none
Load balancing replication connections across multiple DC's at hub site

    Question

  • As a background, I have 4 DC's at my main hub datacenter, 2 are 2003 and 2 are 2008R2. I am transitioning from the 2003 to the 2008 boxes and now wish to decommission them. My site links are set up correctly, my replication connections are "automatically generated", I have no preferred bridgeheads. Prior to yesterday, my 2003 DC's had ALL the replication connections on them, so I thought the easiest way to transition these connections gracefully off of the 2003 boxes would be to set my two 2008 DC's as preferred bridgeheads for IP only. It was my understanding that when the KCC reanalyzed the replication topology, it would distribute replication connections ACROSS both bridgehead DC's. I left it like that overnight, and when I come in here this morning, one of the 2008 DC's has ALL the connections, which were previously distributed evenly across the 2003 DC's. I do not like this at all, and I wish to have them distributed evenly again. Everything is working, but I do not like this arrangement at all. I know if one of the DC's goes down, the replication connection will move the next time the KCC runs, but I do not like that, and I also do not want to manage manual connections either.

    I know there is a command line tool I can use to move these automatically generated site links to another DC, as I did this before years ago, but did not document it. I have removed the preferred bridgehead setting on the 2008 DC's at my hub site. Could someone let me know what that command line utility and string would be to move these replication connections?

    Thanks,

    Wednesday, March 16, 2011 12:02 PM

Answers

  • So I went as far as to delete a number of the replication connections from my hub site and a number of my remote sites, then ran repadmin /kcc site:%hubsite% and repadmin /kcc site:%remotesite%

    Miraculously my connections are now balanced across my 3 DC's at my hub site (I demoted one of the 2003 DC's and now have 2 2008 and 1 2003). Running repadmin /replsum shows no errors and balanced connections across my 3 hub dc's versus a ridiculous number on one and hardly any on the others as before.


    Which begs the question. If the KCC is supposed to run every 15 minutes by default and it's supposed to "load balance the connections" automatically, why did it not do that in my environment?

    Thursday, March 24, 2011 12:06 PM

All replies

  • You can delete all the connection object & run repadmin / kcc * it will generate the connection object automatically on all the sites.

     

    Regards


    Awinish Vishwakarma| MY Blog

    Disclaimer : This posting is provided AS-IS with no warranties or guarantees and confers no rights.

    Wednesday, March 16, 2011 12:07 PM
    Moderator
  • I understand that, but who is to say it wont do the same thing it did before (move everything to 1 dc). The KCC told these replication links to move to 1 dc.
    Wednesday, March 16, 2011 1:26 PM
  • KCC uses site, site link info to create the connection object, so would suggest, re look to the sites, site link & DNS configured on DC & client machine to their site. Take a look at KCC architecture.

    http://technet.microsoft.com/en-us/library/cc755994%28WS.10%29.aspx#w2k3tr_repto_how_uvgz  

    The reason i can think of, it generate all the connection object to single DC is the DC might be the offline/down when KCC(KCC run after every 15 mins), so it generated the connection according to the DC's available.

    Take a look, if you want more.

    http://blogs.technet.com/b/askds/archive/2011/02/04/kcc-offline-bridgehead-behaviors.aspx

    A good discussion on below link.

    http://social.technet.microsoft.com/Forums/en-US/winserverDS/thread/0d10914f-c44c-425a-8344-3dfbac3ed955/

    http://technet.microsoft.com/en-gb/library/cc755994%28WS.10%29.aspx#w2k3tr_repto_how_ludi

     

    Regards


    Awinish Vishwakarma| MY Blog

    Disclaimer : This posting is provided AS-IS with no warranties or guarantees and confers no rights.

    Wednesday, March 16, 2011 1:45 PM
    Moderator
  • "The reason i can think of, it generate all the connection object to single DC is the DC might be the offline/down when KCC(KCC run after every 15 mins), so it generated the connection according to the DC's available."

    Absolutely not...

    All DC's were online during this time, these 4 dc's are in the same AD sites. I understand fully how sites and site links work, and our environment is set up optimally. I have had an ADRAP performed by MS every year for the last 4 years and nothing has changed in the environment except recently forest prep for 2008R2 AD and adding these new 2008 DC's. Furthemore, all DC's in the forest point their secondary DNS server to one of my DC's here at my Hub site, the one which consequently did not receive any connection objects.

     

    Wednesday, March 16, 2011 3:48 PM
  • Any other ideas? I dont want to have to call M$ about this. I wish I would have documented what we did in 2007 when this came up...

    I think I could set the other server as preferred bridgehead only, then delete the automatically generated link one at a time and then run repadmin /kcc from that specific site, that should re-generate that replication link, but I dont want to have to do this 60 something times. The commands we did back in '07 was from one box.

    Thursday, March 17, 2011 7:52 PM
  • Hi-

    So these aren't going to move unless something causes it. 2008 R2 added the load balancing of connections you want for RWDCs. Personally I'd just dcpromo down the 2003 boxes and let AD do its job.


    My Book - Active Directory, 4th Edition
    My Blog - www.briandesmond.com
    Friday, March 18, 2011 3:46 AM
  • I can't think of any other reason which triggered changes if its not the case of offline dc as you can see from KCC architecture overview. I would suggest not to worry much about this, KCC does its job fairly well, so you can removed DC's of 2003 & go ahead with 2008 R2. I have seen so many implementation of windows 2003 o 2008.

    If DNS,replication topology,site links etc. are all well still if something odd you notice it means something is not well, w/o smoke there is no fire,try to do some more analysis as the reason i can see is might be dc went offline due to glitch in the link.

     

    Regards  


    Awinish Vishwakarma| MY Blog

    Disclaimer: This posting is provided AS-IS with no warranties or guarantees and confers no rights.

    Saturday, March 19, 2011 4:56 AM
    Moderator
  • KCC is not doing it's job at all if you say it is supposed to load balance them across dc's in a site. I currently have 44 replication connections on 1 dc and 1 each on the other two. This isnt right...
    Monday, March 21, 2011 3:17 PM
  • So I went as far as to delete a number of the replication connections from my hub site and a number of my remote sites, then ran repadmin /kcc site:%hubsite% and repadmin /kcc site:%remotesite%

    Miraculously my connections are now balanced across my 3 DC's at my hub site (I demoted one of the 2003 DC's and now have 2 2008 and 1 2003). Running repadmin /replsum shows no errors and balanced connections across my 3 hub dc's versus a ridiculous number on one and hardly any on the others as before.


    Which begs the question. If the KCC is supposed to run every 15 minutes by default and it's supposed to "load balance the connections" automatically, why did it not do that in my environment?

    Thursday, March 24, 2011 12:06 PM
  • Thanks for the update & great you made it work, regarding why its difficult to say.

     

    Regards


    Awinish Vishwakarma| MY Blog

    Disclaimer: This posting is provided AS-IS with no warranties or guarantees and confers no rights.

    Thursday, March 24, 2011 2:33 PM
    Moderator