Hi all, I'm trying to setup a failover mysql server, and for that I need replication. Now here's is how I would do it (3 servers, A and B are located next to each other, C is off site):
A is master of B B is master of A B is master of C Now I don't think this will cause much problems, as the documentation describes circular replication is possible, so this should be possible too. To be sure not have collisions, the application only writes to A and will use B from the moment A is down (implementing this using a heartbeat trick with fake ip). As you can see, B is master of C, so changes go from A to B and from B to C. Now the questions come: 1) is it ok to put the following in each /etc/my.cnf file: [mysqld] log-bin server-id=<unique number> log-slave-updates <other, non-replication, parameters> 2) if B goes down, writes from A can't propagate to C. So on C, I would issue "stop slave", "change master to" (with master uid/pass), "start slave". But is this enough for C to be in sync with A then? Or does the "change master to" command also needs the master_log_file and master_log_pos parameters, as mentioned on http://www.mysql.com/doc/en/Replication_HOWTO.html ? 3) I need transactions, so I suppose I need innoDB type of tables/databases, correct? 4) if I create a table, do I need to specify "type=innodb" at the end? Or is this the default with 4.0.15? And if it is not the default, what should I put in my.cnf to make it so? 5) for logfile rotation/deletion (like the replication log, binary log, etc...), I would use the mysql-rotate-logs script (which does the sql command "flush logs") and delete all logs more than 5 days old (if all slaves are up. Is this ok? tx for any responses already! Franky -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]