployed

1 inactive deployment
e2e — 6f5a0a2c Deployed Jun 8, 2026 by Uburro via e2e-test #6873
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

2 participants