bikerhong
asked on
Exchange 2010 DAG failover is not seamless?
Hi
We have a 2 server DAG (all roles on each) on separate subnets.
When a database fails over to the other server, the client's Outlook comes up with a message saying "The Exchange Administrator has made a change which requires you to restart Outlook" - if we do this everything works.
However I know we have had this running before (elsewhere maybe) where it is seamless - the Outlook client might freeze up for 30s but then carries on as normal.
Have we missed a setting somewhere?
We have a 2 server DAG (all roles on each) on separate subnets.
When a database fails over to the other server, the client's Outlook comes up with a message saying "The Exchange Administrator has made a change which requires you to restart Outlook" - if we do this everything works.
However I know we have had this running before (elsewhere maybe) where it is seamless - the Outlook client might freeze up for 30s but then carries on as normal.
Have we missed a setting somewhere?
Do you have an RPC CAS Array? Remember the DAG only protects the database, it doesn't protect the clients. As you are going across subnets, are you also crossing AD sites?
Simon.
Simon.
ASKER
Hi
After posting this question I did some checking and there was 1 server which had a rollup update applied without the other server being updated.
I fully patched both servers last night.
Upon testing, the same thing happens.
Simon, we do not have an RPC CAS Array. Its simply 2 Exchange Servers with all roles, each on a different subnet/AD site configured in a DAG.
No hardware balancers, etc.
After posting this question I did some checking and there was 1 server which had a rollup update applied without the other server being updated.
I fully patched both servers last night.
Upon testing, the same thing happens.
Simon, we do not have an RPC CAS Array. Its simply 2 Exchange Servers with all roles, each on a different subnet/AD site configured in a DAG.
No hardware balancers, etc.
You need an RPC CAS Array then.
You have no client failover. If the clients connect to server 1 for CAS and it fails over, they will not follow reliably. You are then dependant on Autodiscover.
You don't need a load balancer (although one will help), just have a DNS record with a very short TTL. Do failover, change the DNS.
Simon.
You have no client failover. If the clients connect to server 1 for CAS and it fails over, they will not follow reliably. You are then dependant on Autodiscover.
You don't need a load balancer (although one will help), just have a DNS record with a very short TTL. Do failover, change the DNS.
Simon.
ASKER
Hi
I think we are going for the DNS option, but just want to clarify something.
We have 4 host names in question:
DAG.domain.com - DAG IP
mail.domain.com - used for mail services, currently resolves to same IP as primary Exch Svr
primary.domain.com - Primary ExchSvr located at HQ
secondary.domain.com - Secondary ExchSvr located at DR site
When setting up a new Outlook profile, if you put in DAG.domain.com as the Exchange Server, after a few seconds it changes automatically to primary.domain.com as that is where the mailboxes are.
However for obvious reasons we do not want to change the DNS of the actual Exchange Server, rather, we want to change the IP of mail.domain.com.
Is there a way to get clients to always look at mail.domain.com as the Exchange Server, without automatically resolving the the actual hostname of the mailbox server?
I think we are going for the DNS option, but just want to clarify something.
We have 4 host names in question:
DAG.domain.com - DAG IP
mail.domain.com - used for mail services, currently resolves to same IP as primary Exch Svr
primary.domain.com - Primary ExchSvr located at HQ
secondary.domain.com - Secondary ExchSvr located at DR site
When setting up a new Outlook profile, if you put in DAG.domain.com as the Exchange Server, after a few seconds it changes automatically to primary.domain.com as that is where the mailboxes are.
However for obvious reasons we do not want to change the DNS of the actual Exchange Server, rather, we want to change the IP of mail.domain.com.
Is there a way to get clients to always look at mail.domain.com as the Exchange Server, without automatically resolving the the actual hostname of the mailbox server?
You need an RPC CAS Array as I have already said.
That is effectively a virtual Exchange server. It is exists in DNS only. You can then move it around as required.
It is a host name that exists internally only. It doesn't appear on SSL certificates and must not resolve externally at all.
Simon.
That is effectively a virtual Exchange server. It is exists in DNS only. You can then move it around as required.
It is a host name that exists internally only. It doesn't appear on SSL certificates and must not resolve externally at all.
Simon.
ASKER
Can RPC CAS Array's span subnets?
