This article introduce the best practice to remove a physical node from grid.
There are several possible reasons to remove a physical node from a grid including hardware upgrades and server failures. When a server hardware upgrade is needed the original server will no longer be part of the grid and so it should be removed. If only a portion of the existing server will be upgraded it still may be necessary to remove the node completely and re-add the server after the upgrade. Server hardware failures can also require that a server be physically removed from the grid while it is repaired. When a grid node is expected to be down for repair for an extended period of time it should be removed from the grid configuration and re-added after repaired. If the file system on this node are already compromised or the server has cashed then not all steps below apply.
The steps to remove
- Disable the physical node either from BFC GUI or by 3t shell. " 3t srv disable <server name>"
- Restart the applications were running disabled node. Applogic would schedule to application to start up on physical nodes(enabled nodes) which has sufficient resource."3t srv list --map" can display which appliance is running on which node.
- Manually migrate the volume stream using "3t vol migrate" command. "3t vol migrate --all" is used to migrate all volume streams from disabled node to other nodes in the grid. Applogic detects which nodes in grid have sufficient disk space and process the volume stream relocation automatically.
If you would like to know whether remaining nodes in grid has enough disk space, run "3t vol list server=<disabled server>" to display all volume streams accommodated on disabled server, afterward, run "3t srv list --verbose" to display all nodes' server resources including free disk space. In the 3t shell session, "help vol migrate" can display detailed help of this command.
- Run "3t srv list --map" to make sure not no application is running on disabled node. In addition, run "3t vol list server=<disabled server>" to make sure no current data streams exist on the disabled node.
- Reduce minimum/target server number by 1 on BFC GUI if necessary. Additionally, ensure number of remaining nodes is equal or greater than minimum server number specified on BFC to avoid error messages in the grid.
- Delete or quarantine the physical node from grid in the BFC GUI.