Why Ocean doesn’t launch a node automatically

During shutdown hours, Ocean scales the entire cluster down to zero. All nodes are terminated, including the node that hosts the Ocean controller. Because the controller isn’t running during this off period, it can’t send any updates or instructions to the autoscaler.

When the shutdown window ends, Ocean tries to launch a single node from a virtual node group that doesn’t have taints. If all virtual node groups use taints, Ocean falls back to the default virtual node group.

However, if the default group is marked with useAsTemplateOnly, no node is started automatically.

How to resume normal scaling

To resume scaling activity, you need to make sure the controller pod can run:

1. Launch a node manually after shutdown hours.

2. Make sure the node can register to the Kubernetes cluster.

3. Once the cluster has an active node, Kubernetes can schedule the Ocean controller pod.

4. After the controller is running, Ocean resumes normal scaling behavior.

Summary

• Ocean scales the cluster to zero during shutdown hours.

• No nodes means no running controller.

• No controller means no communication with the autoscaler.

• Ocean launches a node only if a VNG without taints is available.

• If not, you must launch a node manually and allow the controller to start.