Usage Guidelines for Switch Export-Import Process
- Pluribus recommends you to backup the configurations on all switches to an external server using the switch-config-export upload-server command to retrieve the configurations in case of a hardware failure.
- The configuration file must use the *.tar.gz extension to be recognized by nvOS.
- Importing the configuration file causes nvOS to restart which results in a brief interruption in traffic. Hence a scheduled maintenance window is advised.
- The switch-config-import command imports the last backed-up configuration and if the configuration has changed after the last backup, you must manually make those configuration changes after restoring the previously saved configuration.
- There are two options that allow you to control how the switch-config-import command modifies the switch:
- apply-system-config — While replacing a faulty switch, apply this parameter to the switch-config-import command if you want all the switch settings (displayed in the switch-setup-show command output ) including hostname, mgmt IP address, in-band IP address, etc., from the old switch to be restored onto the new switch.
- ignore-system-config — Use this parameter if you do not want to restore the additional switch settings from the old switch on to the new switch. This is the default setting.
- The switch-config-reset, software-upgrade, and fabric-upgrade-start commands automatically back-up the configuration before executing these commands.
- To restore the configuration on a non-cluster node, use the switch-config-import process described in this section.
- To restore the configuration on a cluster member-node, Pluribus recommends to use the cluster re-peer process. Use the fabric-join repeer-to-cluster-node command if one of the member-nodes of the cluster is active and none of the nodes in fabric is offline except the node that is faulty, for which the configuration needs to be restored. This is because each cluster node automatically backs-up the other node's configuration. If one node in cluster fails, the above command restores the configuration from the other node. For details, see the Performing the Cluster Re-peer Process in the Configuring High Availability chapter.
However, if both member-nodes of a cluster fail or is being RMAed, or in the cases where multiple fabric member-nodes are offline, then use the switch-config-import command to restore the configuration on the cluster member-node.
Note: As a best practice, Pluribus recommends to periodically backup the configurations on all fabric nodes, including cluster nodes and save them to an external server.