asked
ADVICE FOR SQL 2017 FAILOVER FOR HADR
![alt text][1]Hi guys So i have been given the task to create a clustered environment for our main/critical database. I am new to the topic and thus far from my research would suggest a 3 node window cluster, 2 of which are a sql cluster with shared san storage and the 3rd node on the DR site as a standalone instance with its own storage. Availability groups would be configured on Node1 with secondary replica on the 3rd node in the DR site. San storage is already in place so hardware and cost is not a problem. The server admin is under the impression that the we should have 2 nodes in the cluster, one on the primary site the 2nd on the DR site, each with local storage, eliminating single point of failure on the san and eliminating the need for a 3rd node. Am i correct in saying that we need to use shared storage between the nodes? Appreciate any input on this topic. Thanks in advance. [1]: /storage/temp/4574-sql-failover-and-availability-groups.png
Any chance the server admin is mixing up the Availability groups with Failover Cluster? The storage must be a shared storage between all nodes in the SQL Failover Cluster as all the data and log files of the databases resides in that shared storage. This is why the shared storage is vulnerable to be the single point of failure. Hence, you would need a shared storage for your 2 nodes in the cluster.
Yes that was exactly the confusion but I have explained it to him now and all is good! Given that neither he or I have experience in this I am thinking that the best approach is to simplify it for the initial phase. So configure a 2 node failover cluster. Once I am happy with the running of that and more confident with what I am doing I can then look to add a 3rd node to the WSFC as a standalone instance in the DR site and configure availability groups then. Our data is safe from the point of view of backups and replication of those to the DR site so I think the 2 phase approcah might be the safest. The actual request was to migrate the server to a newer version, I just wanted to do it in more future proof way and provide high availability so if I can achieve the migration with instance failover, I think they will be more than satisfied. Thank you for clearing up the shared storage issue and if you or anyone else has any tips or advice on this topic they are very welcome. I have edited the initial post to include my solution design so if anyone sees any issue with it please do shout :-)
18 People are following this question.
