So, if we want to remove auto-failover from an existing master cluster would it be best to:
- backup the master node
- shut down the master cluster
- remove the auto-failover policy section from the root policy
- build a new (or upgraded) cluster, restoring the master node from backup
- re-load the policies
- complete rest of the build out/upgrade
Out of curiosity, what would be the impact of removing the auto-failover section from the root policy and reloading on the master cluster?
( Have not yet found the remove auto-failover covered explicitly in the docs. )
I would recommend using one of the existing standby nodes and rebuild it as a standby that is not added to the auto-failover cluster. Once it is added as standby, ensure that it has replicated all of the database data (by checking the replay bytes / replay location). Once that is done, shut down the master and other standby, and promote the new standby to the master role using
evoke role promote.
Once this is the master, you can use it as the primary master and add the other nodes that were in the cluster as standbys. If you want to, you can repeat the process on the original master node and promote it to be the master and then reseed the original standby nodes as standby nodes.
When you have the master on the node you want it on, make sure you go onto the standbys in the same site and run
evoke replication sync enable. Finally, run
evoke replication sync start on the master to elect a sync standby within the standard cluster.
Regarding the policy, you can leave the original cluster policy in place if you wish or you can remove it. It is only when the nodes are enrolled in an auto-failover cluster using
evoke cluster enroll that they become auto-failover and cluster configuration aware.
Please let me know if you have any other questions!