Ability to move node between computes in a project

[eantowne]

In a multi-compute environment, being able to move a node of a project from one compute to another would be very useful.

Example:

• Project is running in a multi-compute environment consisting of 3 computes (A,B,C)
• Currently there is a node (instance of device X appliance) running on compute C
• Compute C needs to be decommissioned

Currently, this would require making a backup of the running node's internal configuration, delete the node from the project, then re-add it to the project, but running on compute A or B. Additionally, installation of the appliance may be necessary on the new compute location.

Proposed methodology:

1. If in a multi-compute environment, UI should enable elements to allow for movement of nodes between computes (Menu option, context-menu), assuming node is powered off.
2. Upon user action requesting move, compute checks if required appliance with specific version is installed on requested compute.
3. If not, user is asked if they want to install the appliance to the new compute. If so, installation occurs.
4. Check for node powered state, and power-off if running (user notification).
5. Node "copied" to purgatory (holding area so that two of the same nodes do not exist).
6. Node removed from old compute.
7. Node added to new compute.
8. "Purgatory" copy of the node maintained in case something goes wrong, allowing for an "undo" capability.

[jean-christophe-manciot]

This requested feature is definitely needed.

I have a few comments:

• a node have several parts:
  • its gns3 configuration in gns3_controller.conf
  • its disk image
  • all its files below <project_id>/project-files
• I don't see why it would be necessary to remove ts disk image from the old compute
• I think it would be also easier to keep as is all the old node files below <project_id>/project-files on the old compute; they would simply be replaced by the newer ones in case another move is necessary in the other direction
• the node_id should remain the same on both computes, which is already the case with the project_id.