Hey - we are trying to rebalance some overprovisio...
# harvester
b
Hey - we are trying to rebalance some overprovisioned disks by increasing the replica count for a volume to force a new replica on a empty disk, decreasing the replica count, and deleting the replica on the overprovisioned disk. When we decrease the replica count, longhorn will remove one of the healthy replicas which doesn't appear to line up with the docs "If the current healthy replica count is more than the specified replica count, Longhorn will do nothing". Is this something harvester is doing on top to keep the count aligned with storage class?
The only item we can see in the events page for the volume is Removing unknown replica tcp://10.138.31.19:10694 in mode ERR from engine The volume will happily operate with an additional replica indefinitely until we decrease the replica count so don't think the new replica itself is the issue
Ah looks like replica auto balance, which I thought only applied when new nodes where added but I guess gets triggered on replica count updates as well