Skip to main content

Manage edge instances

Add, edit, remove, and inspect Kubernetes clusters this organization manages.

Add an edge instance​

Sidebar → Edge Instances → Add Edge (top right).

Two paths:

  • Upload kubeconfig — for a cluster that already exists
  • Provision new cluster — only when the install has the Provisioner integration configured (admin sets this up in system settings)

Fill in name, optional description + categories, paste/upload the kubeconfig, save. DT Edge Platform encrypts the kubeconfig at rest with the master key; you'll never see it again from the UI.

Edit an edge instance​

Click the row → Edit. You can change:

  • Name + description
  • Categories (multi-select from the global catalog)
  • Telemetry ID — UUID label that ties this edge's metrics + logs to DT Edge Platform. Editing this orphans existing alarms and Grafana queries pinned to the old ID — only do it during a controlled migration

You can't edit the kubeconfig from the edit drawer — to swap kubeconfigs, remove + re-add the edge.

Re-upload kubeconfig​

If the cluster's apiserver endpoint changed or the embedded service-account token expired:

  1. Go to the edge's row → Replace kubeconfig action
  2. Paste the new YAML
  3. Save

DT Edge Platform re-runs the probe on the new config. The edge's identity (telemetry_id, history, alarms) is preserved.

Remove an edge instance​

Click the row → Remove. Confirm in the dialog.

What happens:

  • The DB row is soft-deleted
  • The edge stops appearing in install dropdowns / lists
  • Helm releases on the actual cluster are not touched — DT Edge Platform just stops managing them. Use helm uninstall directly on the cluster if you want to clean those up.
  • Alarm rules tied to this edge become orphaned (their telemetry_id matchers won't match anything live) — review alarms after removing edges.

Check health​

Each edge row shows a health badge:

  • 🟢 healthy — last probe succeeded
  • 🟡 degraded — probe succeeded but something looks off (e.g. a node is NotReady)
  • 🔴 unreachable — probe couldn't reach the apiserver
  • ⚪ unknown — no probe has run yet (fresh edge)

Click the edge row → Health tab → recent probe history with response times, kube version, node counts, and the last error if any.

The probe runs every health_check_interval_minutes (default 30). To force a refresh, click Recheck in the row actions menu.

See nodes​

Click the edge row → Nodes tab. You'll see every node in the cluster with:

  • IP / role (master / worker)
  • Status (Ready / NotReady / Unknown)
  • Allocatable CPU / memory / GPU count
  • Kubernetes version

This is read-only — node operations (add / remove / drain) live in the upstream Provisioner UI when configured, not in DT Edge Platform itself.

Public vs org-scoped edges​

Edges are org-scoped by default — they belong to one organization and only members of that org can see them.

Public edges (created by super-admins via Admin → Edge Instances) are shared across orgs. Any org user can install on them, with ownership tracked per release. The Public Edges admin page surfaces only when the active license carries MaxPublicEdges > 0.

Common errors​

  • "EDGE_INSTANCE_NOT_FOUND" — the ID doesn't exist or isn't yours. Did the edge get removed?
  • "FAILED_ENCRYPT_KUBECONFIG" — server-side encryption problem; ask your admin to check the ENCRYPTION_KEY env
  • "FLEET_QUOTA_EXCEEDED" — adding this edge would push the license over its node/cpu/memory/gpu cap. See Handle fleet quota exceeded.

See also​