locked
2 member DAG RRS feed

  • Question

  • Greetings all,

    I am planning to create a 2 member DAG. Please tel me what things should i consider before i create the DAG. I first installed exch2003 on my test network and then migrated everything to a exchnage 2010. Now I have added one more Exchange 2010 and want to achive HA for my site. All these servers are on the same domain. I also have a server for FSW, witness server. So i think I am all set to try my hands on DAG. I know many articles are available on this but I prefer to ask u all as u have tried and tested it all.

    Now, let me tel u what i know abt DAG and how to configure it.                                                                            Firstly, both the 2010 servers should be identical, which I have. Call them EXCH01 and EXCHDR.                             Secondly, do I have to enable the failover clustering feature or will it get enabled once DAG is created?               Third, does DAG support my setup? EXCH01, EXCHDR and FSW. Total of 3 members in the DAG(but only two have exchange in it). Also both servers have MB, CAS and HUB roles in it.

    I have read a lot that DAG doesnt work for 2Mailbox setup? Is it true? Also If i in case add the Failover clustering feature, do i have to check it by opening Failover Cluster Manager? Chek wat and how???

    Guys I am straight from university and this my first big project, A rookie an all the ways. Plzzzzz help me out.

    Thanks all, in advance

    Wednesday, June 9, 2010 2:40 PM

Answers

  • There shouldn't be any difference between shutting down the server and turning off the cluster service. Define "whole Exchange is down". Somebody last week had the issue of the fact that Remote PowerShell wasn't working on the secondary server.

    If you open an instance of the Exchange Management Shell on your secondary server, what server does it show it connecting to?

    What is the value of the RpcClientAccessServer for the databases, both before and after the failover? This is the endpoint that clients will try and connect to. If there's no network path to that endpoint, there will be no connections.

    • Proposed as answer by Xiu Zhang Thursday, June 10, 2010 7:05 AM
    • Marked as answer by Xiu Zhang Tuesday, June 15, 2010 3:24 AM
    Wednesday, June 9, 2010 4:47 PM

All replies

  • The failover clustering component will be added when you add the mailbox server to the DAG.  Or you can install it ahead of time if you like.

     

    The DAG does support a 2-node setup like yours, but be aware that you should still load-balance the CAS role and for this you cannot use Windows NLB when the CAS role coexists with the mailbox role in a DAG.  Therefore, you will need a hardware load balancer for true high availability in your design.

     


    Neil Hobson, Exchange MVP
    • Proposed as answer by Xiu Zhang Thursday, June 10, 2010 7:05 AM
    Wednesday, June 9, 2010 2:55 PM
  • Thanks so much for the insight Neil.

    I just created a dag DAG1 and also added the 2 MB servers to it. Now wen i open Failover cluster manager> Validate the cluster. It says the configuration supports the cluster. Good news. Also the Cluster Quorom settings is set to Node and File Sharing Majority. Hope thats correct??? But I still see an error in my DAG1.test.org. it has a small red dot and wen i check the error it says----Cluster network name resource 'cluster name' failed registration of one or more associated DNS name for the following reason: this operation returned because the timeout period expired. Can u plz guide me how i fix tis error.

    Also I have one more question. The 2 servers are EXCH01 and EXCHDR. Now all my test clients are showing exchange account as EXCH01. Thats all rite. Nowwww, if i switchover the DBs will this account change and show EXCHDR?? Coz in my case its not doing it? and if i bring down my EXCH01(complete shut down) then none of the clients can connect to outlook. But if i bring down EXCHDR (assuming that i have active DB on it), then my DB on EXCH01 gets active and users can connect. Can u tell me why is it so? Why doesnt my DR server work wen my primary server is down??? Is ter sumthin wrong in my DAG setup. Plz helpppppppp.

    Wednesday, June 9, 2010 3:17 PM
  • Sharath,

     Error about DNS may be because cluster is trying to register network name in DNS and it is failing. Regarding the second question There will not be any change on the server attribute on user's if the server is changed. If you shutdown the server then Database will not failover to the DR server. You have to create a failure on Exch01 so that the database is automatically failed over to ExchDR. If you are using Outlook 2010 as a client then client will automatically connect to Exchdr if not you have to close and reopen Outlook.

    Thanks

    Bharani

    • Proposed as answer by Xiu Zhang Thursday, June 10, 2010 7:05 AM
    Wednesday, June 9, 2010 3:40 PM
  • Thanks Bharani for the reply.

    We are not using outlook 2010 in our org. I cud not get wat this means -

    There will not be any change on the server attribute on user's if the server is changed. If you shutdown the server then Database will not failover to the DR server. You have to create a failure on Exch01 so that the database is automatically failed over to ExchDR.

    I thot if u have 2 servers in DAG. the other should take over if one goes down. I want to know how HA can be achived if my EXCH01(primary) goes down for sum hardware reasons. Then can my uers still be connected?? Will EXCHDR not pick it up from ter??? It is happening the oter way around though.... And if in a scenario I have to bring down my primary for maintainance how wil my users still be connected?? Wat is the use of switchover and failover options if my DR doesnt work wen primay is down. Plz reply

    Wednesday, June 9, 2010 3:49 PM
  • Also plz let me know if the DNS error in my cluster manager is a serious error or not? Wil this result in my DAG not functioning? If its fatal then plz help me fix this error? Coz i have tried all means without any solution.
    Wednesday, June 9, 2010 3:53 PM
  • Is the netname of the cluster online? If so the DNS error might not be that bad. Can you (from a computer that's not in the Dag) ping the Dag name? If you can't then it's a problem.

