Skip to Content

SAP MSCS High Availability Problem

Dear Experts,

I have some doubts regarding the actual function of SAP HA.

As my understanding:

SAP logon group is used for load balancing, if user1 logon to server3 (DI), and server3 failed at some time, then, user1 will be hanged at SAP system, all his/her data will be lost and he/she must re-logon SAP system.

But for SAP MSCS HA, if currently node1 server provide service, then, user2 will be logon to node1 server, if that server failed, then, cluster will automatically migrate service/data from node1 server to node2 server, user2 will keep operating on the system without any notice about the server changes. am my understanding right?

But currently, we met some trouble on MSCS setup, the cluster on OS level and database(MSSQL) and SAP were all completed successfully, then we go on to the stage to verify whether the function of HA really functions well. Firstly, we move all services/resource to node1, and user1 logon to SAP  system, in the bottom-right corner of SAPGUI, it displays SAPNode1, then, we move all service/resources from node1 to node2 by manual in Microsoft failover cluster manager (windows 2008R2), and then, we thought that if we execute a new tcode in the SAPGUI, the bottom-right corner will be displayed as SAPNode2 instead of SAPNode1, and all customer data will be remained, customer will not be interrupted by this server change and he/she will not be noticed any changes of the server, everything is smoothly changed, but actually, if we type a new tcode and execute it, an error message box displayed that notify us the server is un-available, so we need to re-logon SAP system, this time the bottom-right corner is displayed as SAPNode2.

May I know whether this situation is normal? this is so-called HA?

Besides, could you tell me the ways you use to verify whether your HA functions well? direclty stop some SAP/DB services?

Thank you very much for all.

Best regards,

Freshman

Add comment
10|10000 characters needed characters exceeded

  • Get RSS Feed

4 Answers

  • Aug 09, 2012 at 04:42 PM

    Dear all,

    Could anyone help me? really thank you and will sure to award.

    Add comment
    10|10000 characters needed characters exceeded

    • Former Member

      Hi,

      Yes, you can directly stop the services and monitor if the services fails over successfully to another node. One more thing is to verify if the lock table entries get copied. You can also check the log files for any error messages during the process.

  • avatar image
    Former Member
    Aug 10, 2012 at 06:07 PM

    Hello BA

    HA means to cluster Single point of failures i.e Message service, enqueue service(ASCS) and database services. CI and app server can be installed anywhere.

    Now suppose we have 2 nodes Node 1 and Node 2 and we have clustered our SAP and database services on these 2 nodes. Whenever Node1 fails all the services will failover to Node 2 and vice versa.Remember ASCS and database are addressed by virtual names and not physical host names. So whenever nodes failover , ASCS and database services will still be addressed with same virtual host names. Failover takes few seconds and user would get impacted but only for a very short span of time.The issue might come if user is connected to an instance which is installed on Node1 or Node2  and Node1 or Node 2 goes down , in that case user would have to login again as I understand(Use logon group in this case).But if user is connected to the instance which is not on either of these nodes there should be no impact.

    Also we have to install enqueue replication server on both nodes.So suppose if ASCS is active on Node1, ERS will be active on Node2 and vice versa. ERS actually contains a replication of lock table. If enqueue server fails on Node1 , then enqueue server starts on node 2 and will retrive the data from replication table on node 2 and will write in its lock table and hence would make sure that nothing gets lost in which user was working

    Also to check ur HA, SAP in its guide has suggested couple of tests, please go through them

    Also remember we are not clustering our CI or DI, we are only clustering Message,Database,enqueue and database services

    Hope it helps

    Thanks

    Add comment
    10|10000 characters needed characters exceeded

  • avatar image
    Former Member
    Aug 11, 2012 at 02:22 AM

    This has been answered quite a few times ..on couple of occassions by me..

    Application server (where user actually logs in) in not normally clustered and isn't high available. Also the concept is high available and not ALWAYS available.

    Regards.

    Ruchit.

    Add comment
    10|10000 characters needed characters exceeded

  • Aug 14, 2012 at 10:48 PM

    Simple answer for this question is as below..

    When dialog instance fail it will not put you to other system because you basically tie user to one of the dialog on specific server.

    Once dialog instance is down you need to log in again and logon group will put you to the available dialog instance.

    Cheers,

    Yogesh

    Add comment
    10|10000 characters needed characters exceeded