Errors and troubleshooting¶

What the console shows when something fails, and what to do.

Who can do this: anyone using the console. Changing a cluster still requires super_admin.

Status codes and error bodies for scripts are in the API reference.

Create is refused on the form¶

The create form can refuse the request before anything is started:

The name is already taken — pick another name. Names cannot be changed later.
The Kubernetes version or machine type is not in the catalog — the wizard only offers catalog values.
Prepaid credit is insufficient — contact your Door account team.
Your role is member — ask a super_admin.

Cluster stays in Provisioning¶

Stay on the provisioning screen. There is no published maximum duration. If the steps do not move for a long time, contact Door support with the cluster id. Do not start another cluster with the same name while this one is still being created.

Messages on a failed cluster¶

When setup does not finish, the provisioning screen and Overview show a message written for you, never raw platform text.

Message	What to do
`We couldn't set up networking for this cluster in this zone. Our team has been notified — please retry or contact support.`	Retry create, or delete and recreate. If it repeats, contact support with the cluster id.
`Setup timed out waiting for the cluster to become healthy. You can retry, or delete and recreate.`	Delete and recreate, or contact support.
`Network routing didn't finish setting up. Support can help recover this cluster.`	Contact Door support with the cluster id. Do not keep retrying the same broken row without them.
`Something went wrong during setup (ref {cluster_id}). Contact support if this persists.`	Give Door support that cluster id. This is the generic message for any other failure.

kubectl times out on a private cluster¶

The Kubernetes API is Private. Use the VPN endpoint on the Access tab. The private API is on port 6443. If you expected the public hostname under clusters.dks.door.africa, switch to Public on Access when the cluster is Ready, then download kubeconfig again.

Fewer workers than you asked for¶

The Nodes tab shows the workers that have joined. Refresh until the count matches what you asked for. The Status column shows Failed for a worker that did not come up. If the count stays lower after the pool change has finished, contact support with the cluster id.

Access button shows Managed by your organization policy¶

Use the door CLI, or ask your account team.

Adding an address is refused¶

If the zone has no addresses left, contact your Door account team for capacity. Internet addresses on a cluster created with Private (VPN) are not possible.

Delete takes a long time¶

Wait. The cluster leaves the Clusters list while Door removes its resources. The dialog offers a forced delete when Door hub environments are still attached.

Contact Door support¶

Use your Door account team / Door support channel (no public ticket URL is published here). Always include:

Cluster id (UUID), not only the name
Organization slug (acme)
Zone (abidjan)
Approximate time and the failure message (never tokens)

The generic setup failure sentence already embeds ref {cluster_id} so support can find the cluster.

Related