We are looking at implementing some redundancy into our mail system, which is currently running on a sendmail/postfix platform off a SAN on individual smtp and pop servers. We would like to build redundancy by either mirroring the configuration on the smtp and pop servers to other nodes that could be load balanced to accept the incoming traffic (we usually have an average of 1.25 million emails traverse our mail system daily, most filtered) using NFS for shared storage, or building a cluster on the Redhat GFS2 file system, or possibly use some method that is better??
The whole idea is to make this system as simple as possible to manage and not have multiple layers of failure to deal with in the case of an emergency, but we cannot seem to find documentation that would push us in any direction over another in regards to building a shared storage system.
I would really like some information on the IO constraints that are apparent with NFS, and whether or not building NFS file servers to attach to the FC disks would be a good way to server the files to the mail servers, or if using GFS2 would be more efficient and use better file locking.
Thanks in advance!
