Multiple identical Windows 2008 Clusters for file sharing, how to create a failover?

Home Forums Server Operating Systems Windows Server 2008 / 2008 R2 Multiple identical Windows 2008 Clusters for file sharing, how to create a failover?

Viewing 1 post (of 1 total)
  • Author
    Posts
  • Avatar
    isaaca
    Member
    #164284

    (Full thread title: Multiple identical Windows 2008 Clusters for file sharing, how to create a failover, since DNS CNAME is not an option?)

    Hello,

    We have multiple SAN devices that expose block storage to multiple Windows 2008 Storage Edition clusters, which consume them via iSCSI. To simplify, here is what we have:

    machine1+machine2 –> cluster –> storage1 (accessible via \storage1.domain.nameshares)

    machine3+machine4 –> cluster –> storage2 (accessible via \storage2.domain.nameshares)

    The 2 clusters have replication from active (storage1) to standby (storage2), so in theory they both include pretty much all the files.

    We have hundreds of applications that are pointing to \storage1shares. In case of a disaster, however, it is an immensely laborious task to change all the end-points from \storage1shares to \storage2shares, and then back.

    I was trying to declare a CNAME (\activestorage) in our internal DNS, and point it to the currently active storage system. I tried both an alias and an IP pointing to the cluster. However, when I try to access \activestorageshares, I got an error that the network share does not exist. I have been reading about this, and now I understand that CNAMEs are not possible in 2008 and only allowed in 2012 for this particular purpose.

    I also understand that I may have to register a CAP (Client Access Point) to create an “alias” for my storage. However, the CAP approach does not seem to address (or does it?) my initial problem – I am not looking to create an alias just for a particular storage cluster, but rather for the active storage cluster. This way, if disaster strikes, I can simply go and change an IP somewhere, and all the applications will starting using the standby storage system.

    We were suggested to use DFS (only for namespacing resolution, not for replication), and this is something we’re currently looking at. If there is something a little easier, and less dependent on other machines (such as AD for DFS), that would be great.

    I realize that there is an extra administration of ensuring that the two clusters are identical from shares perspective, and that’s something we are already doing. All I need is a clue into how to make this architecture work in a less painful way.

    Thank you for your help,

    Isaac

Viewing 1 post (of 1 total)

You must be logged in to reply to this topic.