What options are out there when Prod SQL Network and Standby SQL Network names are different and you need to failover quickly

  • DNS.

    You can look in wiki to get a very brief overview on it.

    http://en.wikipedia.org/wiki/CNAME_record

  • Perry Whittle (10/10/2011)


    parissa_bakhshi (10/10/2011)


    So in either way, if you use CNAME or A recore, you shall change the orriginal A record IP address.

    this is incorrect.

    The CNAME record does not affect the IP address assigned to the host (A) record.

    as I know in networking, there's no difference, what you wanna do can be done using both A records and CNAMEs.

    If you use CNAMEs, to have a failover, you have to change the CNAME to point to another A record! And if you use A records you have to change the IP address that the A record points to.So, ..no difference.

    The only matter is Clustering with to different networks!!!!! Clustering is a solution that needs no admin operations to fail-over, what you say about DNS are useful for mirroring, logshiping, replications,....not clustering. even in geo clustering you have to make a unique network to implement a real cluster. not a cluster that needs admin operations to failover!

  • The only matter is Clustering with to different networks!!!!! Clustering is a solution that needs no admin operations to fail-over, what you say about DNS are useful for mirroring, logshiping, replications,....not clustering. even in geo clustering you have to make a unique network to implement a real cluster. not a cluster that needs admin operations to failover!

    Do not overcomplicate things. Topicstarter does not have geo-cluster, they have to different clusters one of which is production and other is standby. Consider them as separate servers, replace 'cluster' to 'server' in the first post, if you wish.

    It doesn't matter how they deliver data to standby server - replication, mirroring, log shipping, service broker or anything else. They do not want to unite production and standby into one cluster no matter why.

    The problem is to switch between production and standby servers in minimal downtime time. If application does not support mirroring failover features the optimal solution (IMO of course) will be cName. That is all.

  • Can we assume you have a trust set up between the two domains? Obviously your users are in one domain and their user accounts must be known to and trusted by the standby clusters domain or windows authentication won't work. If this were the problem, it could explain why you were able to log in with the sa account if your sa passwords were the same.

  • They are. We are doing another failover test tomorrow. I will probably fail again.

    What I am going to do is this:

    1. Failover the databases to Standby server/Instance.

    2. Bring down Prod and Standby sever/Instances.

    3. Network guys are changing DNS.

    4. Bring up Prod and Standby server/Instances. <-It's either going to break here

    5. Users connect with application to Standby server/Instance. <- or break here.

  • Then in that case, you need to do what Perry Whittle suggested. You need to tell your DNS Administrator (if it's not you) to add a CNAME record that points to your production SQL Cluster Name. Then you will have to have all of your users make a one time change to their connection strings so that they now point to the new alias name.

    Then, in a fail over scenario, you would bring your standby databases online, same as before, and your DNS Administrator would change the value of the CNAME record to point to your Standby Cluster Name.

    I guess the only real question is, can you compell your users to make the necessary one time change on their end?

Viewing 6 posts - 16 through 20 (of 20 total)

You must be logged in to reply to this topic. Login to reply