Is this page helpful?

Graceful System Reboot#

Important

This section applies only when the cluster and applications are already deployed and you need to reboot the system, or when the system reboots unexpectedly.

Before Rebooting#

Before you reboot the system, scale down the deployments so that all pods terminate gracefully:

kubectl scale deployment --all --replicas=0

After Rebooting#

After the system reboots, wait for the cluster to fully initialize. Then scale up the deployments:

kubectl scale deployment --all --replicas=1

This command sets all deployments to one replica.

Recovery Procedure#

If the system reboots before you complete the graceful reboot steps, wait for NVIDIA Cloud Native Stack (CNS) to fully initialize. Then delete all pods that do not have the Running status:

kubectl delete pods --field-selector status.phase!=Running