Preferred Leader leads to node unavailability problem #85
Description
The initial state of the broker group
n0 is master, preferredLeader
n1 and n2 is slave
The status after the re-election is as follows
n0 and n2 is slave
n1 is master
Problem Description
n0 node cannot synchronize data.
If the cluster restarts, the brokerId is always -1.
Problem Analysis
When the preferredLeader is triggered to re-initiate the election, if the ldeger entry index lags behind the leader node, the vote fails.
As the term increases by 1, the node is always larger than the other node's term and cannot be re-elected as a follower. Therefore, the preferredLeader will always be in the candidate state, resulting in the node being unavailable.
The log of the n0 node is as follows
The log of the n1 node is as follows
