# `Raft` Leader Election - Cont'd

![`smart`](RaftLeaderElect03_aa_002_20220511.png)

------------------------------------------------------------------------------
Next, we look at a special case, where after the Election Timeout triggers the election, more than two nodes increase their term and become candidate roles, and vote for themselves. And they just happen to receive the same number of follower agreement messages. At this time, the system will initiate a new round of voting until a unique leader is produced. In order to avoid the situation that a leader can never be elected or repeated elections, we can use an algorithm called random election timeout to reduce the probability of repetition. This algorithm can be understood as that each node randomly sets an election timeout within a range when starting the leader election, instead of using a unified value, so that it can avoid the situation that multiple candidates have the same number of votes with a high probability.