Am Mittwoch, 19. März 2008 12:02 schrieb Chris Sarginson - Technical Support: > Hi Michael, thanks for the quick response > > Michael Schwartzkopff wrote: > > Hi, > > > > no problem to accieve this setup > > > > 1) Loadbalancer: Why? Throw it out. Or use LVS with ldirectord as > > heartbeat resource. > > The loadbalancers are running ldirectord and heartbeat (Version 1), and > are configured solely as loadbalancers and need to stay that way - I > failed to mention this is a solution that is already in place that I am > having to rather work around.
So you have all together 5 machines. building one big cluster from all machines and use the localhost feature of LVS. But anyway, your setup 2 LVS and 3 HB machines will do very good also. > The setup that is required is to have all 3 webservers up and running at > any one time, all connecting to mysql on localhost for reads and the > high availability IP for writes. As it's using Master -> Slave MySQL > replication Web and MySQL should be running on all servers. I may > change the setup to do all reads and writes from the Virtual IP that > heartbeat manages, as I think this may make the whole thing more > manageable. Master/Slave instances of databases are difficult to handle in heartbeat because the resource agent cannot do this. You have to create two slave instances and one master instance of the database and group all with their according webserves. Add the virtual IP address to the group of the master database. Add constraints that forbid runing two groups on the same nodes. You also have to take care that the master instance run in any case, even if one slave instance has to be taken down. This should be possible with location constraints giving more points to the master than to the slave. I don't know exactly, you have to try this on a test cluster. > > > a) Make both into a group together with the IP address of the cluster. > > If you want to you can pin the whole group to one node with a location > > constraint. But this is not nescessary. > > This is where I'm falling apart, I really don't get the setup of the > cib.xml file, what processes interact with it, and how I can use it to > do a daisy chain type setup (server a -> server b -> server c). I know > once I have done this once I will *get it*, it's just getting this first > time out of the way. http://www.linux-ha.org/GettingStartedRevisedV2 If you still do not find your way, take one week of time to really read things - http://www.clusterlabs.org/mw/Documentation - German O'Reilly book to be published end of may. - go to a seminar, - or hire somebody to do the initial setup. This could also be done remote. > > > b) Make a group with MySQL and IP address 1 and an other group with the > > webserver an the IP address of the cluster. Make a colocation > > constraint that forbids both groups running on the same node. The > > webserver will access the database under the first IP address. > > I think this is closer to what I need - although I don't need a floating > IP for the web requests - the separate loadbalancer using ldirectord > takes care of http failing on any server, so I think my cibs.xml needs > to monitor only ping and the MySQL service in order to handle the failover You would need the floating IP address for the SQL queries, not for the webservers. In this setup you have one database that takes all the reads and writes. use the respawn method for pingd. See: http://www.linux-ha.org/pingd > > In both cases you have to think of o good database replication. DRBD is > > your friend! > > DRBD is something I have had a brief glance at, and does look quite > amazing (it's given me some ideas for future projects), but as I stated > earlier, I don't think this is a situation I can currently use DRBD in. Beware: DRBD only works on two nodes, not three! > > > Have monitoring installed in all ressources and pingd in all nodes. Make > > constraints that make the ressoruces running on the "healthiest" nodes. > > Again, I think this is set in the cibs.xml if what I have read is > correct, I'm just having a little trouble wrapping my head round it. Yes. Just add a monitor operation your resource. -- Dr. Michael Schwartzkopff MultiNET Services GmbH Addresse: Bretonischer Ring 7; 85630 Grasbrunn; Germany Tel: +49 - 89 - 45 69 11 0 Fax: +49 - 89 - 45 69 11 21 mob: +49 - 174 - 343 28 75 mail: [EMAIL PROTECTED] web: www.multinet.de Sitz der Gesellschaft: 85630 Grasbrunn Registergericht: Amtsgericht München HRB 114375 Geschäftsführer: Günter Jurgeneit, Hubert Martens --- PGP Fingerprint: F919 3919 FF12 ED5A 2801 DEA6 AA77 57A4 EDD8 979B Skype: misch42 _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
