Server lifecycle events

Last modified: 21 August 2024

This topic describes Datalore server lifecycle events, such as restart or shutdown.

note

None of the currently running computations and notebook agents will be affected by Datalore restart. Once the connection is lost, the agent will attempt to re-establish the connection back, unless either it succeeds or the agent is terminated manually.

As Datalore's pod is controlled by a StatefulSet, there are two ways to restart Datalore:

  • Do a rollout restart: 

    $
    kubectl rollout restart sts datalore

  • Delete the pod manually: 

    $
    kubectl delete pod/datalore-0
kubectl wait pod/datalore-0 --for=delete --timeout=600s

tip

When using the second approach, the pod will be automatically redeployed by Kubernetes, as the StatefulSet replicas condition will not be met once the original pod is deallocated.