Scaling down a cluster

General
#1

How can you scale down a cluster? I haven’t found any tools that’d obviously remove/evict a node from the cluster. The documentation says that a downtime of more than 10 minutes will cause the data to be rebalanced in the remaining nodes; however we still saw the node listed in the WebUI after some downtime. Additionally, I believe this should be configurable, too. An automatic rebalancing on an arbitrary time feels dangerous.

** Note: this is a copy of a community asked question on a different communication channel

#2

We are actively working on this - specifically making draining on node shutdown and how we handle scaling down more graceful. This will be ready by 1.0. Regardless, even though the removed node is still showing up in the admin UI, it’s harmless. Simply removing the nodes (one at a time and waiting for the number of underreplicated ranges to go to 0 before removing the next one) is enough to scale down the cluster.

#3

And it’s worth noting that you can see the number of under-replicated ranges in the “Ranges” graph on the “Replication” dashboard of the admin UI.