Collects diagnostic information from your CMK Slurm cluster and packages it into a zip archive for troubleshooting.
curl -sL https://raw.githubusercontent.com/crusoecloud/cmk-support-bundler/main/collect.sh -o collect.sh
bash collect.sh <namespace>Replace <namespace> with the namespace where your SlurmCluster is deployed. If you omit it, the script will prompt you.
The script will show which cluster and namespaces it will access, then ask for confirmation before proceeding.
| Category | Details |
|---|---|
| Slurm CRs | SlurmCluster, Slinky Controller/LoginSets/NodeSets |
| Slurm status | sinfo, scontrol (config, nodes, partitions), sdiag |
| Configuration | Slurm-related ConfigMaps and mounted /etc/slurm/* files |
| Logs | Container logs for controller, login, and worker pods; operator and topograph logs |
| GPU | nvidia-smi hardware queries, GPU topology, NVLink status, Xid errors |
| Network | InfiniBand status (ibstat, ibstatus, ibv_devinfo), NCCL environment |
| System | CPU, memory, disk, mounts, network interfaces, dmesg, sysctl |
| Kubernetes | Services, endpoints, nodes, PVCs, events, pod listings |
| Helm | Status, history, values, and manifests for operator Helm releases |
- Secret contents or names
- Customer workloads or job data (squeue, job queues, process lists)
- Environment variables from pods
- Deployments, StatefulSets, DaemonSets
- GPU process listings (nvidia-smi process table is stripped)