asked on

Question about Exchange 2013 Failover

Hi guys.

I will be carrying out the first Failover testing since implementation of DAG and HA for our Exchange 2013 CU1 Server. I plan on following the steps from this website:

http://msexchangeguru.com/2013/01/17/e2013-dag/

In this documentation, he suggested we kill the Microsoft.Exchange.Store.Worker process then check the mailbox status to see if the failover was successful. That is all well and good. What he left out is:

1. What happens when I start the Microsoft.Exchange.Store.Worker service back on on CONTOSO1? Will it revert back to the CONTOSO1 database, or do I have to manually perform a switchover using the "Activate" feature on Exchange Control Panel?
2. In his example, the "ContentIndexState" goes to a "Failed" state after performing the failover. Does this fixes itself? Do I have to wait for ContentIndexState to be back to "Healthy" state before manually performing a switchover?

For example purposes, my first server and Active database is CONTOSO1, second server and passive database is CONTOSO2, Database name is CONTOSODB1, DAG name is CONTOSODAG1 and Witness server is CONTOSOWITNESS1. Both CONTOSO1 and CONTOSO2 are CAS and MB.

Thanks,

Rod

ASKER CERTIFIED SOLUTION

Simon Butler (Sembee)

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

Rodrigo Costa dos Prazeres

ASKER

Great answer as usual, thanks Simon.

One other question, and I am happy to open another thread if it may be. I've inherited this config and first time working with DAG and HA.

Can you confirm that the namespace (let's call it mail.contoso.com) should be resolving to to the Cluster IP address, when pinging internally?

The setup has no hardware load balancer, and the way the system was setup was with Failover Cluster Manager (Cluster name CONTOSODAG01).

I've performed the failover test just before and realized Outlook lost connectivity when I switched the first server off. The moment I changed the DNS record so mail.contoso.com points to the cluster IP it reconnected fine and I could see it was connected to the second server.

Can you confirm this is a supported scenario? And that once I implement CU5 (it is currently with CU1) this will not stop working because of some changes that were introduced?

Thanks

SOLUTION

Simon Butler (Sembee)

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

Rodrigo Costa dos Prazeres

ASKER

Is there a compatible software based load balancer I could use? I read that I can't use NLB. Both servers sits on a data centre and adding a hardware load balancer is impossible.

M A

Please check this. You can have a KEMP virtual appliance for load balancing. You need a VM for this.
http://kemptechnologies.com/latam/reviews/ms-exchange-2013-useful-blog-post-about-load-balancing-2013-vs-2010

Simon Butler (Sembee)

Almost all of the load balancer vendors have virtual appliance versions of their products. Kemp and Jetnexus do, as I am running them both at home.

Disclaimer - both Kemp and Jetnexus have given me NFR copies of their products to use.

Simon.