Jeremy Zawodny ([EMAIL PROTECTED]) wrote: > >> >http://jeremy.zawodny.com/mysql/managing-mysql-replication_files/slide0121.htm >> > >> >Just make B a slave of A, C a slave of B, and A a slave of C. >> >> This is a good approach. > >Except that it's fragile. One machine can go down and "break the >cycle" unforunately. :-(
Yes, this is the problem and why a spread solution should be more robust, we have just to find a procedure to re-sync a node after a crash. >> We are thinking about a parallel replication by use the Spread >> Toolkit (www.spread.org) to broadcast the writes queries on all the >> servers in the same time and only once with the guaranty that the >> messages are well received by each server, and load balance the read >> queries. Then we can have a very scalable clustering solution with a >> minimum network load. > >Yeah, putting Spread support in MySQL for something like that might >not be a bad idea. Perhaps we can use the binary log mechanism in output and input with Spread to broadcast the changes, then the load balancer send the queries on the pool servers and only the write queries should be broadcasted by the server who received it, then each server become master or slave or both since Spread has ordered messages. First we have thinking to use Spread to manage locks and thread communications among nodes in a cluster to use distributed filesystem but it's a big deal ! >I've talked with two of the MySQL developers (Monty and Nick) about >solving auto-increment problem. After my Replication talk at OSCON >2002, we kicked around some ideas that seemed quit promising. I just >don't know when they'll be implemented. Hopefully Nick has them on >his TODO list. :-) Good ! Perhaps we can begin a new project... Sam. --------------------------------------------------------------------- Before posting, please check: http://www.mysql.com/manual.php (the manual) http://lists.mysql.com/ (the list archive) To request this thread, e-mail <[EMAIL PROTECTED]> To unsubscribe, e-mail <[EMAIL PROTECTED]> Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
