runai-bgu delete Manual

Introduction

runai-bgu delete is a command-line interface (CLI) for deleting workloads on the BGU HPC cluster. This command allows you to remove jobs that are no longer needed, freeing up cluster resources and cleaning up your workspace. The command automatically detects the workload type (workspace or training) and safely removes it from the cluster along with associated resources.

This manual explains how to use runai-bgu delete to remove workloads from the cluster.

Quick Start

To delete a workload, use:

$ runai-bgu delete my-workload

Deletes the specified workload from your default project.

Basic Usage

Delete a Workload

Remove a workload by name:

$ runai-bgu delete research-job

Deletes the workload named research-job from your default project.

Verify Before Deletion

Check workload details before deleting:

$ runai-bgu describe my-workload
$ runai-bgu delete my-workload

Review workload information to ensure you’re deleting the correct job.

Understanding Delete Operation

When you delete a workload:

Resource Release: All allocated CPU, memory, and GPU resources are immediately freed and returned to the cluster.
Data Preservation: Data stored in persistent volumes remains intact, but temporary container storage is lost.
Process Termination: All running processes within the workload are terminated.
Network Cleanup: Associated network resources and port forwards are cleaned up.

Common Use Cases

Clean Up Completed Jobs

Remove finished training jobs to keep your workspace organized:

$ runai-bgu delete completed-training

Free Resources

Delete resource-intensive workloads when not needed:

$ runai-bgu delete gpu-workload

Remove Failed Jobs

Clean up jobs that failed to start or completed with errors:

$ runai-bgu delete failed-experiment

Development Cleanup

Remove temporary development workspaces:

$ runai-bgu delete test-workspace

Pre-Deletion Checklist

Before deleting a workload, consider:

Save Important Data: Ensure any important results or logs are saved to persistent storage.
Check Dependencies: Verify no other processes or team members depend on the workload.
Review Logs: Extract any needed log information before deletion.
Confirm Workload Name: Double-check the workload name to avoid accidental deletion.

Best Practices

Verify Workload Information

Always check workload details before deletion:

$ runai-bgu list
$ runai-bgu describe target-workload
$ runai-bgu delete target-workload

Save Important Output

Extract logs and results before deletion:

$ runai-bgu logs my-workload > workload-logs.txt
$ runai-bgu delete my-workload

Regular Cleanup

Maintain a clean workspace by regularly removing unnecessary workloads:

$ runai-bgu list --status Completed
$ runai-bgu list --status Failed

Team Communication

Inform team members when deleting shared resources or workloads.

Deletion Safety

Immediate Effect

Workload deletion is immediate and cannot be undone. The workload will:

Stop running immediately
Release all allocated resources
Remove associated network configurations
Clean up temporary storage

Data Persistence

While the workload is deleted:

Persistent Volumes: Data in mounted persistent volumes remains available.
Shared Storage: Files in shared storage systems are preserved.
Container Storage: Temporary container file systems are permanently lost.

Recovery Limitations

Once deleted, workloads cannot be recovered. You would need to:

Resubmit the workload with the same configuration
Restore any lost temporary data from backups
Reconfigure any custom environment settings

Troubleshooting Deletion

Workload Not Found

If the workload cannot be found:

$ runai-bgu list
$ runai-bgu list -A  # Check all projects

Permission Issues

Verify you have deletion permissions:

$ runai-bgu describe my-workload  # Check ownership and project access

Deletion Failures

If deletion fails, the workload might be in a protected state or have dependencies. Check the workload status and wait for any ongoing operations to complete.

Alternative Actions

Before deleting, consider these alternatives:

Suspend Instead: Use runai-bgu suspend to temporarily stop the workload while preserving it for later use.
Resource Reduction: Some workload types allow resource modification without deletion.
Scheduling Changes: Modify workload scheduling rather than deleting entirely.