Do you have to use a shared disk with AlwaysOn FCI in a 2-Node FCI? If not and if you have a 2-Node FCI and each node has a local SAN Disk can failover be automatic and will the whole instance failover automatically? My concern is the single point-of-failure with a single shared disk.
Thanks!
Best Answer
There's two main forms of failover you can use with SQL post 2012,
The Windows Failover Cluster (WFC) which has been around for a while and the Always on High Availability Group(AO)
WFC uses a single database which can move between multiple machines this method requires the database to be installed on shared drives (such as a SAN) so that when the cluster moves the node the other machine has direct access to the information that makes up that database
AO uses two completely seperate instances of SQL Server, both instances run at all times, (depending on how close they are you can set them up to be synchronous or a-synchronous, (if they're on the same network synchronous is better if they're separated globally a-sync is a better choice), due to there being two databases you do not require any shared storage, the two machines work completely independently of each other, we currently have ours set up with a quorum witness which is actually on a shared drive but is not technically required (it is advisable to have a witness however)
Worth Noting about using AO, if you use the secondary node for any data access / backups (which you can as it can be accessable through readonly) you will require the extra licences to access it, if it sits there purely as a redundancy for when your primary node fails you do not require any extra licences as only one database will be accessible at any one time, with WFC since there is only actually one instance you only require one set of licences for that instance as it moves, you do have the disadvantage of only one disk failure will cause problems but using a RAID 10 on those disks will limit this from becoming a problem
Ste