UI Guide
A screen-by-screen tour of the Slurm Cluster console: what every page, tab, button, dialog and field is, and where to find the control you are looking for. Use it as a reference rather than reading it end to end.
Navigating There
Training Cluster in the left sidebar, under PRODUCTS. The project selector at the top-left scopes everything — a cluster only appears in the project it belongs to.
Clusters are project-scoped. If a colleague's cluster is missing from your list, check the project selector before anything else.
The Cluster List Page
Headed Manage Training Clusters.
Header
| Control | What it does |
|---|---|
| Refresh icon | Reload the list |
| Documentation | Opens these docs |
| Create Training Cluster | Starts the create form |
Table
| Column | Shows |
|---|---|
| Cluster Name | Click it to open the cluster |
| Created By | Who created it |
| Plan | The plan, with (Private Cluster) appended for reserved capacity |
| No of Nodes | Worker count |
| Created At | Creation time |
| Connect | A terminal icon that opens the SSH panel. Disabled on a Failed cluster |
| Status | A coloured pill — see statuses |
| Actions | The ⋮ menu |
Rows per page and pagination sit below. The list refreshes itself every few seconds while any cluster is Creating or Deleting.
Bottom cards
More about Training Clusters — three cards linking to product documentation, the API reference, and FAQs.
The Create Form
One page, top to bottom. Fully documented in Create a Slurm Cluster.
| Section | Fields |
|---|---|
| Cluster Name * | The name |
| Image * | Image family and version, plus a refresh icon |
| Plan Configuration | GPU / Private Cluster tabs; plan cards; Configuration; Workers stepper |
| Pricing | On-Demand and any savings plans, then the amber note "Charges apply from the moment the cluster is created." |
| Access | SSH Keys *, Parallel File System * |
| Advanced Settings | Security Group *, Lifecycle Script, Slurm Configuration, Shared File System, Dataset Storage |
| Submit | Create Training Cluster |
* marks a required field. The submit button stays disabled until every required field is valid.
That GPU type has no free nodes right now. Pick another, or reserve capacity with a Private Cluster.
The Cluster Details Page
Header
| Element | What it is |
|---|---|
| Select Training Cluster | Switch clusters without going back to the list |
| Status pill | The cluster's current state |
| Refresh icon | Reload the current tab. The details page does not auto-refresh |
| Connect | The SSH panel |
| Restart Workers | Shortcut for Restart All Workers. Disabled unless Running |
| ⋮ | The Actions menu |
Tabs
Ten tabs, left to right:
| # | Tab | Contents | Reference |
|---|---|---|---|
| 1 | Details | Cluster Details, Plan Details, Connection Details | Manage |
| 2 | Cluster Overview | Summary cards, Node Health, Slurm Partition Info | Cluster Overview |
| 3 | Nodes | Node-Level Metrics (DCGM), filters, per-GPU drawer | Nodes |
| 4 | Jobs | State counters, filters, job table | Jobs |
| 5 | Monitoring | GPU and scheduler charts | Monitoring |
| 6 | Alerts | Attach alerts | Alerts |
| 7 | Logs | Login-node and worker logs | Logs |
| 8 | Volumes | Datasets, SFS, PFS sub-tabs | Storage |
| 9 | Network & Security | Network and Security Groups sub-tabs | Network |
| 10 | Login User Management | Members and Groups | Login users |
Most tabs show a message such as "Cluster overview is available when the cluster is running." until the cluster is up. Details always works. The exact tab set varies slightly by region — some regions show a dedicated Partitions tab in place of Cluster Overview.
The Connect panel
Opened by Connect. Headed Use Shell to run freeform commands, with a copy button per row:
| Row | Command |
|---|---|
| Connect to the head instance using SSH | ssh root@<ip> |
| To view information about nodes and partitions. | sinfo |
| To view and modify configuration and state. | scontrol show nodes |
| To submit a batch script for later execution | sbatch <PATH OF SCRIPT> |
The sstat command can be used to track the resource usage of your jobs. | sstat -j <Job ID> |
The scontrol command can be used to track your jobs. | scontrol show job <Job ID> |
The Actions Menu
Available from the ⋮ on a list row and from Actions on the details page. Full reference: Cluster actions.
| Item | Opens | Restarts anything? |
|---|---|---|
| Update SSH Keys | Update SSH Keys dialog | No |
| Update Image | Update Image: <name> → Confirm Image Update | Rolling, node by node |
| Slurm Configuration | The Slurm Configuration dialog | Reloads slurm.conf; may restart the controller |
| Scale Cluster | Scale pool: <name> with Scale up / Scale down | Scale-down restarts workers |
| Convert to Committed | Convert to Committed Plan | No — billing only |
| Restart All Workers | Confirmation | All workers |
| Restart Cluster | Confirmation | Everything |
| Restart Login Service | Confirmation, confirm button Restart | Login node only |
| Reconfigure Cluster | Confirmation | Re-applies current state |
| Clone Cluster | Navigates to a pre-filled create form | No |
| Terminate Cluster | Confirmation, confirm button Terminate | Ends the cluster |
| Delete Training Cluster | Confirmation with an acknowledgement checkbox | Removes the record |
The last two are red, below a divider. On the details page Delete Training Cluster is labelled Delete Cluster.
