System Design Interview: How to persist leader state during failover?
May 10, 2021
3 Comments
I understand we can use external linearizable store like Chubby, Zookeeper for leader election. But what is the best way to persist leader state?
Example: in the UniqueID problem, say there is a leader handing out ranges of unique ids to many sharded distributed servers. It needs to persist the ranges that it handed out, so that during failover the new leader knows where to pick up. How does it persist that state for failover?
Note: I am not interested in knowing a different solution for UniqueID, rather want to understand how to do the failover correctly. UniqueIdD is just an example here
#systemdesign #leaderelection
comments