Question

Design a leader-election / coordination service so that a cluster of 30 stateless scheduler replicas elects exactly one active leader to run cron-like dispatch, with standbys ready to take over within seconds of a leader failure. Constraints: at most one leader acting at any instant (double-dispatch is harmful), failover under 5s, and the design must tolerate network partitions without two leaders both dispatching. Describe the election mechanism, how the leader proves it still leads, and the partition behavior.