Exchange Design - trying for more High Availability
Current Configuration
We are running Exchange 2007, SP1.
Mailbox Server
We are running CCR for our mailbox server. The Active Node is located in our central office, and the Passive Node is located at our DR Site, which is a separate AD Site.
Hub Transport and CAS
We currently only have one CAS and Hub Transport Server, which is located at our central office. This is clearly a concern, as it represents a single point of failure.
Edge
We are not running any edge servers. We utilize third party solutions for our anti-spam solution.
My main objective at this point is to increase our availability in regards to the CAS and Hub Transport Servers. My initial thought was to just install one Hub Transport and CAS server at the DR Site, which is also home to the Passive Node of the CCR. I have read that it is good design to have two Hub Transport and CAS Servers for every AD Site that has an Exchange Mailbox. While this is certainly desirable, I am not sure it is within our budget. I figure I would be far better off adding an additional Hub Transport server at our DR Site, as again, we have only the one right now.
Also, once I get an additional Hub / CAS server setup - which would be located at our DR site - could I configure that Hub / CAS server to point to another SMTP gateway which would also be located at the DR site? The objective being if the central site were unavailable, then the DR site would allow mailbox access, and still function in regards to mail transport.
Thanks in Advance.
Martin
All Replies
Hi Martin,
You are on the right path, you will need to have a CAS & HUB server at the DR site otherwise you wont be able to access the mail or route mail. You will also need to consider in a true DR situation you will need to change your DNS to point to the DR CAS servers. As for HUB server being able to point to a different SMTP gateway that shouldn't be an issue.
Also having a second CAS and HUB server at each site would help you achieve a true High Availability but if not in your budget at this time then just make sure management understands the risks.
Thanks
Will##
Mailbox Server
We are running CCR for our mailbox server. The Active Node is located in our central office, and the Passive Node is located at our DR Site, which is a separate AD Site
##
It is an Exchange limitation that all nodes of an Exchange Server 2007 failover cluster have to be in the SAME AD site.
Hi,
I am not if it is limitation, bcoz so far did not implemented CCR in Live environment.
Please could you give more info on the Limitation?
Because that was the MS intention ..to keep active server in one site and passive server in other site ( for DR)
please could you give more info on it.
Thanks
Prabhakar kamuni
Martin,
As per the scenario you implemented good Mailbox server availabilty solution.
But how abt CAS/HUB, you should go for it in DR site.
So that would be a good solution.
Inform management about the risk involved in the Mail routing, if the current HUB/CAS go down
Cheers!!!
Prabhakar kamuni
http://technet.microsoft.com/en-us/library/bb125149(EXCHG.80).aspx
#
Geographically dispersed clusters also require that a single Active Directory site be stretched between the datacenters. However, only the clustered node needs to be in the site in the second datacenter. Third-party hardware and replication technology is required to deploy a geographically dispersed SCC solution.
#
For CCR this also has to do with the transport dumpster. After a lossy failover the mailbox role will ask all HT in the SAME AD site to resubmit messages.
Think what would happen if the two nodes in a CMS were in differnent AD sites. You take the active down for maitnence and all of a sudden your clients can't connect to their mail. Why? Because the Exchange Server switched sites. Exchange does have a thread in a service somewhere which checks to make sure that the site field for the server in the AD matches what the site for the physical machine is. How often does this value change? about never. So from what I can tell the thread must run about every fifteen minutes. So you'll have a time of at most fifteen minutes before anyone knows that the server is in the different site. Then once the AD is updated only computers in that site know that it has changed. But they've got their own caches, so lets say for arguments sake it takes them 15 min to pick up the change. Then an hour later when the AD replicates you'll begin to get mail to get redelivering to the databases. This is very not highly avaliable.
So how do you place the back up in a different site? Have the passive node be physically in a different site, but logically in the AD be in the same site. That way if the original goes down, everything will have failed over and you can manually change the site of the node, if it's worth it to change the site.
- Dear Martin ,
Lets first clear out differences between a Disaster Recovery Solution and a high Availbility Solution. What you require as per my understanding is the high availability solution.
High Availbility solution ensures availbility of services incase a SERVER goes down. whereas Disaster recovery is re-initiating of services in which the primary site is completly down and wil take a considerable amount of tym to be operational again.
High Availability Solution :
Mailbox server : CCR
HUB and CAS on same server : NLB
DR solution:
Mailbox Replication : SCR.
HUB and CAS : Install seperate servers for HUB and CAS.
regards
Ali- Proposed As Answer byS. Alizaib Alam Wednesday, October 28, 2009 5:28 AM
In a nutshell ...
Mailboxes:
- HA = CCR
- DR (resilience) = SCR.
Note: To achieve both (HA + DR) you can combine them (CCR + SCR).
Regarding transport/client acces:
- HA = Additional Hub/CAS in primary site. NLB CAS servers, Hub Transports do load balancing of their own.
- DR = Additional Hub/CAS in secondary/DR site- Proposed As Answer byJon-Alfred Smith Thursday, October 29, 2009 7:05 PM
Agree completely with Michel de Rooij. Just for clarity, Hub/CAS roles are on the same server. Hub Transports by default do load-balancing of outbound e-mail from Exchange, not of inbound. You will need to configure you third-party anti-spam solution to load-balance, for instance by assigning MX records to the Hub servers with equal values (MX preference = 10)
For scanners or applications that need to send mail though your Exchange organization (SharePoint, payroll etc), you should configure NLB for the receive connector.
Jon-Alfred Smith MCTS: Messaging | MCSE: S+M
