Action Availability and Blockers
An action can be missing from the menu, greyed out inside a dialog, or rejected after you confirm. This page is the single reference for why — start here when a control you expected is not there.
Quick Diagnosis
Work down this list — the first match is almost always the answer.
| Symptom | Check |
|---|---|
| The action is not in the menu at all | The cluster status. Most actions need Running — see what each status allows |
| Scale Cluster is missing on a Running cluster | The cluster is on a committed plan. Scaling is not supported for committed plans |
| Convert to Committed is missing | The cluster is already committed, is a CPU cluster, or runs on a private cluster plan |
| Slurm Configuration is missing | The cluster is not Running |
| Update Image is missing | The cluster is not Running |
| No actions at all, or the page is read-only | You do not have update permission on Training Clusters in this project |
| The dialog opens but the confirm button is grey | Something in the dialog is incomplete — see disabled-confirm reasons |
| You confirmed and got an error | See blockers and their messages |
| A tab shows "available when the cluster is running" | Cluster Overview, Partitions, Nodes, Jobs, Monitoring, Volumes and Network & Security need a Running cluster |
| The cluster is gone from the list | It was deleted. From the moment Delete is issued the cluster is removed from the list and every action on it fails |
Cluster Statuses
| Status | Meaning |
|---|---|
| Creating | Provisioning is in progress |
| Running | The cluster is live and accepting jobs |
| Failed | Creation did not complete and was rolled back |
| Terminating | Termination in progress |
| Terminated | Terminated. The record remains so you can clone or delete it |
| Deleting | Deletion in progress |
| Deleted | Deleted. No longer listed |
The list page refreshes itself every 5 seconds while any cluster is Creating or Deleting, so you can watch provisioning without reloading. The details page does not auto-refresh — use the refresh icon in the header.
Creating usually finishes in a few minutes
If a cluster sits in Creating much longer, try Reconfigure Cluster — it is available on a Creating cluster and re-applies the intended state. If that does not help, note the cluster name and contact support.
What Each Status Allows
| Status | Actions offered |
|---|---|
| Creating | Update SSH Keys · Convert to Committed¹ · Reconfigure Cluster · Clone Cluster · Terminate Cluster · Delete |
| Running | Everything |
| Failed | Update SSH Keys · Clone Cluster · Terminate Cluster · Delete |
| Terminating | Update SSH Keys · Clone Cluster · Terminate Cluster · Delete |
| Terminated | Update SSH Keys · Clone Cluster · Terminate Cluster · Delete |
| Deleting / Deleted | Not listed. Once Delete is issued the cluster leaves the list and no action on it is possible |
¹ Only on hourly, non-private-cluster plans.
Failed clusters are for cloning, not fixing
A Failed cluster cannot be repaired into a Running one. The productive path is Clone Cluster — you get the same configuration in a fresh create form — then Delete the failed record. A Failed cluster is not billed for compute.
Visibility Rules by Action
| Action | Visible when | Hidden when |
|---|---|---|
| Update SSH Keys | Always | Never |
| Update Image | Status is Running | Any other status |
| Slurm Configuration | Status is Running | Any other status |
| Scale Cluster | Running and (hourly plan or private cluster) | Committed plan, or not Running |
| Convert to Committed | Hourly plan and not a private cluster and status is not Failed / Terminating / Terminated | Committed plan, CPU cluster, private cluster, or one of those statuses |
| Restart All Workers | Running | Any other status (the header button is disabled rather than hidden) |
| Restart Cluster | Running | Any other status |
| Restart Login Service | Running | Any other status |
| Reconfigure Cluster | Running or Creating | Any other status |
| Clone Cluster | Always | Never |
| Terminate Cluster | Always | Never |
| Delete | Always | Never |
| Add / Edit / Delete Partition | Running; row actions only on your partitions | Platform partitions show Managed and have no actions |
| Node reboot | Nodes tab, and the node is not already rebooting | While a reboot is in flight |
| Mount / unmount a volume | Volumes tab, cluster Running | — |
| IP and security-group actions | Network & Security tab, cluster Running | The VPC section is not shown in every region |
| Connect | Always in the list; the details button needs Running | The list icon is disabled on a Failed cluster |
Why a Confirm Button Is Disabled
| Dialog | Confirm stays grey while |
|---|---|
| Update SSH Keys | Nothing is selected, or the selection is unchanged from the current keys |
| Update Image | No version chosen, or the chosen version is already the current one |
| Slurm Configuration | The configuration is still loading or being saved. Validation errors are listed above the buttons and block the save |
| Scale up | The total is not above the current node count, no inventory is available, or you asked for more than is free |
| Scale down | No node selected, or every node selected — at least one must remain |
| Convert to Committed | No committed plan is available, or the acknowledgement is unticked |
| Delete Partition | The acknowledgement checkbox is unticked |
| Delete Training Cluster | The acknowledgement checkbox is unticked |
| Any action | A previous submission is still in flight |
Blockers and Their Messages
These are checked after you confirm.