Most actions require Running. See Action availability and blockers for the full matrix.
The Slurm Configuration Dialog
Reached from Actions → Slurm Configuration on a Running cluster, or from Configure Slurm in the create form. Full reference: Slurm Configuration.
Title: Slurm Configuration
Intro: "Applied to slurm.conf on every node of the cluster. Platform managed settings (cluster
name, accounting, node and topology wiring) cannot be overridden."
Tab 1 — slurm.conf
| Element | Detail |
|---|---|
| Heading | Extra slurm.conf |
| Helper | Custom Slurm configuration appended to slurm.conf on all nodes. One Key=Value per line. |
| Link | slurm.conf reference → SchedMD's manual |
| Field | A text area, one Key=Value per line |
Tab 2 — Partitions
| Element | Detail |
|---|---|
| Heading | Add Partitions — Adds a PartitionName= line to slurm.conf. |
| Button | + Add Partition |
| Empty state | No new partition yet. Use Add Partition to define one. |
| Existing partitions | Listed under Already configured as read-only chips; click one to jump to the partition table |
Each new partition card is headed NEW PARTITION with an Enabled toggle and a remove button:
| Field | Control |
|---|---|
| Name | Text |
| State | UP (default), DOWN, DRAIN, INACTIVE |
| Max time | Text — Slurm time format or UNLIMITED |
| Default time | Text |
| Nodes in this partition | Text, with a node-count badge |
| Extra partition options — one Key=Value per line | Text area |
| Make this the default partition | Toggle |
Tab 3 — Prolog & Epilog
Four script editors, each requiring a shebang:
| Editor | Runs |
|---|---|
| Worker Prolog | On Slurm workers before each job starts |
| Worker Epilog | On Slurm workers after each job ends |
| Controller Prolog | On slurmctld at job allocation |
| Controller Epilog | On slurmctld at job completion |
Footer
An info panel headed Cluster Reconfigure Required, then Cancel and Apply Configuration (Save Configuration in the create form).
A partition names specific nodes, and those do not exist until the cluster runs. Add partitions afterwards from the Cluster Overview tab.
Where Is That Control?
| I am looking for | It is at |
|---|---|
| The SSH command | Details → Connection Details, or the Connect button |
| The cluster's IP | Same, or Network & Security → Network |
| Which Slurm version I have | Details → Cluster Details → Image Version |
| Add a partition | Cluster Overview → Slurm Partition Info → Add Partition |
Edit slurm.conf | Actions → Slurm Configuration → slurm.conf |
| A prolog script | Actions → Slurm Configuration → Prolog & Epilog |
| Why a job is pending | Jobs, then squeue -o "%R" over SSH |
| A broken GPU | Nodes → filter XID Errors |
| Reboot one node | Nodes → View GPUs → Reboot Node |
| Add nodes | Actions → Scale Cluster → Scale up |
| Mount a volume | Volumes → the type → Mount |
| Add a teammate | Login User Management → Members → Create Member |
| Open a port | Network & Security → Security Groups |
| Attach an alert | Alerts → Select Alerts → Attach |
| Stop paying | Actions → Terminate Cluster |
| Recreate a cluster's configuration | Actions → Clone Cluster |
Icon and Badge Glossary
Status pills
| Pill | Colour | Meaning |
|---|---|---|
| Running | Green | Live and accepting jobs |
| Creating | Amber | Provisioning |
| Terminating / Deleting | Amber | Shutting down |
| Failed | Red | Creation did not complete |
| Terminated | Red | Shut down; the record remains |
Node state badges
| Badge | Meaning |
|---|---|
| Idle | Up, nothing allocated |
| Allocated | Fully allocated to jobs |
| Mixed | Partly allocated |
| DOWN | Slurm has marked it unavailable |
| UNKNOWN | Slurm is not reporting it |
Health badges
| Badge | Meaning |
|---|---|
| Healthy | No faults reported |
| XID Error | A GPU on this node reported an NVIDIA fault |
| Rebooting | A node reboot is in progress |
Other markers
| Marker | Meaning |
|---|---|
| Floating on an IP | Assigned from a pool; can change |
| Reserved on an IP | Held for your account; stable |
* beside a partition | The default partition |
| Managed in a partition's Actions | A platform partition you cannot edit |
| not live on a partition | Configured, but the controller is not reporting it |
| DISABLED as a partition state | One of your partitions with Enabled switched off |
∞ in MAX TIME | UNLIMITED |
| Latest / Current on an image version | Newest available / running now |
| Inventory not available on a plan card | No free nodes of that type |
– or — in a cell | No data |