ci-ns-ttl-controller

This tool implements a Kubernetes controller that manages hard and soft times-to-live (TTLs) for namespaces in order to ensure that namespaces do not persist in order stop resource leaks from test workloads. TTLs can be configured for a namespace by annotating the namespace either at creation time or at any later point:

TTL Type	Annotation Key	Annotation Value	Description
Hard	ci.openshift.io/ttl.hard	time.Duration-formatted string	The time after namespace creation when the namespace should be reaped.
Soft	ci.openshift.io/ttl.soft	time.Duration-formatted string	The time after Pod termination or last active marker where, if no other Pods are Pending or Running, the namespace should be reaped. The last active marker is optional and stored as a RFC3339-formatted date in the ci.openshift.io/active annotation.

Whichever TTL is reached first will be used by the controller to delete the namespace.

The controller will only consider pods matching the !ci.openshift.io/ttl.ignore selector for soft TTL calculations, allowing non-matching pods to keep running while considering the namespace to be inactive.

Deployment

Deployment of these components requires system:admin level control, as it requires creating wide-reaching ClusterRoles and ClusterRoleBindings. To deploy, run:

make deploy