Skip to content

Home Dashboard

The Home Dashboard is your starting point — it gives you a real-time snapshot of GPU availability, your resource usage, active jobs, and recent platform activity.

Access: All authenticated users.

Dashboard Layout

graph TD
    subgraph TopRow["Top Row"]
        A["GPU Availability Card\n(cluster-wide GPU units)"]
        B["Project Summary\n(active projects count)"]
    end
    subgraph MiddleRow["Middle Row"]
        C["Resource Usage Chart\n(quota used vs. allocated)"]
        D["Activity Feed\n(recent platform events)"]
    end
    subgraph BottomRow["Bottom Row"]
        E["My GPU Jobs\n(running workspaces & jobs)"]
    end
    TopRow --> MiddleRow --> BottomRow

Widget Reference

Widget What It Shows Refresh
GPU Availability Card Cluster-wide free vs. allocated GPU units, plus preemptible-queue indicator On page load and every 10 seconds
Per-Node GPU Allocation Grid of every GPU node with free SM percentage and busy/free counts On page load and every 10 seconds
Project Summary Total projects you belong to On page load and every 10 seconds
Resource Usage Chart GPU/CPU/memory quota consumed vs. total allocated per project, plus plan-window status On page load and every 10 seconds; countdown ticks every second
Activity Feed Chronological log of recent platform events (workspace starts, project changes, storage operations) On page load and every 10 seconds
My GPU Jobs Table of your currently running workspaces and batch jobs with GPU, SM share, and memory On page load and every 15 seconds

Reading GPU Availability

The GPU Availability Card shows the current state of the cluster's GPU pool.

Home dashboard GPU availability card Figure 1: Home dashboard showing GPU availability, resource usage, and active jobs. Replace this placeholder with an actual screenshot.

Steps to read the GPU card:

  1. Look at the allocated / total fraction (e.g., 6 / 8 GPUs).
  2. Free GPUs = total − allocated. If free GPUs is 0, your workspace launch will be queued.
  3. The card also shows a Preemptible queues count — when GPUs are exhausted but preemptible queues exist, higher-priority workloads can still be scheduled by evicting lower-priority pods.
  4. Click the card to navigate to the Admin Monitoring page (admin users only) for per-node breakdown.

DRA effective GPU units

The card reports effective GPU units from DRA labels and live ResourceClaim bindings. A 1-GPU claim at 50% SM counts as 0.5 GPU only while DRA has bound it to a live Pod.

Per-Node GPU Allocation

Beneath the availability card, the Per-Node GPU Allocation grid shows every GPU node as a small tile with:

  • Node name and GPU model
  • A free SM percentage bar ({{pct}}% free) per device
  • An Available / Busy badge so you can pick a less-contested model
  • Total GPU count on the node

Use this grid to spot which models are idle before submitting a workload, or to confirm a specific node has capacity for a fractional claim.

Resource Usage Chart

The chart displays your usage across all projects you belong to.

  • Each bar represents one project.
  • The bar fill shows consumed quota (GPU-hours, CPU-hours, or memory).
  • Each project row shows whether the bound resource plan is always active, closing, opening, or expired.
  • Plan-window countdowns update once per second without requiring a page refresh.

Activity Feed

The activity feed shows the 20 most recent events relevant to you:

  • Workspace started / stopped / deleted
  • Project member added
  • Storage permission changed
  • Request status updated

Click any activity item to navigate directly to the related resource.

My GPU Jobs

The My GPU Jobs table lists your active workspaces and batch jobs:

Column Description
Name Job / workspace name
Project Owning project
GPU Effective GPU units consumed
Status Running, Pending, Stopped, Error
Started Start timestamp
Actions Stop / Open links

Click Open next to a running workspace to open the IDE in a new tab.

Navigating from a job to its project

Click the project name in the Jobs table to jump to the Project Detail page where you can see quotas and config files.

Common Questions

Why is my GPU count 0 on the availability card?

The card shows effective DRA GPU availability. If all compatible GPUs or SM shares are allocated, you'll see 0 available. Check the Activity Feed — a workspace that finished recently may have freed capacity within a minute.

The activity feed is empty — is something wrong?

No. A fresh account with no project activity will show an empty feed. Create a project or launch a workspace and events will appear.

My running workspace is not in the jobs table.

The table refreshes every 15 seconds. Refresh the page (F5 or the browser refresh button) if the workspace still does not appear after the next poll.