I'm trying to setup a load balanced Highly Available network on Red Hat Linux Enterprise 4. I've been trying to read up on whats available and whats possible, but still am unsure about the best route.

Configuration

I currently have two servers that ideally will be configured to run

• Apache
• Tomcat
• MySQL
• Shared Disk Space

I'd load balance requests to the server machines while concurrently accessing/updating the mySQL databases (one each machine for redundancy) and making reads and writes to the shared disk space.

From my research so far it looks like there are many products (both commercial and otherwise) that handle a piece of this problem. I'm looking for peoples suggestions for a whole package. I'm open to commercial and open source products.

To me it looks like I may want to implement Heartbeat for the server monitoring, with mySQL enterprise for database mirroring, ultra monkey or pirhana for load balancing, and DRBD for shared disk space. The biggest problem I see in this is the DRBD cannot support active/active clusters. Since we expect the number of servers to scale up, I can't do active/passive systems since the hard-drive space requirement would go up exponentially to do a primary->secondary for 10 machines each mirroring their data on the other 10 instead of a single shared data access.

I read some about GFS, but am not sure if it works in practice or not. Any advice is very much appreciated.

What is the nature of the shared diskspace? Is this something that the nodes can rsync from rather than actually share, even if there is a process that is updating images or what not?

Are you going to run every node with its own mysql server, an apache server and tomcat server?

Based on queries to the apache server, applications within another cluster must begin processing data. The output then needs to be available to any of the web server from that point forward.

The thought was to run an Apache/Tomcat cluster, mySQL cluster, and shared access cluster (where the processing will happen) After a request is made into tomcat the process cluster will be accessed to farm out the work.

As of this afternoon I have found an article that describes a partial solution, but I need write privs from any of the nodes, not just a master. Setting Up A Highly Available NFS Server | HowtoForge - Linux Howtos and Tutorials

what are the capabilities of rsync? could I rsync to the master's NFS drive from one of the other cluster nodes?

Hi All,
at work, we are facing the same problem for sharing filesystem. We decided to go forward with GFS. I think that's the most secure way, especially if your system is really critical.
I'll give my feedback, after installation