Leonardus Setiabudi <[EMAIL PROTECTED]> wrote on 04/07/2006 05:33:50 AM:
> Hi All, > > I have a problem with my multi master replication plan. > I have set my server (linux) to run 4 instance of mysqld, each with a > different port, socket, tmpdir, log file, relay log file, master info > and relay info. > > the illustration as follow : > > Server 1 (non slave) > -------- > db1 > db2 > db3 > db4 > > Server 2 (slave from other machine) > -------- > db2 > > Server 3 (slave from other machine) > -------- > db3 > > Server 4 (slave from other machine) > -------- > db4 > > each slave run well and replicate successfully > but the problem is, Server 1 unable to see the update from the other > slave server, although i can see the update from each originating slave > server. > > Is this a problem with mysql (locking issue?), or file system issue? > I am using mysql 4.1.11 on mandrake linux, with ext3 fs and all the > tables were MYISAM > > Server 1 was supposed to be somekind of data warehouse, to provide data > from many distributed servers phisically separated. > > Would someone please help me, i have googled this and red the manual, > but couldnt find similiar situation. > > TIA > - > Leo > Just so that we are on the same page, let's review a few definitions as they relate to database replication: master - the server from which changes will propogate during replication. slave - the server to which replicated changes will be applied during replication. According to the rules of MySQL replication, as I understand them, each master can be a host to many slaves but each slave can only receive changes from a single master. You said: > each slave run well and replicate successfully I assume you mean servers 2-4? > but the problem is, Server 1 unable to see the update from the other > slave server, although i can see the update from each originating slave "other slave server"? I don't quite understand. If Server 1 is supposed to be receiving updates then it is the slave and the process it is getting the changes from is called the master. A single server process can actually have both roles as in this situation: Server A -> replicates to -> Server B -> replicates to -> Server C In this case A is the master of B. B is a slave of A. B is also a master to C which is the slave of B. Multi-master replication would be something like this: Server A -> Server C Server B -> Server C In this plan, Server C would be recieving changes from both A and B. However the current design of MySQL limits you to having only *ONE* master per slave for many excellent reasons (search the recent archives of this list for a few). Can you possibly re-explain your situation using the terms master and slave as I just used them? Shawn Green Database Administrator Unimin Corporation - Spruce Pine