Exchange Exchange
A community dedicated to Exchange and related technology.
Forums » General Exchange and PowerShell » Exchange 2007 » CCR Heartbeat Options

CCR Heartbeat Options

rated by 0 users
This post has 2 Replies | 2 Followers

Top 75 Contributor
Points 0
Reply
lashah28 Posted: 02-07-2007 11:39 AM

Can the CCR Cluster Heartbeat use Static IP Addresses on the public network. Our Nodes resides on different subnets and different physical locations. They both reside within the Firewall so they are able to communicate with each other with no problem.

’ve been reviewing a few TechNet documents on MS Exchange 2007 CCR (Installing Cluster Continuous Replication, How to Configure Network Connections for Cluster Continuous Replication, Planning for Cluster Continuous Replication). There is a section in the document “Installing Cluster Continuous Replication” called “Configuring the Cluster Heartbeat,” and in table 2 the third option (All Communications (mixed network) it says that this option is selected by default for all network.

Marcus Walker System Manager
  • | Post Points: 0
Top 10 Contributor
Points 3,290
Reply
Joel Stidley replied on 02-07-2007 3:11 PM

The limitation with Windows 2003 is that all servers in a MSCS cluster have to be on the same subnet.  This is true for the public interface and the private interfaces.

The heartbeat packets have a TTL of 1 so any network device that it would transverse would expire the packet. 

In order to have this same subnet at both locations some network trickery will need to be accomplished.  More information about 2003 clusters here.

http://download.microsoft.com/download/7/6/f/76f3db2f-6f43-4624-bfde-ff731e3c1f96/GDClusters.doc

The good news is that Longhorn and the subsequent service pack of Exchange 2007 should change these limitations.

 

- Joel

Top 10 Contributor
Points 3,290
Reply
Joel Stidley replied on 02-08-2007 3:48 PM

Marcus and I traded some emails on this.  Hope you don't mind Marcus :)  My last e-mail to him went something like this:

Since both nodes need to be on the same non-routed subnet that configuration isn't going to function.  If you think about it, since the nodes are on different subnets what IP address would you be using for your Exchange/CCR instance and how would you fail it between the sites?  You need to be able to fail-over that ip to the secondary site.
 
This information is in the Microsoft documentation (this talks about using a MNS cluster at three sites... having the voting node at the third site):
 
  1. The private and public network connections between cluster nodes must appear as a single, non-routed LAN (for example, using technologies such as VLANs to ensure that all nodes in the cluster appear on the same IP subnets) between the three sites.
    1. All three-site public interface must be on same subnet
    2. All three-site private interface must be on same subnet
  2. The network connections between the sites must be able to provide a maximum guaranteed round trip latency of no more than 500 milliseconds.
As for the heartbeat network there is still a router between the two servers, and each hearbeat has a TTL equal 1.  This will cause the heartbeart to be dropped at the first router.  http://support.microsoft.com/?id=307962
 
Another document that I found very helpful is this one.
 
I feel your pain on this, I was very excited about CCR when I first heard of it but then reality set in and it lost its appeal, at least for now.  I think the best answer (without having to run OSPF on each server and doing some crazy network routing) would be to wait for Longhorn to implement the multi-site CCR and stick to a single site for now.  Longhorn promises to fix many of these as well as other limitations.
 
Hope this was helpful.

- Joel

  • Filed under:
  • | Post Points: 0
Page 1 of 1 (3 items) | RSS
© 2003-2008 NamedPipes Consulting. All other company and product names are property of their owners.
Powered by Community Server (Non-Commercial Edition), by Telligent Systems