In the process of upgrading the OS of all the CAPM systems. In order to upgrade the OS of the DR cluster, the plan is to build a new node with RHEL 8, to replace an old node with RHEL 6.
The DR cluster is a 3 node cluster. Our plan is to replace node 3, with the new node_3a. Once that has been replaced and synced, then replace node 2 with new node 2a and lastly, replace node 1 with node node 1a.
The above process will result in no outage as the DR cluster can run on 2 nodes?
How do we go about replacing a node in the cluster with a new node? What is the step by step process?
DX NetOps CAPM Release : 21.2 or later
There's no real support for a node by node, one at a time upgrade. One way is to install a new RHEL8 node, add it to the cluster, rebalance, then remove the original RHE6 node, and then rebalance. You'll somehow have to do a re-ip from new IP to old IP for each node as you go. Repeat for the 2 other nodes.
However, this would be painstakingly slow and will affect DA performance the whole time. Plus all the nodes get renamed from 1,2,3 to 4,5,6 which makes copy-cluster difficult. So we do not recommend this approach.
Instead, we recommend that you follow the 3rd upgrade option suggested in the Vertica KB :
Prior to starting, have backups created as per our TechDocs:
Then update the OS and reinstall the DR then upload the backup.
Make a backup of the current 3 DR nodes, bring down the DA, then upgrade the OSs on all 3 nodes (if VM's, take a snapshot after shutting down first). Then once all 3 are upgraded to RHEL8, then install any OS packages specified by our TechDocs:
Then restore the DB from backup so we make sure nothing got corrupted. Start the DA. If everything runs fine, remove the VM snapshots as snapshots slow down the machine.