The scenario is, three physically separate peers (Gluster servers). 1 volume replicated across all three servers, with 1 brick on each server.
If 2 of the 3 peers were to loose network connectivity indefinitely and the volume needed to be used ASAP, what would be the correct way to get that volume back up and running on just the 1 remaining peer?
The solution I have come up with is to remove the bricks on the missing peers and reducing the replica count down from 3 to 1. Then, if any of the mounts on the clients were pointing to one of the peers that went down, unmount these and remount to the remaining live peer.
If/when the 2 missing peers come back up, add the bricks back into the volume and up the replica count back up to 3.
Is this the correct thing to do? Does anyone see any issue with this or know of a better or 'proper' way?