Exchange 2010 and HA/DAG's


I've put questions here before about this but I'm still not quite getting it.

I have a single server now running Ex2010 in a lab. It's functioning as a real valid mail server, using a real test domain that I created.

I want to implement DAG's because, as I understand it, it will give me real-time database copies of my mailbox database.

So, that means I need to install a second server, join it to the domain and install Exchange 2010 on it.
The first question is: do I only need to install the Mailbox role on the second server in order to implement DAG's? I think this is true from what I've seen.

The second question is: Can I install also the HUB Transport and CA roles on this second physical box? The reason I'd want to do that would be to have a High Availability solution in place in the event my first Exchange box failed for some reason.
So this leads me to question 3 -- can I have a HA solution for Ex2010 with just two servers? I remember reading someplace that Microsoft recommends 3 servers at a minimum.

If I can't have HA with just two servers, but I can set up DAG's with just two servers, what would I have to do in the event Exchange server 1 went down? Would the DAG's automagically redirect to Exchange server 2 or would I need to intervene to make this happen?


Who is Participating?

Improve company productivity with a Business Account.Sign Up

Paul MacDonaldConnect With a Mentor Director, Information SystemsCommented:
NLB is not clustering.  NLB is load balancing and implies traffic is divided among several hosts, all running concurrently.  Clustering implies traffic sent to one node (the active node) among several hosts.

I don't know how else to assuage your fears other than to say we run two 2010 servers here.  Each server hosts Hub, CAS and Exchange services.  Both nodes hold copies of all mail databases and the Hub and CAS roles are clustered so if I lose either server, the other one can take over completely.  For additional disaster recovery/HA, we contracted for a remote node about 100 miles away with which we replicate our mail databases via CCR.  Unless something takes out Virginia from Norfolk to Richmond, I expect I'll have mail.

In your case, the loss of a single server should go unnoticed.  Mail is sent to the DAG.  Clients are sent to the DAG.  Loss of a node within the DAG just means another node takes over.  Both Exchange servers have copies of the mail databases.  Corruption on one server shouldn't imply corruption on all servers so you can just roll over, kill the bad copy and force a new one.  If all copies are bad, restore from backup.  
Paul MacDonaldDirector, Information SystemsCommented:
Yes, the second server must have the mailbox role.

Yes, you can install the Hub and CA roles on the second server.

Yes, two servers is essentially an HA the extent the facility they are in isn't destroyed or access to the servers isn't cut off.  Two servers is better than one.  Two (or more) servers spread out geographically is even better.
nachtmskAuthor Commented:
Thanks for the quick reply Paul.

Excuse the simplistic question here -- so, if I have all of the roles installed on Both servers, which server is actually functioning as my "Exchange" server -- are they both? Is there some setting I have to choose when installing to designate one server as the primary.

Are you saying I don't need a third server to have HA? I thought I remember reading something about windows clustering and the CA role regarding a third server??  That might also have been talking about Load Balancing. I honestly don't remember and can't find it again.

Has Powershell sent you back into the Stone Age?

If managing Active Directory using Windows Powershell® is making you feel like you stepped back in time, you are not alone.  For nearly 20 years, AD admins around the world have used one tool for day-to-day AD management: Hyena. Discover why.

Paul MacDonaldDirector, Information SystemsCommented:
IIRC, Microsoft wants you to offload the CA role to an external server (or cluster of servers) for HA.  We simply clustered the role on our extant Exchange servers.

The gist is that mail coming in to the network gets queued by the Hub, pending the mail going into the mail database.  Once the mail is in the database and replicated, it's de-queued.  That way, if Exchange crashes before the message is stored/replicated, there's still a copy of it in the Hub, which can then pass the message on to any Exchange servers still running.

Check these items for more information:
nachtmskAuthor Commented:
Thanks again.
So, will two physical servers give me redundancy? I'm reading other stuff online. Please look at this quote and tell me what you think:
I got it from:

"You see, since DAGs make use of the Windows Failover Clustering (WFC) component to an extent (primarily heartbeat and the cluster database), you can’t configure the two servers as nodes in a Windows NLB since it’s unsupported to use both WFC and WNLB on the same server. This has been unsupported since Windows NT 4.0 and is due to potential hardware sharing conflicts between the Cluster service and WNLB. Read more in KB article:

This means that you must use an external load balancing/fail-over device such as a hardware-based load balancer. Also note this balancer should be redundant, so you need a minimum of two devices."
Paul MacDonaldDirector, Information SystemsCommented:
In that article, they're using NLB to allow them to run two Hub/CAS servers simultaneously because they (evidently) have so much mail/client traffic that they need two servers to handle it all.  There's nothing wrong with that design and it may be typical for larger networks.

I don't know what a reasonable cut-off is for making that decision, but on our network, we have the Hub/CAS servers clustered so only one is running at a time.  We get HA/fault tolerance but we must not have as much traffic as they do.  For the record, we're supporting less than 100 mailboxes.
nachtmskAuthor Commented:
I don't need to use NLB for performance. I have about 200 users but most don't check their mail often.  I'm more concerned about HA.

My plan is this:
Run two physical servers
Run all Ex2010 roles on each server
Create DAGs across servers.

In the event of a server melt down or DB corruption I would be able to use the other server or one of my database copies.
it seems like this is all possible using just two physical servers to me, yes??
But -- if a DB gets corrupt or if a physical server dies, what will I have to do in order to get the other server to take over? A few mins or even an hour downtime is acceptable in my situation.
nachtmskAuthor Commented:
You said:
"we have the Hub/CAS servers clustered so only one is running at a time."
It's possible to cluster CAS/HUB on only two servers while using DAG's? From what I'm reading (and quite possibly not understanding correctly), it says that clustering (NLB) can't happen if it's running on the same server as the CAS.
I'm still confused...
nachtmskAuthor Commented:
I see now! I thought NLB was clustering. That's why I kept thinking I needed 3 servers, because everything I read said three servers were needed for NLB.

One last question.
You said both your HUB and CAS roles were clustered. How do you cluster these roles? Simply by having them on two servers that are domain members?

Thanks. You have been very helpful.
Paul MacDonaldDirector, Information SystemsCommented:
Yes, I've created a DAG cluster on the Exchange servers themselves to host the Hub and CAS roles.
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.