We’re upgrading EventStore to 4.1.4 in our production environment and are in the process of rebuilding indexes. To avoid loss of redundancy during the upgrade, we’re rebuilding indexes on a standalone node based on a backup of the live data as suggested in the upgrade instructions.
When this process is complete, we can either take a backup of the state in the standalone node and deploy that as one of our instances joining the cluster - this would then catch up with the other 2 nodes and add the events that had been added since it was “detached”.
Alternatively, it would be easier in our case to transfer only the new index from the standalone instance to one of the cluster nodes and restart.
Is this safe? Would EventStore detect that the index is not complete and fill in the new data, would it attempt to rebuild the index again from scratch, or is there any risk of corruption by doing it this way?