How to setup a failover/replication server

Hello experts,

We are trying to setup a failover environment in our office for disaster recovery or hardware problems.

Currenlty we have one SBS 2003 unit which is the DC and runs Exchange (In the process of migrating to EBS or SBS 2008), two server standard 2003 (one runs SQL and a few apps, and the other is file storage), and one server 2008 Standard running terminal services.

I know that in order to setup a cluster we have to upgrade all of our servers to Datacenter or Enterprise, and have identical hardware on the units (which we don't have).

What will be the most cost-effective way of achieving this?
Will DFS work?
I've also read about Double-Take, and was wondering about Hyper-V or VMware.

Your input is appreciated.
Who is Participating?
qf3l3kConnect With a Mentor Commented:
AD Redundancy

If you want to provision redundancy for Domain Controllers you can just promote second machine to be a DC as well.
It will provision exact copy of the AD database and in case of failure you will be able to use second DC for authentication.
Also, you can add DNS and DHCP on both machines to provision redundancy for those services.

Recommended setup would be:
DC01 - existing - holds DNS and DHCP
DC02 - additional one - holds DNS and DHCP

If you place DNS on both servers there is no harm for environment as you can just point clients to one of those. I presume that DNS zone is AD-integrated as you are using AD as directory service which means that it will replicate itself with AD replication.

With DHCP you have to options if you want to provision redundancy:

Option 1
place DHCP server on both machines. COnfigure both machines with same scopes and options. Disable scopes on one of DHCP servers and in case of failure just enable scopes to assign IP addresses to clients (if first machine dies)

Option 2
place DHCP server on both machines. Configure both machines with same scopes and options. On first DHCP server exclude from scope first half of the IP addresses, on second machine exclude from scope second half of IP addresses. Keep scopes on both DHCP server active. It will provision simple semi load-balancing and redundancy as in case first box fails then second will be still operational, so you don;t need to swing service manually.

In regards to Active Directory there is one more thing to remember. FSMO roles. Once FSMO roles holder will die you have to seize roles to last live domain controller.
Also if you will have 2 domain controllers make both to be GLobal Catalogs (in case one will go down second will not need to rebuilt global catalog resources).

File Services Redundancy
For that purpose, as it seems to relatively small infrastructure, I would recommend domain based DFS and at least 2 replicas.
And also here you have two options. Once DFS will be in place to share files and you will define replicas, each folder shared on DFS (called link) will be pointing to 2 servers. You hav 2 options to use that:
you can have both replicas enabled for link which means that DFS will do semi-load balancing and will be using both servers to allow users to access files. In case one server will go down second replica will be able to handle file shares. Possible data loss of most recent files for users connected to server which goes down.
you can have one replica set to active and second set to passive. In case of failure you have to enable replica of files and it will take a while for users to re-connect to live machine. Also, data loss is possible as operational server goes down.
Exchange Redundancy
As you are on the way to get Windows 2008 and Exchange 2007 I would recommend to have a look into Standby Continuous Replication as this is available in Exchange 2007 Standard as a part of EBS you mentioned. So, before you will decide to implement Double Take or any other 3rd party software have a look into built-in features.
More detailes about SCR: 
As a summary I can say that currently my company operates on Exchange 2007 CCR (unfortunately not available in EBS) and this is very reliable solution which is exactly the same as Double Take with one exception - built-in into Exchange. Also we have DFS in few different geographical locations which works ok for us.
Hope that helps a little bit.
If you need more information le us know.
Mike KlineCommented:
SBS doesn't support clustering but you can add another domain controller to your network that would provide another AD box for you.  Generally you like to have a DC on a dedicated box (virtual or hardware) but I've seen small shops run AD and file services on the same box (they could not buy another box)
montekaneAuthor Commented:
I read about Exchange 2007 CCR, and it sounds like another viable solution, CCR runs on Datacenter right? If that's the case we could migrate to a Datacenter or Enterpise edition if that makes things easier.

Any recommendations for SQL failover?

That is correct. CCR works on Enterprise version, so if you have budget for that I would go for CCR. It works for us perfectly. We've tested it during production hours and Outlook dropped connection only for about 10-15 seconds.

For SQL server you have few options as well :)

If you don't have SAN or any kind of storage which can be connected to both servers same time then:
 - Database mirroring
 - Database Logshipping

Both technologies are based on transferring transaction logs from production server to standby and recovering changes on standby database.
In case of failure standby recent logs will be restored on database and database will become read-write production database.

Difference between those 2 is that database mirroring is done by SQL Server engine and Database logshipping is done by SQL Server Agent as scheduled jobs, so I think more reliable would be to go with Database mirroring.

Using one of those two solutions you can have production database for entering a data and standby database for reporting as this database remains read-only until failure happens.

If you have SAN then you can go with normal cluster solution as for cluster you have to share same storage space physically connected to both servers.
montekaneAuthor Commented:
Thanks for all your help.
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.