SAN vs. NAS Shared Storage for Failover Clustering: A Detailed Comparison

There are several elements that make this architecture complex to implement:

• on failover, switching the shared storage requires low level instructions which are storage manufacturer dependent,
• recovery procedure on the file system (FS) must be passed before restarting the application,
• if both file systems on both nodes access the same raw disk at the same time, the full file system will be corrupted,
• to avoid a double access, a quorum disk must be configured.

There are several elements that make this architecture simple to implement:
• on failover, switching the shared storage means only remounting the external file system,
• no recovery procedure on the file system must be passed before restarting the application,
• if both nodes access the same shared file system at the same time, the full file system will be not be corrupted,
• however, there is still the possibility of a double execution of the same application corrupting its data in the shared storage when nodes are isolated.

There are no such issues with SafeKit because its replication and failover solution does not require a shared storage.

However, if SafeKit must manage a shared storage:

• use a NAS SMB shared storage or a NAS NFS shared storage,
• put in the restart scripts the mount/umount of the external file system,
• configure the SafeKit split brain checker to avoid a double execution of the same application accessing the shared storage when nodes are isolated.