This topic describes how to back up and restore the VMware Tanzu Kubernetes Grid Integrated Edition (TKGI) infrastructure.
Overview
Backing up and restoring the TKGI infrastructure includes the following components:
- NSX-T Data Center. See Backing Up and Restoring NSX-T Manager.
- vCenter Server. See Backing Up and Restoring the vCenter Server.
Note: In terms of TKGI, the primary focus is the back up and restore of NSX-T Data Center. vCenter back up and restore is outside the scope of this documentation, but a mention of it is included to ensure it is part of your end-to-end back up and restore workflow.
Test Considerations
As part of your TKGI back up and restore planning and testing, consider the following test scenarios for NSX-T.
- Back up NSX-T.
- Shut down (or delete) NSX Manager nodes.
- Deploy new NSX-T Manager node and restore configuration from backup.
-
Restore additional NSX Manager nodes.
-
If you deploy a Kubernetes application with a service endpoint, and it is included in the NSX-T backup, on restore of NSX-T there is no loss in application connectivity. However, the application must be redeployed if it was deployed after the backup was taken.
-
If you create a Kubernetes namespace, and it is included in the NSX-T backup, on restore of NSX-T, the namespace is present. However, the namespace must be recreated if it was created after the backup was taken. If you delete a namespace when NSX-T is down, on restore of NSX-T the namespace is removed.
-
If you create a Kubernetes cluster, and it is included in the NSX-T backup, on restore of NSX-T the cluster is reachable. However, if the cluster was created after the NSX-T backup was taken, the cluster is not reachable. In this case, delete cluster and create new one.
-
If you create a Kubernetes namespace, and backing up NSX-T, then deploy pods, on restore of NSX-T, restarting pods brings them to running state since the namespace was created before backing up. However, if the namespace was not created before backing up NSX-T, the pods are stuck in ContainerCreating state. Both the namespace and the pods must be recreated.
-
If you create a load balancer service before backing up NSX-T, on restore the service is present. However, if you create the service after backing up NSX-T, on restore of NSX-T the load balancer VIP is pending. You must redeploy the service for it to work. If you delete a load balancer service in-between taking a back up of NSX-T and restoring NSX-T, on restore the load balancer VIP is deleted.
