Skip to main content

Ops & Clusters

Ops is where apps actually get onto edge clusters, and where you watch whether they landed. When you deploy something from the App Store, Ops is the machinery underneath, and this view is how you see what it did.

It is available to platform staff and tenant administrators. Engineers get a read-only view; Operators do not see it.

Two deployment modes

How a deployment reaches a cluster depends on how that tenant is set up.

ModeUsed forHow it works
InternalTenants running on platform-managed infrastructureThe app is installed directly onto the local cluster
ExternalTenants running their own clusterThe deployment is packaged as a bundle and picked up by an agent running on the customer's cluster

You do not choose the mode per deployment. It follows the tenant's configuration, and the console routes automatically. The practical difference is visibility: external deployments go through a sync step, so they surface a Pending state that internal deployments skip.

Reading the summary

The Ops summary is a row of counters covering everything you have visibility into:

CounterWhat it means
Clusters ReadyDownstream clusters healthy and checked in
Clusters PendingRegistered, but no agent check-in yet
Clusters ErrorRegistered and failing to connect
Bundles DeployedDeployment units successfully applied to their targets
Bundles PendingUnits waiting to be applied
Repos ActiveConfiguration sources with all targets in sync
Repos ErrorConfiguration sources failing to sync

A healthy fleet sits at zero in every Error and Pending column. Pending that never clears is the signal worth chasing, because it usually means an agent cannot reach the platform rather than that a deployment is wrong.

Cluster status

The cluster list shows every downstream cluster, with:

FieldDescription
NameCluster identifier, DNS-compatible
Display NameThe human-readable label
StatusReady, Active, NotReady, or WaitCheckIn
Node CountNodes in the cluster
Bundle CountDeployment units targeted at it

WaitCheckIn means the cluster is registered but its agent has not reported yet. On a new site this is normal for the first few minutes. On an established site it means connectivity, so check ArcNet tunnel status before you look at anything else.

NotReady is different: the agent is talking, but the cluster is not in a state that can accept work. That one is a cluster problem, not a network problem.

Deployment status

Each deployment unit reports its own state, and a unit targeting several clusters reports per-cluster. A unit is only as healthy as its unhealthiest target, so a single unreachable site will hold an otherwise fine deployment in a non-ready state. Drill into the unit to see which target is responsible rather than redeploying the whole thing.

Next steps