These steps are tested on Bright 9.0 with CentOS 7
PXE Boot the new head node from the current head node in RESCUE mode
- After the node boots successfully, we will run “cm-clone-install” script to start cloning the head node
Case 1: the new head node will have the same hostname as the former head node
- In this case, we will set the cm-clone-install parameter “–hostname” to the same name as the current head node:
- After the clone is done and the node is rebooted successfully, the ifcfg-* files (for the interfaces) should be updated with the correct MAC addresses, device names (if necessary) and IP addresses (if necessary):
- CMDaemon will initially fail to start on the new head node because the MAC addresses of the interfaces will not match the MAC address in the license, so “request-license” will need to be run (after unlocking the product key from the customer portal) to activate the license with the new MAC address.
- If the new head node is being accessed via BMC, please make sure that SetupBMC in cmd.conf is set to false so that when CMDaemon starts it will not overwrite the current IP address with the IP address stored in the database.
- If the IP addresses of the interfaces will not change, please make sure to power off the former head node to avoid having conflicting IP addresses.
- After updating the ifcfg-* files and re-activating the license, the MAC addresses of the interfaces on the active head node needs to be changed:
Reboot the cloned head node so that the interfaces will come up cleanly:
- After the head node has come up properly, all the compute nodes should be rebooted to get a new certificate signed by the new license on the new head node, in case the admin decided not to re-use the private key from the existing license after running request-license.
Case 2: the new head node will have a different hostname than the current head node
- In this case, in addition to the same steps in Case 1 and after updating the ifcfg-* files, the hostname in the cmdaemon database has to be updated: