| Cluster stays in Creating status |
Quota, node pool, or provisioning issue |
Recheck requested resources and contact administrator. |
| Connection to Ray cluster fails |
Cluster not yet ready or wrong client endpoint |
Verify cluster status and endpoint details. |
| Distributed training fails on checkpointing |
No shared persistent storage |
Use the configured external storage path for checkpoints. |
| GPU workload does not use accelerators |
CPU node pool or GPU count set to 0 |
Create cluster on a GPU pool with GPU resources requested. |