Offline Node Timeout During Maintenance Mode #1412
Description
Is your feature request related to a problem? Please describe.
During the maintenance mode, partitions are assigned to nodes based on current states. It is possible for certain offline nodes to come back to life during the maintenance mode, in which case Helix doesn't differentiate them from other live nodes, and will assign them partitions based on their previous current states; this type of operations lead to state transitions from OFFLINE to another live state for the assigned partitions. From the customer point of view, this may cause new data bootstraps: if a node containing certain partitions has been offline for a certain period of time, data bootstrap may be required when it becomes live again. That is against the customer expectation of the maintenance mode.
Describe the solution you'd like
Offline node time-out feature: if a node has been offline for a certain period of time, the maintenance mode rebalancer will not assign partitions to it.