Windows Server 2012 : Continuous availability (part 2) - Failover Clustering enhancements - Simplified cluster management, Active Directory integration

Clustering and Active Directory improvements

In Windows Server 2012, Failover Clustering is more integrated with Active Directory. There are improvements made based on past experiences that administrators were running into.

One of the big call generators to Microsoft Support is the creation of the Cluster or names within the Cluster. When the creation of the Cluster or the names occurs, it would only create the Active Directory object in the default Computers OU. In many domain environments, the default Computers OU is locked down because domain administrators did not want objects created in this OU. When this is the case, you had to pull in a domain administrator to pre-create objects in the OU where the object needed to be, set permissions on the object, and do a few other tasks.

This tended to be a long, drawn-out process if there were issues because you had to wait for someone else to fix your problems before you could continue. Now, Clustering is smarter about where it is going to place objects. When creating a Cluster, it will look in the same OU where the cluster node names are located and create the Cluster Name in the same OU. So now you no longer need to pre-create the objects in a separate OU—Cluster is doing it for you.

Let’s take this a step further. Say in your domain environment, you wanted to separate the physical machines (OU called Physical) and the Clustered names (OU called Clusters). This is not a problem because you can pass the OU information during the creation of the Cluster. When doing this through the Failover Cluster Manager interface, you would input the name in this fashion:

If you wanted to do this in Windows PowerShell, the command would be:

New-Cluster -Name "CN=MyCluster,OU=Clusters,DC=Contoso,DC=Com"

Another call generator is the accidental deletions of the Virtual Computer Object from Active Directory. When a name comes online, Failover Clustering checks the objectGUID that it has for it to match it with the one in Active Directory. If this account is deleted, it would fail to come online. You had to go through a utility such as ADRESTORE.EXE, restore it from the Recycle Bin (if enabled), do an Active Directory Restore, or simply delete the resource and create it again.

This is no longer the case in Windows Server 2012 Failover Clustering because we have built-in “repair” functionality for just these instances. If the name has been removed from Active Directory, the resource will still come online. It will still log an event about the resource so that you are notified. However, it will give you time so that you can repair the object and not experience downtime. There is a “repair” option you can select where it will go into Active Directory and re-create the object for you.

Failover Clustering is no longer dependent on a writeable domain controller. In some environments where perimeter networks are in place, the perimeter network will usually contain a Read Only Domain Controller (RODC). Failover Clustering will now work with those environments because the requirement has been removed.

Along those same lines, we can talk about virtualized environments. For many companies, moving to virtualized environments is proving to be cost effective. However, there were “gotchas.” In some cases, cluster design was not planned to consider the need for a writable domain controller.

So, let’s say you want to virtualize all your domain controllers and make them highly available by placing them all in a cluster and storing them on CSVs. In the event that all nodes of the cluster are down, you are placed in a catch-22 situation: Cluster services and CSVs depend on a writable domain controller for domain authentication in the beginning, but your virtualized domain controllers need the cluster services running in order to start. The Cluster Service would not start because it could not get to the domain controller, and the domain controller would not start because the Cluster was down!

In Windows Server 2012 Failover Clustering, this has changed. The Cluster Service will now start using a special internal local account. All other nodes in the Cluster will start and join as it is using this special account. The CSVs would also come online. It is almost like we have our own hidden domain just for ourselves to use. Because the Cluster Service is started and the CSVs are online, the domain controllers can start.

We have made big strides in the way we integrate in Active Directory, and all of it is for the better. Cluster administrators spoke, and Microsoft listened.