Priority election

Priority election is an alternative to the default raft leader election, where leader election is implemented by a random timer-based algorithm.

It aims to achieve a more uniform leader distribution by assigning each node a priority per partition and modifying the election algorithm to ensure nodes with higher priority have a higher chance of becoming leader.

Configuration

Priority election is enabled by default. To enable or disable it explicitly, set camunda.cluster.raft.priority-election-enabled to true or false in your config, or set the equivalent environment variable CAMUNDA_CLUSTER_RAFT_PRIORITYELECTIONENABLED. For details, see camunda.cluster.raft.

If you are using the fixed partitioning scheme (experimental), you may need additional configuration.

Limitations

With priority election enabled, election latency and thus failover time increases.

The result of a leader election is not deterministic, and priority election can only increase the chance of having a uniform leader distribution, not guarantee it.

Factors such as high load can prevent high-priority nodes from becoming the leader.