Preinstall Soperator utility scripts to jail

[rdjjke]

Here’s a summary of the changes made in the PR:

1. Default Values for Slurm Configuration

• The default value for the field taskPluginParam in SlurmConfig is changed from an empty string ("") to Autobind=Cores in:
  • Go struct tags (api/v1/slurmcluster_types.go)
  • The generated CRD YAMLs (config/crd/bases/slurm.nebius.ai_slurmclusters.yaml, helm/soperator-crds/templates/slurmcluster-crd.yaml, helm/soperator/crds/slurmcluster-crd.yaml)

2. Slurm Scheduler Configuration

• Added CR_ONE_TASK_PER_CORE to the SelectTypeParameters in the Slurm config generation (internal/render/common/configmap.go) to improve core allocation granularity.

3. Jail Docker Image Enhancements

• Two utility scripts are added to /opt/soperator_utils in the jail container:
  • soperator_instance_login.sh: Script to SSH/chroot into the systemd namespace of a worker node.
  • slurm_task_info.sh: Prints Slurm task resource bindings (node, rank, CPU, GPU, CUDA devices).

4. User Creation Script Update

• createuser.sh now supports a --without-internal-ssh flag to skip generating an internal SSH key pair.

5. New Utility Scripts

• images/jail/scripts/slurm_task_info.sh: Prints details about Slurm task resource bindings.
• images/jail/scripts/soperator_instance_login.sh: Allows logging into the instance associated with a specific worker node or instance ID.

Summary:
The PR sets Autobind=Cores as the default for Slurm’s taskPluginParam, tweaks Slurm core scheduling, adds two admin/diagnostic scripts to the jail image, and enhances user creation with an option to skip internal SSH key generation.

[Copilot]

Pull Request Overview

This PR preinstalls Soperator utility scripts into the jail container, enhances user creation with optional internal SSH key support, and updates the Slurm CRD defaults for taskPluginParam.

• Adds two helper scripts (soperator_instance_login.sh, slurm_task_info.sh) and integrates them into the jail image
• Extends createuser.sh with a --without-internal-ssh flag to control SSH key generation
• Updates Helm charts, CRD YAMLs, and Go API types to default taskPluginParam to "Autobind=Cores"

Reviewed Changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
internal/render/common/configmap.go	Included CR_ONE_TASK_PER_CORE in SelectTypeParameters
images/jail/scripts/soperator_instance_login.sh	New script for interactive instance login
images/jail/scripts/slurm_task_info.sh	New script to display Slurm task resource bindings
images/jail/scripts/createuser.sh	Added --without-internal-ssh option and conditional key gen
images/jail/jail.dockerfile	Preinstalled Soperator scripts under /opt/soperator_utils
helm/soperator/crds/slurmcluster-crd.yaml	Set default taskPluginParam: Autobind=Cores
helm/soperator-crds/templates/slurmcluster-crd.yaml	Mirror default update for Helm chart
config/crd/bases/slurm.nebius.ai_slurmclusters.yaml	Mirror default update for CRD base
api/v1/slurmcluster_types.go	Updated Kubebuilder default annotation for TaskPluginParam

Comments suppressed due to low confidence (1)

images/jail/scripts/slurm_task_info.sh:9

• Nested double quotes inside the command substitution will break the shell syntax. Use single quotes around the grep and sed patterns (e.g., grep 'Cpus_allowed_list:' | sed 's/Cpus_allowed_list:\t//') or escape the inner quotes.

    "$(cat /proc/self/status | grep "Cpus_allowed_list:" | sed "s/Cpus_allowed_list:\t//g")" \