locked
SC2012CM (still can't get used to that...) RTM Site Design for Resilience/DR - Help Please RRS feed

  • Question

  • Hi All,

    I'm a little confused about some of the resilience/DR concepts regarding SCCM 2012 so I am hoping someone has already built this out.

    I have a design to create to cater for one live geographical site and one DR geographical site, the DR geographical site is online and accessible (20Gbps link) but there are no clients there until DR is invoked.

    As I see it I have two options:

    1. Create a central admin site and a primary site for both the live and DR sites.

    2. Span one primary site over the live and DR sites with an MP, DP etc... at each site.

    I would prefer the second option for a couple of reasons; the customer already has an RC2 primary site (POC) which has a lot of config that can be upgraded and re-used (I know the content can also be migrated), and we're also only talking about 1500 devices so I would like to avoid the complexity of 3 SCCM sites if poss. I also understand that the POC primary site (once upgraded to RTM) can't be attached to a new central site so wouldn’t be reusable.

    What I don’t understand fully in the second option of spanning a single SCCM site across both live and DR sites, is how the SQL would work. I read that SCCM doesn’t support SQL mirroring , a geocluster is off the cards, could I just set up a SQL mirror anyway and point the SMS provider at a DNS alias which can be updated in the event of DR? But then that raises the question of the SMS provider, can you still only have one? In which case if I lost the live site (assuming the provider is installed there) would I then have to manually install the SMS provider on the MP in the DR site.

    I hope someone can shed some light…

    Thanks

    Wednesday, May 23, 2012 10:12 AM

Answers

  • You can set up multiple SMS Providers.
    See also "Planning for High Availability with Configuration Manager": http://technet.microsoft.com/en-us/library/hh846246.aspx

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

    Wednesday, May 23, 2012 10:26 AM
  • A few things here.

    - There is no supported upgrade path from RC2 to RTM so that should not be a factor at all in your decision.

    - Adding two primary sites under a CAS does not in any way give you HA. It gives you some DR, but it is not a clean DR in any way as you must re-assign all of the clients to the new site which is not an automated task. In general, I would never recommened this approach for DR.

    - If you can't cluster SQL, then there's no other built-in method for HA of SQL. What other technologies are in use to provide HA/DR at the customer? Perhaps of those can solve te entire issue. What is the SLA for recovery? With proper backups, ConfigMgr doesn't take that must to restore as this is truly the primary method of DR designed into ConfigMgr.


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

    Wednesday, May 23, 2012 3:14 PM
  • A primary site will continue working even if the CAS is down.
    Clients will not be reassigned automatically once they are assigned to a site.

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

    Wednesday, June 20, 2012 10:32 AM

All replies

  • You can set up multiple SMS Providers.
    See also "Planning for High Availability with Configuration Manager": http://technet.microsoft.com/en-us/library/hh846246.aspx

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

    Wednesday, May 23, 2012 10:26 AM
  • A few things here.

    - There is no supported upgrade path from RC2 to RTM so that should not be a factor at all in your decision.

    - Adding two primary sites under a CAS does not in any way give you HA. It gives you some DR, but it is not a clean DR in any way as you must re-assign all of the clients to the new site which is not an automated task. In general, I would never recommened this approach for DR.

    - If you can't cluster SQL, then there's no other built-in method for HA of SQL. What other technologies are in use to provide HA/DR at the customer? Perhaps of those can solve te entire issue. What is the SLA for recovery? With proper backups, ConfigMgr doesn't take that must to restore as this is truly the primary method of DR designed into ConfigMgr.


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

    Wednesday, May 23, 2012 3:14 PM
  • What about just using recovery based on a backup to the DR site?  Robocopy the backup file over there each night and call it good.  If the main site blows, up, run recovery on the DR box.

    Microsoft MVP - ConfigMgr

    Wednesday, May 23, 2012 7:07 PM
  • A few things here.

    - There is no supported upgrade path from RC2 to RTM so that should not be a factor at all in your decision

    - Adding two primary sites under a CAS does not in any way give you HA. It gives you some DR, but it is not a clean DR in any way as you must re-assign all of the clients to the new site which is not an automated task. In general, I would never recommened this approach for DR.

