Names and statuses

Names

• 3 to 50 characters: lowercase letters, digits and hyphens, starting with a letter or digit. End the name with a letter or digit too; see Failed.
• Unique in the boundary.
• Cannot be changed after the group is created.

Namespace names

Each resource group becomes the namespace rg-<prefix>-<name> on its clusters, where <prefix> is the first nine characters of the boundary's slug (or the whole slug when it is shorter). With the 50-character name limit this is at most 63 characters. stsh rg get <name> -o json shows it as kubernetesNamespace.

The namespace name must be unique on the platform. Two boundaries whose slugs start with the same nine characters therefore cannot both have a group with the same name.

Statuses

Status	Meaning
Provisioning	The group's namespace is being created on the boundary's clusters.
Active	The namespace exists on every cluster the group was placed on. Resources can be created in it.
Failed	The namespace could not be created on at least one cluster. The platform keeps retrying — see Failed.
Deleting	The group's namespaces are being deleted. Once every cluster has accepted the deletion, the group's access is removed and the group disappears from the list. If the access cannot be removed, the group stays Deleting and is retried — see Follow the deletion.

The platform works on groups that are being created or deleted right away, and checks every group again every minute.

When a group is Failed

A failed group is retried every minute and turns Active by itself as soon as its namespace can be created on every cluster. The reason is recorded on the group — stsh rg get <name> -o json shows it as statusReason — and on the failed operation on the boundary's Operations tab. Common causes:

• The name ends in a hyphen. Kubernetes refuses such a namespace name, so the group can never become Active. Delete it and create it with a name that ends in a letter or digit.
• A namespace of the same name already exists on the cluster. Right after a group of the same name was deleted, the old namespace may still be on its way out; the new group recovers when it is gone — see Using the name again. If the namespace is not on its way out, it does not belong to any resource group, and a platform administrator has to remove it from the cluster.
• The cluster could not be reached. The group recovers when the cluster is back.

A group that stays Provisioning

A group that was refused because the boundary had no active cluster is still listed, as Provisioning, and nothing will ever happen to it. Delete it, make sure the boundary has an active projection, and create it again.