That sounds like an ideal application for DRBD, not using MySQL's replication at all. In my application, since it was mostly read-only and was heavily loading a single server, we wanted active-active servers both serving the same data for load sharing. It sounds like you're just planning on active/standby. I would set up a DRBD filesystem and let MySQL think it's standalone. You can have linux-ha stop and start the MySQL server at the same time as switching the Virtual IP address and the DRBD master.
- Dave On Thu, Jun 28, 2007 at 05:26:58PM +0200, Adrian Chapela wrote: > Hi Dave! > >What does it mean to be a master server when there is only one machine? > >Do you mean you want it to be able to do updates and send the changes > >back to the other when the other comes back up? > > > >MySQL has (at least) 2 ways of doing replication. One is master/slave > >where everything that changes on a master is copied to a slave. > >I've got an application set up where most of the operations are > >read-only, so if the master is down we just assume that updates > >cannot be done. I don't think MySQL reliably supports swapping the > >master/slave relationship with this kind of replication. > > > OK, but You can configure both servers to replicate in both directions > but you can only use one at the same moment. > > A replicate to B and B replicate to A, but you can only use A. When A > goes down, you can use B as a master because when A goes up it will be > replicating all again and it will be up to date soon. But in this moment > you can only use B until A was up to date. When A is updated you can use > it again but You must leave B as slave now... In my problem I only want > solve the first step, A -> goes down -> Heartbeat detects the problem -> > Heartbeat execute the scripts to make B a master -> B is a master with > the same VIP as A (and now the applications run OK!!). The second step > is to automate this second change (this is a bit more difficult than the > first change..) > >The other way MySQL does replication is called MySQL cluster. > >With that method the various servers are peers and either one can do > >updates and they get automatically copied to the others even as the > >servers leave and join the cluster. Maybe you need to use this method. > > > > > MySQL is a "bit" complex than replication and it requires more hardware > than replication. (Too much memory for a large database or too much > machines!! I have a about 130 GB database. It isn't large for MySQL but > for me could be...) > > Now I am documenting more about Heartbeat V2 and I am trying to solve > some basic steps. Then I think it will be more easy. > > Regards... > >- Dave > > > >On Tue, Jun 26, 2007 at 01:36:17PM -0400, Madison Kelly wrote: > > > >>Adrian Chapela wrote: > >> > >>>Hello, > >>> > >>>I am trying to make a MySQL system with high availability. One Master > >>>server and one slave server wich will assume the master functions when > >>>the master server goes down. > >>> > >>>My problem is to make a rule for the execution in slave server when > >>>master server goes down. When master server goes down, the slave server > >>>assume the VIP and then must execute a script to test something and then > >>>the server can go up but it can't go up until the first test has > >>>finished. > >>> > >>>Can you say me how to write this rule ?? > >>> > >>>Thank you! > >>> > >>Can you give a bit more details on what you've gotten setup so far? Are > >>you using heartbeat v2? Are you using the /etc/ha.d/haresources file or > >>CRM? Are you using DRBD? > >> > >>In my case (still working on DRBD issues) my ha.cf file is: > >> > >>http://tle-bu.org/misc/ha.cf.vash > >> > >>And my haresources file is: > >> > >>http://tle-bu.org/misc/haresources.vash > >> > >>Do note though that the haresources file above may be broken. As I > >>mentioned, I am currently working on a DRBD problem. However, the > >>commented out line (second from the bottom) works fine. > >> > >>If you can provide more details and/or ask specific questions I might be > >>able to help you more. At the moment though, I'm still learning as well, > >>though I have gotten past a few of the early trip-ups. > >> > >>Madi _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
