I think it’s worth mentioning that you can answer ‘Am I the leader?’ by looking at the number of affected rows on the insert/update. Only the leader will change the row values so you can update the row and check leadership in one statement. In JDBC something like …

Properties connectionProperties = new Properties();
connectionProperties.put(“user”, username);
connectionProperties.put(“password”, password);
connectionProperties.put(“useAffectedRows”, “true”);

Connection connection = DriverManager.getConnection(url, connectionProperties);

PreparedStatement statement = connection.prepareStatement(INSERT_ELECTION_STATEMENT);
statement.setString(1, id);
statement.setInt(2, timeout);

boolean leader = statement.executeUpdate() > 0;

Note that you need to set the useAffectedRows (MySQL specific) property since JDBC specifies that the the number of selected rows should be returned.

—

Also, another benefit of your solution over GET_LOCK is that GET_LOCK doesn’t work with MySQL Cluster since the lock is not visible outside of a single MySQL process.

> I don’t want any server who is not leader to alter any data

Which data is that? Any data?

I wouldn’t incorporate an SQL leader logic into every single query in your app. Instead, your app can have a flag, frequently updated, saying “am I the leader”.
If you’re OK that the app will cease writing for a second as the leader changes, you can further require that a new leader only begins writing at least one second after becoming leader, during which time previous leader completes its own writes.

IF (SELECT count(*) as is_leader from service_election WHERE anchor=1 AND service_id=’my_service_number_7′) = 1 THEN
do update like insert or update or delete
END IF

In this way, we can prevent malfunctioning master polluting data. But I don’t know if we can express it in SQL