failover replication

  • Hi Folks,

    this is probably a silly question but I need to make sure

    If you wanted to switch a application pointing from a publisher database to its subscriber could you do this without interrupting the people that are accessing the database through the application.

  • Actually it depends on your DB design.

    For instance you have an Identity column in the publisher, The subscriber will not be an identity column. It will either be an int or a Bigint. You have to alter the table. You also will have issue when the DSN is switched, You might have a lose of connectivity.

    In this case, why not use Mirroring or clustering? Clustering is the best for you. If one goes down, the other node will take over without human intervention.

    -Roy

  • thanks Roy.. I believe clustering is expensive - I don't think my boss will go for that.

    Presently we are using mirroring and having a small problem with applications having connection settings inside the app (therefore not transparently failing over). So when we want to failover changes to the app need to be made.

    This is a slight problem because our infrastructure team don't want to give me access to the application server so they need to change the settings. This is annoying because we are a 24 hour operation and downtime (failover) has to be late at night.. which mean someone from infrastructure needs to be here went I want to failover (usually by patching).

    I was just wanting to make sure that replication would not handle this better. That is to say if there was some way to do replication without causing any outage it would solve this problem.

  • Clustering is expensive and it will handle the automatic failover, though it can be 10-60 seconds to move.

    Mirroring is the best solution, but you have to be using the SNAC client, not older clients. If you move to this client (I think ADO.NET 2.0 has this) it takes two connection strings at startup and then if one fails, it tries the 2nd. Failover is very quick, seconds.

    With replication, you can move the data, but it won't change the client connection strings. There's no failover with replication.

  • In addition failover clustering does NOT recover client connections that could time-out.

    Connection interruption code *must* be present on the client-side code!


    * Noel

Viewing 5 posts - 1 through 4 (of 4 total)

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