Add SLURM_MEM_PER_CPU to keep_slurm variables

[ASLeonard]

I noticed my slurm jobs started being OOM all of a sudden despite using only a fraction of the allocated memory. Eventually seems to be the changes in d0c515c, where the slurm-jobset (srun) was not inheriting the memory allocation of the slurm-executor (sbatch). I have a memory value in my config default-resources, which was being used instead.

I confirmed this change correctly forwards the memory requested to the actual jobstep via sacct. Our HPC does not allow passing --mem=0, so this may a fragile approach to more generally passing on the mem/mem-per-cpu variable.

Summary by CodeRabbit

• Bug Fixes
  • Preserved the SLURM memory-per-CPU setting for nested job launches, helping child srun commands inherit the expected memory allocation.
  • Improved consistency for jobs that rely on nested SLURM execution.

[coderabbitai]

📝 Walkthrough

Walkthrough

A single environment variable, SLURM_MEM_PER_CPU, was added to the keep_slurm allowlist in Executor.run_job, so it is now preserved rather than stripped when sanitizing the environment for nested srun invocations.

Changes

Environment allowlist update

Layer / File(s)	Summary
Allowlist SLURM_MEM_PER_CPU snakemake_executor_plugin_slurm_jobstep/__init__.py	Adds SLURM_MEM_PER_CPU to the keep_slurm allowlist so it is retained for nested srun calls instead of being removed.

Estimated code review effort: 1 (Trivial) | ~2 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: preserving SLURM_MEM_PER_CPU in the keep_slurm allowlist.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (1)

snakemake_executor_plugin_slurm_jobstep/__init__.py (1)

208-220: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win

Consider also keeping SLURM_MEM_PER_NODE.

SLURM_MEM_PER_CPU and SLURM_MEM_PER_NODE correspond to the mutually-exclusive --mem-per-cpu and --mem flags respectively, and only one of them is ever set for a given job. Since this allowlist currently keeps only SLURM_MEM_PER_CPU, jobs submitted with a plain --mem allocation will still have their memory setting stripped before the nested srun, reproducing the same OOM regression this PR fixes for --mem-per-cpu jobs.

Proposed fix

             "SLURM_SUBMIT_HOST",
-            "SLURM_MEM_PER_CPU"
+            "SLURM_MEM_PER_CPU",
+            "SLURM_MEM_PER_NODE",
         }

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@snakemake_executor_plugin_slurm_jobstep/__init__.py` around lines 208 - 220,
The allowlist in the Slurm environment preservation block currently keeps only
SLURM_MEM_PER_CPU, so jobs submitted with --mem still lose their memory setting
before nested srun. Update the keep_slurm set in __init__.py to also include
SLURM_MEM_PER_NODE alongside the existing memory variable so both mutually
exclusive Slurm memory settings are preserved.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@snakemake_executor_plugin_slurm_jobstep/__init__.py`:
- Around line 208-220: The allowlist in the Slurm environment preservation block
currently keeps only SLURM_MEM_PER_CPU, so jobs submitted with --mem still lose
their memory setting before nested srun. Update the keep_slurm set in
__init__.py to also include SLURM_MEM_PER_NODE alongside the existing memory
variable so both mutually exclusive Slurm memory settings are preserved.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: ccc9a18d-6d0e-44b7-8bf7-250f16375444

📥 Commits

Reviewing files that changed from the base of the PR and between c09349d and 6795442.

📒 Files selected for processing (1)

• snakemake_executor_plugin_slurm_jobstep/__init__.py