Racked & Stacked - 3
Both fileservers are connected via independent FC switches to both controllers on each SATABeast array.
The SATABeasts are operating in "all LUNS, all ports" mode which exposes all drive volume LUNS across both internal controllers.
This allows each fileserver to see each RAID6 volume twice via independent paths through different switches, controllers and HBA's. Linux multipath and device mapper code handle load balancing and failover should one path fail.
The idea being that this gives us a nice level of safety and redundancy.
The final bit of secret sauce is a RHEL fileserver cluster using clustered GFS and LVM to stitch the volumes together into a single ~90 terabyte volume.
Racked & Stacked - 3
Both fileservers are connected via independent FC switches to both controllers on each SATABeast array.
The SATABeasts are operating in "all LUNS, all ports" mode which exposes all drive volume LUNS across both internal controllers.
This allows each fileserver to see each RAID6 volume twice via independent paths through different switches, controllers and HBA's. Linux multipath and device mapper code handle load balancing and failover should one path fail.
The idea being that this gives us a nice level of safety and redundancy.
The final bit of secret sauce is a RHEL fileserver cluster using clustered GFS and LVM to stitch the volumes together into a single ~90 terabyte volume.