    - If you can't cluster SQL, then there's no other built-in method for HA of SQL. What other technologies are in use to provide HA/DR at the customer? Perhaps of those can solve te entire issue. What is the SLA for recovery? With proper backups, ConfigMgr doesn't take that must to restore as this is truly the primary method of DR designed into ConfigMgr.


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

    There's really no supported upgrade? So the RTM setup gives you an upgrade option but that's not supported? :-S

    I understand that two primary sites won't offer HA, what it would offer is a hot standby site with it's own DB that we could replicate content two and stand clients up in if we needed to, but I think this is overkill anyway.

    Again I'm not really looking for HA, just multiple copies of data. I'm going to show my SQL inexperience, but would a mirror not allow us to replicate the DB to the DR site to be made primary in the event that the live site is lost. Or is a SQL mirror just flat out not supported for ConfigMgr irrespective of how it's accessed.

    I've also looked into DR with the underlying virtual platform, but nothing helpful going on there. I'll double check their expected SLA's my current understanding is that manual recovery should be OK and seems to be the best solution at the mo. I've recovered sites to the same server but are their any caveats to the restore, for example, would it be OK if the name/IP of the server changes?

    Thanks for your help!

    Thursday, May 24, 2012 9:37 AM
  • RC2 to RTM is not supported any you might run into issues (even it the "upgrade" seems to work fine). You would have to run scripts that were only made availble to TAP and CEP in production customers.
    Have you read the article I provided? It lists some options for high availability.
    IP can change in a recovery scenario, but the server name not.

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

    Thursday, May 24, 2012 10:29 AM
  • RC2 to RTM is not supported any you might run into issues (even it the "upgrade" seems to work fine). You would have to run scripts that were only made availble to TAP and CEP in production customers.
    Have you read the article I provided? It lists some options for high availability.
    IP can change in a recovery scenario, but the server name not.

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

    OK thanks, that's good to know. I assumed that the option there to be used.

    And no I haven't read the article but I'm about to.

    Thanks again for your help!

    Thursday, May 24, 2012 10:48 AM
  • Hi All,

    Thanks for the information above. I've just come back from holiday and the scenario has changed a little and I was hoping to run some of my thinking past you.

    The customer has a production and DR site, both of which now are hosting live services, so an active active configuration. Amongst other things ConfigMgr will be used to deliver App-V applications into VDI sessions at both sites.

    Based on the information provided and the additional reading I did, my design has the core ConfigMgr infrastructure at the primary site, with off box SQL, no SQL clustering (we have mirroring although I understand this is of little use) or anything clever or useful going on at the virtual platform level, I have now placed an MP and DP at the DR site for the clients that will be there. The DR process is to perform a site recovery using the setup media and a local backup at the DR site.

    The customers concern is that the VDI solution at DR will already be being utilised so in the event of DR there will be a period of time where SCCM is being recovered and new applications cannot be provisioned to VDI sessions. I was wondering whether the management point and DP will continue to provision applications at the DR site (for policies already configured), while the ConfigMgr infrastructure is recovered, I expect not. Or is this now (as I suspect) a good candidate for multiple primary sites.

    In the case of multiple primary sites, would the DR primary site continue to operate if the CAS and prod primary site (at the primary geographic site) were lost? Also I've read it might be possible to overlap the boundaries of the two primary ConfigMgr sites to provide a level of resilience, so for example if ConfigMgr sites in both datacentre's A and B both included the boundary for client site C, then clients at site C would be randomly assigned to either primary site, but in the event of failure would they reassign to the working site or just remain assigned to a non-working site? I am also making the assumption that all administration could be performed at the CAS making the configuration of the two primary sites identical.

    As you can probably tell I'm starting to confuse myself. I have the opportunity to bear this all out in testing but I was hoping someone has already done this.

    Thanks for your patients and your help!

      

    Wednesday, June 20, 2012 10:02 AM
  • A primary site will continue working even if the CAS is down.
    Clients will not be reassigned automatically once they are assigned to a site.

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

    Wednesday, June 20, 2012 10:32 AM