ASKER
http://www.tino.nl/index.php/2010/06/11/answers-to-the-10-most-common-questions-on-the-exchange-cas-array/
Q4. Can I stretch a CAS Array over multiple IP-Subnets?
A4. Depends. There can be only one CAS Array per AD-Site. So if both IP Subnets are in a different AD-Site, you cannot.
Seems I cannot as each site is in a different subnet, with each site as a seperate AD site.
What would be your recommendation in this instance?
Q4. Can I stretch a CAS Array over multiple IP-Subnets?
A4. Depends. There can be only one CAS Array per AD-Site. So if both IP Subnets are in a different AD-Site, you cannot.
Seems I cannot as each site is in a different subnet, with each site as a seperate AD site.
What would be your recommendation in this instance?
CAS Array is AD site specific.
However if you are failing over the CAS Array to another site then that is fine. It is just a DNS entry. What you cannot do is have that configuration in place all the time.
So you could have a CAS array if
outlook.site1.example.loca
outlook.site2.example.loca
In the event of site 1 going to site 2, just move the CAS Array to point to a server in site 2. Don't change anything else and it will work fine.
Simon.
However if you are failing over the CAS Array to another site then that is fine. It is just a DNS entry. What you cannot do is have that configuration in place all the time.
So you could have a CAS array if
outlook.site1.example.loca
outlook.site2.example.loca
In the event of site 1 going to site 2, just move the CAS Array to point to a server in site 2. Don't change anything else and it will work fine.
Simon.
ASKER
Thanks Simon I really appreciate this - will look into configuring this up in the next couple days.
Am I right in saying that in future then all new Outlook profile setups should be pointed at the cas address instead of the dag? What about all the current pre-existing clients who are already configured?
Am I right in saying that in future then all new Outlook profile setups should be pointed at the cas address instead of the dag? What about all the current pre-existing clients who are already configured?
New clients will get the CAS Array address from Autodiscover.
Existing clients will have to be updated by hand. That means doing a repair of the profile in Outlook. Annoying I know - I usually tell the helpdesk of clients to do it if they have to touch a workstation for any other reason. That usually gets about 80% of the clients pretty quickly.
Simon.
Existing clients will have to be updated by hand. That means doing a repair of the profile in Outlook. Annoying I know - I usually tell the helpdesk of clients to do it if they have to touch a workstation for any other reason. That usually gets about 80% of the clients pretty quickly.
Simon.
ASKER
Thanks. I just looked up some guides to creating these and I'm a little confused about the CAS object!
First link:
http://blogs.technet.com/b/ucedsg/archive/2009/12/06/how-to-setup-an-exchange-2010-cas-array-to-load-balance-mapi.aspx
Sounds very simple really, 1 shell command to create the CAS Array object, and set the DNS to point to the right IP.
However a second link:
http://exchangeserverpro.com/how-to-install-an-exchange-server-2010-client-access-server-array/
This stipulates the CAS Array cannot sit on servers with DAG, because of NLB. We only have 2 servers, each with all roles installed.
So can I or can I not create this CAS Array? My confusion comes from the point that the CAS Array is not an actual server, essentially just a DNS entry. But then the NLB + DAG incompatibility throws me....!
First link:
http://blogs.technet.com/b/ucedsg/archive/2009/12/06/how-to-setup-an-exchange-2010-cas-array-to-load-balance-mapi.aspx
Sounds very simple really, 1 shell command to create the CAS Array object, and set the DNS to point to the right IP.
However a second link:
http://exchangeserverpro.com/how-to-install-an-exchange-server-2010-client-access-server-array/
This stipulates the CAS Array cannot sit on servers with DAG, because of NLB. We only have 2 servers, each with all roles installed.
So can I or can I not create this CAS Array? My confusion comes from the point that the CAS Array is not an actual server, essentially just a DNS entry. But then the NLB + DAG incompatibility throws me....!
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
http://social.technet.microsoft.com/Forums/exchange/en-US/98990dbb-82e6-4047-880a-a6919ae8904e/dag-failover-behavior-for-outlook-clients
And you are correct by the End User Experience, Outlook users should not be prompted to Restart Outlook. CAS should redirect to the online DAG member.
Hope this helps