SERVER-25516, we should document a change to the "remove shard" procedure for versions 3.2.10 through the end of the 3.2 line (but not 3.4 and beyond).
Specifically, when removing a shard, if any mongos or mongod nodes start producing annoying log messages about hosts from the removed shard being unreachable, or if the operator wishes to re-add the shard using different hosts at a later date, the operator has two choices:
- Restart all the nodes with the annoying log messages
- Toggle the timeOutMonitoringReplicaSets server parameter introduced in
SERVER-25516from "false" to "true" on each offending node for 2 minutes, and then toggle it back.
This behavior will only be available in 3.2.10+, and will be unnecessary in the 3.4 and subsequent releases.