    To test failing over of the databases you can stop the cluster service on the Dag member that has the mounted databases.

    Wednesday, June 9, 2010 4:04 PM
  • I can ping DAG1 just fine from other systems. Tried ur suggestion for testing failover, worked fine. Thanks for that.

    But, I want to know wat if my primary serevr is shut down for any reasons, wil this switchover happen automatically? and even if it does happen wil the clients still be able to connect??? Coz i have tried to completely put the primary out of order(by shutting down , or removing from LAN network) and then my whole exchange is down. I need to know wat happens in tis scenario????

    Plz reply all. Thanks so much for the help so far. Its been very usefull.

     

    Wednesday, June 9, 2010 4:16 PM
  • There shouldn't be any difference between shutting down the server and turning off the cluster service. Define "whole Exchange is down". Somebody last week had the issue of the fact that Remote PowerShell wasn't working on the secondary server.

    If you open an instance of the Exchange Management Shell on your secondary server, what server does it show it connecting to?

    What is the value of the RpcClientAccessServer for the databases, both before and after the failover? This is the endpoint that clients will try and connect to. If there's no network path to that endpoint, there will be no connections.

    • Proposed as answer by Xiu Zhang Thursday, June 10, 2010 7:05 AM
    • Marked as answer by Xiu Zhang Tuesday, June 15, 2010 3:24 AM
    Wednesday, June 9, 2010 4:47 PM
  • What I mean is that when i turn the clustering service OFF on my EXCH01(primary server), then the DB on the DR server becomes active. But if i shut EXCH01 down(power off), then although the DB on DR server becomes active, the users cannot connect to outlook. It says- Disconnecetd, connection to exchange server lost. Now, u r saying that in both the cases the DR shud take over. Rite? But here its not. Can u plz explain y?

    The EMS is working fine on the DR server. It says connected to EXCHDR even if the primary is shut down. No problems with the EMC also. But clients cannot connect and thats my only concern. And i have not created a CAS array since we cannot have both array and DAG on these sytems. So will the DAG not function witout the CAS array.

    Also how do i chek the RpcClientAccessServer for the databases? Plz guide...

    Thursday, June 10, 2010 8:26 AM
  • Get-MailboxDatabase | ft name,RpcClientAccessServer.

    If the client access server is the server you just shut down, that would be why clients can't connect.

    There's an idea that's been thrown around on other threads of this forum. Create a ClientAccessArray and give it the name of the Dag. That way (if all of the Dag members are Cas's as well) the clients will connect to the cluster group owner. You don't get load balancing (which you may not need) but you will get high availability.

    Thursday, June 10, 2010 2:24 PM
  • Yes u wer rite. Tried the cmd Get-MailboxDatabase | ft name,RpcClientAccessServer  and it gave me EXCH01, my primary server. So, if my Primary is shut down, my clents cannot connect. Thanks a lot for this help.

    Now tel me how do i fix this thing? Can this nly be fixed my a client access Array? Or is ter any other way??? How can I point my clients to the DR CAS server wen my Primary CAS server is shut down?

    Thursday, June 10, 2010 4:19 PM
  • You can manually change the RpcClientAccessServer with the Set-MailboxDatabase cmdlet.

    To create a Cas Array you use the New-ClientAccessArray cmdlet.

    Thursday, June 10, 2010 5:07 PM
  • So every time my primary goes down i have to manually point my clients to the DR server using Set-MailboxDatabase  cmd? So u mean it wont happen automatically? That is sad....

    I tried to create CAS array, but i am unable to. I read sumwer that u cant create CAS array on the servers which are members og DAG? Then how do i create CAS array and give it the same name as the DAG? The solution which u had suggested above

    There's an idea that's been thrown around on other threads of this forum. Create a ClientAccessArray and give it the name of the Dag. That way (if all of the Dag members are Cas's as well) the clients will connect to the cluster group owner. You don't get load balancing (which you may not need) but you will get high availability.

    Plz reply soonnnnnnnnnnnn.

    Friday, June 11, 2010 8:31 AM
  • Jader,

    I tried the above steps. I shut down the Primary server EXCH01 and the DB on secondary server EXCHDR mounted fine. Then i used the Set-database DB1 -RpcClientaccessserver EXCHDR cmd, which changed my CAS server for DB1. But my clinets still cannot connect through outlook. It says disconnected. Any reasons why???

    Friday, June 11, 2010 8:49 AM
  • Hello Sharath,

    What is the version of Outlook client you are using? If it is Outlook 2007 then do the following.

     

    1) Right Click on Outlook icon at the right end of the desktop by holding ctrl.

    2) This should give you an option to test email configuration.

    3) See which server outlook is talking to.

     

    Thanks

    Bharani

    Friday, June 11, 2010 12:30 PM
  • Windows NLB and Windows Failover Clusters can't share nodes. There's nothing blocking you from creating a Cas Array. It's just an FQDN, it's what you put behind the Cas Array that matters.
    Friday, June 11, 2010 2:28 PM