Suspend & Resume Clusters
Clusters can be manually suspended (stopped) or resumed (started) whenever you need to, independently of the automatic threshold described in Auto Suspension. Suspending releases the underlying instances so your cloud provider stops billing for them, while the cluster's configuration survives untouched: its name, its compute shape, its attached resources and its access grants are all still there when you come back.
Resuming is therefore cheap. We restart instances against a definition that never went away, which is why a warm start completes in roughly 10 seconds where cold provisioning takes 4 to 5 minutes.
Destroying is the irreversible one. It releases the definition as well as the instances, so a destroyed cluster can't be resumed and the status moves through DESTROYING to DESTROYED rather than back to STOPPED.
Suspend (Stop)
- Stop the cluster after clicking on stop button in action menu of cluster
- Helps reduce cost
Resume (Start)
- Starts the cluster after clicking on start button in action menu of cluster
- The Warm Start mechanism brings a stopped cluster back in about 10 seconds
- Useful for running job or notebook
Destroy
- Destroy the cluster by clicking on destroy button from action menu of cluster
Steps
- Go to the Cluster Dashboard
- Click Start or Stop next to the cluster
- Yeedu handles the underlying infrastructure automatically