Plan and cluster-type blockers
| Message | Cause | Fix |
|---|---|---|
| Scale cluster is not supported for committed cluster | Scaling a committed plan | Size the cluster before committing, or create a second hourly cluster |
| Cluster plan upgrade not allowed for committed clusters | Asking for plan options on a committed cluster | None — the commitment window has to run out |
| Convert to committed is not supported for committed cluster | Already committed | — |
| Convert to committed is not supported for CPU cluster | CPU clusters have no committed plans | — |
Capacity and quota blockers
| Message | Cause | Fix |
|---|---|---|
| Inventory not available for the selected configuration | No free nodes of the right type on the same rack as your existing nodes | Retry later; inventory is refreshed frequently. For guaranteed capacity, use a Private Cluster |
| Inventory Unavailable in the Scale dialog | Same, surfaced before you submit | As above, or scale to the maximum the dialog offers |
| A credit-related message | Not enough credit for the requested nodes or the upfront committed charge | Top up, then retry |
| A BOM or limit message | An account-level limit on the requested configuration | Contact support |
State blockers
| Message | Cause | Fix |
|---|---|---|
| Slurm configuration can only be updated on a running cluster | Editing Slurm config on a non-Running cluster | Wait for Running |
| Minimum one node required for cluster to remain in running | Scaling down to zero | Keep at least one node, or terminate instead |
| Invalid request, Cluster does not exist | The cluster was deleted, or belongs to a different project | Check the project selector |
| A VPC-IP-in-progress error | A VPC IP is mid-attach, which blocks terminate and delete | Wait for the attach to settle, then retry |
Storage blockers
| Message | Cause | Fix |
|---|---|---|
| Cannot remove storage. At least one storage volume must remain mounted on the cluster. | Unmounting the last volume | Mount a replacement first, then unmount |
| Cannot unmount this storage as it is used for persistent login. Disable persistent login on the cluster first. | The volume holds the home directories | Disable Login User Management first |
Network blockers
| Message | Cause | Fix |
|---|---|---|
| At least one IP (reserved or floating) must be attached. | Detaching the last IP | Attach a reserved IP first, then detach the other |
| A "at least one security group" error | Removing the last security group | Attach a replacement first |
Partition validation
See Manage partitions for the full list, including reserved names, the 20-partition limit, and the node-membership rules.
No Action Checks Your Job Queue
This is the most important operational caveat on the page.
Nothing stops you restarting a cluster with jobs running
No action inspects the Slurm queue before it runs. Restart All Workers, Restart Cluster, Scale down and Terminate all proceed with jobs in flight, and those jobs are lost. Only Update Image drains nodes as part of its rolling restart.
Before any disruptive action:
- Open the Jobs tab or run
squeue. - Let jobs finish, or drain first:
scontrol update nodename=<node> state=drain. - Confirm your jobs checkpoint to a PFS or SFS mount so a loss costs minutes, not days.
What each action interrupts
| Action | Interrupts jobs? |
|---|---|
| Update SSH Keys | No |
| Convert to Committed | No — billing only, nothing on the cluster changes |
| Slurm Configuration | No — running jobs continue; queued jobs use the new settings |
| Reconfigure Cluster | No |
| Scale up | No |
| Mount / unmount a volume | Yes — the nodes restart |
| Update Image | Yes, gracefully — nodes are drained one at a time |
| Restart Login Service | Only shells on the login node; Slurm jobs keep running |
| Restart All Workers | Yes — all running jobs |
| Restart Cluster | Yes — all running jobs |
| Scale down | Yes — on removed nodes, and remaining workers restart too |
| Terminate / Delete | Yes — everything |
Permissions
Actions are gated by your role on Training Clusters in the project.
| You need | For |
|---|---|
| Read | Viewing clusters, all tabs, and Connect |
| Update | Every action in the Actions menu except Clone and Delete; partitions; volumes; IPs; security groups; node reboot |
| Create | Clone Cluster — it leads to a create form |
| Delete | Delete Training Cluster |
If the whole Actions menu is missing, or the details page is read-only, you have read access only. Ask a project administrator to review your role in the IAM Panel.
Related Resources
Last updated on September 10, 2026.