Want to protect your cyber security and still get fast solutions? Ask a secure question today.Go Premium

  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 1786
  • Last Modified:

Restore exchange cluster to cluster with different name

Hello, I have Exchange 2003 on 2003 Server in a three node cluster with 2 EVS's located on east coast. Lets call the EVS's "mail1" and "mail2". I also have a 2 node cluster on the west coast with one EVS called "west1". They are also Exchange 2003 and 2003 Server. Both sites use a MSA-1000 for shared storage. The sites are linked via a 100mb MPLS. All offices connect to the east coast for mail, the west coast is just for failover. My problem is that I need to have failover capability with the most current data should the east coast (mail1 and mail2) office go down. My thought was to rename the west coast EVS to "mail1" take it offline and perform a restore with the data from the original "mail1" server. Then take a snapshot with shadow copy every 2 - 3 hours. I would then use HP Open View Storage Mirroring to replicate the data to the appropriate locations on the west coast. I am thinking that if I ever had to bring the east coast online it should have no problem mounting the stores. Does that make sence or have I just lost my mind? My other question is how would I get the data from "mail2"  to restore onto the east coast server?

Thanks for any help.
  • 2
2 Solutions
As long as the "Org" name "Administrative Group" name "Storage Group" name and "Store" names are identicaly, it should mount. As long, of course, it's a "Clean Shutdown" and the databases are consistent. It's very similar to creating a "Recovery" Environment. You backup data every day right?

Here is a quick giude to do this for Thousands of users. It's our process for a "Disaster Recovery " scenario

1. Server Crashes - But it can be powered on, howerver all databases corrupt
2. First thing, Rename all Storage Group folders and trans logs to another name
3. Create empty folders with the same name.
4. Now go to ESM and MOUNT EMPTY Databases. This will allow users to get Email SERVICE back up again withing Minutes if needed.
5. Now try to repair your "Corrupt" databases offline. If you can't do it . Go for a restore
6. At an evening time STOP SMTP Virtual Service on cluster admin to stop all mail flow
7. Exmerge ALL DATA current for that date to a temp folder
8. Perform a Full Restore from whatever backup software you from the evening prior OR Last avail.
9. Once all stores are mountable (may have to patch as last resort) Mount the stores.
10. Customers now have last nights email.
11. NOW, Exmerge all the Data you saved BACK into the database.
12. Now you've caught them up to that point.

Bottom line, i think there is an easier solution than what your trying to do. Also, with so many clusters the chances of Exaclty 2 clusters of the same Virtual server crashing at the very same time is kinda unusual. Unless you don't have stability in your cluster in the first place.

best of luck.

STB - The game that you are going to try and play with the restores is a dangerous one.  What clustering software are you using?  Are you using an F5 or similar device to "stretch/geostretch" the subnet to the passive node in the west?  Expectations are that you have Active Directory, DNS, etc. out there.

The restore could take a significant amount of time to perform, dependent upon the amount of data.  Where would you perform it if the East Coast office is down?  Even Delta Dash of the tapes to the west coast could take nearly 1/2 a day after you get them to the airport.  If it's 6PM and, say, an earthquake happens on the East Coast, you may not even be able to get to your backup tapes for a day or two.
If it was something less significant, such as the power outage in NY, what is to say that you are going to be able to connect out there in order to do the restore?  If it is something less significant, such as a fiber line that gets cut by a backhoe, you have to ensure that mail1 is down and will not come back up.  Then there is the issue of the IP addresses, updating DNS, etc.  When the system restores it may not be able to authenticate the new secure channel password if the PDC Emulator FSMO is down.

Realistically you should look at something like Veritas VVR and VCS (Veritas Volume Replicator and Veritas Cluster Services) in order to do a geographically stretched cluster.  With the bandwidth and hardware, and correct SAN configuration, Exchange should be able to handle this without a problem.  Optionally, dependent upon the Storage Mirroring, your infrastructure, and the equipment, you may be able to have the SAN replicate with another SAN on the other end and west1 would just "take over" without a problem when it detects mail1 and mail2 are down.  The problem will be when they come back up.  If you do a restore of mail1 to west1, then what happens to the west1 cluster?  What happens if mail1 is brought back up and someone forgets to pull the physical network cabling and it and the restored mail1 collide (this is similar to split-braining...but it is a bit worse).

Many companies face these questions and they are all tough questions.  Once you have the answers to those questions you will be closer to hammering out a solution.
stb3222Author Commented:
Im sorry, I dont think I explained it properly. This is not a geo-cluster. The east and west cost servers are completly independent from each other. An exampe of what I am trying to accomplish. Florida gets hit by Hurricane and looses power for 2 weeks. Before power goes out I need to be able to fire up the servers on the west coast and have all users from accross the country connect to there, via OWA for users that are not in the east coast office.
NOTE:  This scenario (DR planning) is not cheap, either way you put it.  If you are worried about just the hardware, go with good hardware from vendors like HP/Compaq.  if you really want uber-redundancy go with something from Stratus (www.stratus.com).

So, it is not a cluster at all?  Where does the three node cluster that you mentioned fit in?  Is it just the SAN replication?

First, you can't do what you want to do with Microsoft Clustering services.

Well, I would suggest that you look @ a geographical cluster-type application if you are serious about protecting your infrastructure (or if your organizations are in Florida, or within 150 miles of any of the costal areas, or a geological fault line).

Veritas should talk to you and demo their product for free...but, I warn you, their consulting rates are not cheap should you decide on their product.  Also, the training is needed as well.  If you go with Veritas, make sure to specify a VCS/VVR expert that is also an Exchange Expert after they give you the quote, but before you accept it.  Too many times you will get someone who is an expert with VCS/VVR, but they do not know all the intracacies of how Exchange interfaces with their software (but they may be experts with SQL, Oracle, etc).  

Legato should talk to you and demo their product for free...but their consulting rates are not cheap, should you decide on their AAM solution.  AAM would be able to do what you want, but it takes a solid infrastructure (DNS, Active Directory, SAN, replication, etc) to function properly.  AAM would show mail1, mail2, and west1.  It is an active/passive environment, meaning that one of the servers (west 1) would have to be passive.  It "tricks" AD and it updates DNS via script when there is a failure.

Make sure that you have a demonstrated performance clause in there...that they have to get this working for x amount of $$.  Also, make sure that two members of your staff go to the training class...people do leave and people do get fired/laid off.

The restore stuff that you want to do sounds good, but it can cause you a nightmare in a real DR or test DR scenario (see my notes above).

The downsideis that you can't failover an active Exchange node (mail1 or mail2) to an active Exchange node (west1).  The node that you failover to will have to be a passive node.  Active/Active/Passive clustering is only available on Microsoft clusters.  Another downside is that Microsoft will not allow a passive node to function for two separate clusters.  I think the same applies with Legato and Veritas.


Featured Post

Hire Technology Freelancers with Gigs

Work with freelancers specializing in everything from database administration to programming, who have proven themselves as experts in their field. Hire the best, collaborate easily, pay securely, and get projects done right.

  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now