1

If I have a 2-Node AlwaysOn Automatic Failover Cluster Instance with a shared disk and RAID-5 or RAID-10 to account for a SAN failure with the same setup in a second datacenter with data propagation via SAN Replication, is there a way to benefit from using AAG in this scenario?

We're not looking to use AAG for scaling-out, offloading backups or READ-ONLY requests. Scaling-out can be implemented via new nodes and NLB. A new node can be added to offload backups. I'm not seeing a use for AAG here. I think the one feature that we'd miss is automatic page repair. I'm not seeing any other uses. But I am new to AlwaysOn.

Thanks for the help!

PASQLDBA
  • 53
  • 1
  • 6
  • No, you have mentioned all the beneficial features of Always ON and that you wouldn't benefit from them. In that event, no, it's not worth it for you. I prefer FCI as it keeps everything in sync (logins, agent, SSIS) across nodes. Always ON does solve problems such as if you have no SAN replication possible or need to off-load reads or backups, but if you aren't going to utilize those features, then FCI probably hits all of your requirements. – Jonathan Fite Apr 07 '16 at 15:31
  • Actually no, FCI Only hits your requirements if you can live with the database going down due to data corruption. There is a LOT that can go wrong with SAN level replication (such as the SAN happily replicating a corrupted database thanks to a controller malfunction or server RAM error). AOAG are explicitly there to avoid the single point of failure that the database file is. – TomTom Apr 08 '16 at 05:10

1 Answers1

1

If I have a 2-Node AlwaysOn Automatic Failover Cluster Instance with a shared disk and RAID-5 or RAID-10 to account for a SAN failure with the same setup in a second datacenter with data propagation via SAN Replication, is there a way to benefit from using AAG in this scenario?

That depends, are you worries about any of the following scenarios?

  1. Strict RPO/RTO
  2. Does your SAN support write ordering across multiple LUNs?
  3. OS/SQL Patching issues
  4. Controller level IO corruption
  5. Downtime for Maintenance

If your answer to the above is, "Not worried" then I don't see any reason to use any AlwaysOn technology (FCIs or AGs). If your answer is, "Yes, one or more worry me" then I would consider looking into a set of technologies to meet your specific concerns and scenarios.

Sean Gallardy
  • 32,288
  • 3
  • 40
  • 77