locked
High Availability and CM 2012 RRS feed

  • Question

  • Hi

    I am currently designing our new CM2012 installation. One of the demands I have on me is that the installation needs to be up 24/7. As in we have international offices that use our systems when our people are sleeping. So we need a way to ensure that our systems are operational at all times. We have two data centers and I have been looking into SAN mirroring but these methods only cover hardware or VM failure. If the servers for some reason, against all odds, blue screens the mirroring will not solve this problem.

    What is the usual way to ensure that all parts of CM is redundant? I would really love it if the solution to this did not involve increasing the complexity of administrating CM2012.

    Thanks

    Edit:

    I am currently looking into having all of the following available.

    1. SAN mirroring (This will enable us to recover is one of our data centers burn down with a identical copy, but not from a failed OS)

    2. SAN snap (This will let us recover from a failed OS. But with a dated copy of the failed server)

    3. TSM restore. This will enable us to recover any machine from the state it was at the last backup

    4. CM restore, this is the built inn restore option. This is cumbersome and time consuming and is by me considered a last resort.

    I was checking out having multiple sites with a CAS. But this increases complexity and recovery is not instantaneous. As you do have to assign clients to the new site, thus the user will notice the failure.

    All suggestions are welcome.

    • Edited by Nico_ Tuesday, July 24, 2012 8:13 AM
    Tuesday, July 24, 2012 8:00 AM

Answers

  • Planning for High Availability with Configuration Manager: http://technet.microsoft.com/en-us/library/hh846246.aspx

    Torsten Meringer | http://www.mssccmfaq.de

    Tuesday, July 24, 2012 8:09 AM
  • You can almost make sccm 2012 full redundant, there are som site systems that you can't make redundant.

    You can read about it here: http://technet.microsoft.com/en-us/library/hh846246.aspx



    Tuesday, July 24, 2012 8:16 AM
  • Place it on a clustered hypervisor that handles the failover if one node goes down.

    Torsten Meringer | http://www.mssccmfaq.de

    Tuesday, July 24, 2012 8:23 AM
  • One of the demands I have on me is that the installation needs to be up 24/7.

    Not discounting this requirement at all, but I would push back. ConfigMgr is not Exchange, it's not all or nothing. HA is an excercise of defining available functionality vs. cost of that functionality not being available and the acceptable recovery times. Thus, you need to look at the various roles and the functions they provide in ConfigMgr. A primary site server, hosting no other roles in ConfigMgr 2012, does not prevent *any* operational activity. Clients will still be able to receive software and software updates, OSD will still function, reproting will still be available, etc.

    Thus, trying to make the primary site server itself redundant is an excercise that provides no benefit except some extra money for your hardware vendor. What you really need to do is define each piece of functionality that is required to be highly available and then design the roles for that accordingly.

    Politically, this is not necessarily an easy task because those that are dictating this requirment probably don't care (they probably barely know what HA really means), but if you can break it down into a simple table with dollars/costs associated by function then let them make the decision with the fact thrown in that there is no built in way to make the application as a whole HA (except using Torsten's suggestion below which gets even more expensive).

    A couple of other notes

    CM restore, this is the built inn restore option. This is cumbersome and time consuming and is by me considered a last resort.

    Not at all. This is the standard method used by many and is actually fairly quick (depending upon your DB size) and painless -- or at least straight-forward. If you've properly separated your roles for HA, then this is the prefered choice.

    thus the user will notice the failure.

    How? User's will never notice anything if your entire ConfigMgr hierarchy goes away. I am by no means advocating using multiple primaries for a psuedo-redundancy solution (far from it because I think it is a terrible solution), but a total ConfigMgr failure only minimally impacts the users and that's only because they won't be able to receive new software.


    Jason | http://blog.configmgrftw.com | Twitter @JasonSandys

    Tuesday, July 24, 2012 1:27 PM

All replies

  • Planning for High Availability with Configuration Manager: http://technet.microsoft.com/en-us/library/hh846246.aspx

    Torsten Meringer | http://www.mssccmfaq.de

    Tuesday, July 24, 2012 8:09 AM
  • You can almost make sccm 2012 full redundant, there are som site systems that you can't make redundant.

    You can read about it here: http://technet.microsoft.com/en-us/library/hh846246.aspx



    Tuesday, July 24, 2012 8:16 AM
  • I have read the technet article. As far as I can see there is no good way of ensuring that the site server is redundant. If the site server goes down there is no way of instantaneously recovering from that?
    Tuesday, July 24, 2012 8:18 AM
  • Place it on a clustered hypervisor that handles the failover if one node goes down.

    Torsten Meringer | http://www.mssccmfaq.de

    Tuesday, July 24, 2012 8:23 AM
  • One of the demands I have on me is that the installation needs to be up 24/7.

    Not discounting this requirement at all, but I would push back. ConfigMgr is not Exchange, it's not all or nothing. HA is an excercise of defining available functionality vs. cost of that functionality not being available and the acceptable recovery times. Thus, you need to look at the various roles and the functions they provide in ConfigMgr. A primary site server, hosting no other roles in ConfigMgr 2012, does not prevent *any* operational activity. Clients will still be able to receive software and software updates, OSD will still function, reproting will still be available, etc.

    Thus, trying to make the primary site server itself redundant is an excercise that provides no benefit except some extra money for your hardware vendor. What you really need to do is define each piece of functionality that is required to be highly available and then design the roles for that accordingly.

    Politically, this is not necessarily an easy task because those that are dictating this requirment probably don't care (they probably barely know what HA really means), but if you can break it down into a simple table with dollars/costs associated by function then let them make the decision with the fact thrown in that there is no built in way to make the application as a whole HA (except using Torsten's suggestion below which gets even more expensive).

    A couple of other notes

    CM restore, this is the built inn restore option. This is cumbersome and time consuming and is by me considered a last resort.

    Not at all. This is the standard method used by many and is actually fairly quick (depending upon your DB size) and painless -- or at least straight-forward. If you've properly separated your roles for HA, then this is the prefered choice.

    thus the user will notice the failure.

    How? User's will never notice anything if your entire ConfigMgr hierarchy goes away. I am by no means advocating using multiple primaries for a psuedo-redundancy solution (far from it because I think it is a terrible solution), but a total ConfigMgr failure only minimally impacts the users and that's only because they won't be able to receive new software.


    Jason | http://blog.configmgrftw.com | Twitter @JasonSandys

    Tuesday, July 24, 2012 1:27 PM