AX 2012 R3 NLB for services

(0) Share

Report

Posted on by Thomas Gundolf

225

Hello everyone,

i have a small problem with understanding NLB for Services. I have an Environment with 3 AOS Servers - they are all in an AOS-Cluster for RPC (TCP Port 2712) and i wanted to also install NLB for Services as described in https://msdn.microsoft.com/en-us/library/hh397322.aspx

I installed NLB Feature on all three Servers and made a multicast NLB Cluster for the TCP-ports 8101 and 8201.

On the Reporting Server i added the two configuration keys in the registry as described in the article, i refreshed WCF config which worked and could find the NLB address in the configFile.

Do i also use this config-file for all my Clients or just on Servers that Need to connect with WCF? Do my ordinary Client sessions (TCP 2712) Need to know of the NLB Cluster for Services? For example if i open a Report in the Client - does the Client use Services or does just the Reporting Server use Services?

I also had the Problem that if i stopped AOS 1, no Connection could be made also with the normal AOS-Cluster in the application (TCP 2712) - the Client crashed when i tryed to connect to AX. But sometimes the Client opened after a while without crashing - randomly it worked or not?!

Is there anybody out there who can describe me NLB for Services in Detail. Also a good questions is to use Unicast or Multicast?!

Thank you,

Thomas

*This post is locked for comments

I have the same question (0)

All responses (4)

Answers (0)

Community Member on at

Like (2)

Report

-----

Q: Do i also use this config-file for all my Clients or just on Servers that Need to connect with WCF?

A: You could use your load balancer to balance the port 2712, but you would loose the functionality of balancing equally your users across your AOS servers. Otherwise, I think you should use the AOS Cluster solution

-----

Q: For example if i open a Report in the Client - does the Client use Services or does just the Reporting Server use Services?

A: This link explains well the Reporting Services architecture technet.microsoft.com/.../dd361996.aspx

Add note : The client initiates WCF and RPC connections to the AOS.

Source: blogs.msdn.com/.../are-your-ax-2012-aos-servers-load-balancing.aspx

------

Q: I also had the Problem that if i stopped AOS 1, no Connection could be made also with the normal AOS-Cluster in the application (TCP 2712)

A: Do you have all other AOS listed in your configuration file ?

------

Q: Also a good questions is to use Unicast or Multicast?!

A: Select the unicast method for distributing client requests, unless only one network adapter is installed in each cluster host and the cluster hosts must communicate with each other. Because Network Load Balancing modifies the MAC address of all cluster hosts to be identical, cluster hosts cannot communicate directly with one another when using unicast. When peer-to-peer communication is required between cluster hosts, include an additional network adapter or select multicast mode. When the unicast method is inappropriate, select the multicast method.

Source: technet.microsoft.com/.../cc782694(v=ws.10).aspx

Was this reply helpful? Yes No
Thomas Gundolf 225 on at

Like (0)

Report

Hello Matt,

thank you for your answers. There are some questions left:

The Client calls the Reporting Services with the URL i think - is that true? So the Client still doesn't need to know of the NLB-Cluster the reporting Services are connecting to.

I had all my AOS-Servers listed in my configuration file. Thats what was a bit strange to me. My understanding is, that the AX-Client uses for Startup the "normal" AOS-Cluster (port 2712). I have to do more Investigation on this Kind of Problem as soon as i'm at the customer site. Maybe deinstall the NLB Feature on the AOS and try again just to test the AOS-Cluster (without NLB).

Kind Regards,

Thomas

Was this reply helpful? Yes No
Suggested answer

Brandon Wiese 17,788 on at

Like (1)

Report

Never use Unicast for NLB, especially if they are Hyper-V guests, as the switch flooding can be significant. Hyper-V virtual switch appears to start dropping packets at 100k multicast packets/second (this is how the packets will be classified under Hyper-V perfmon counters), which is easily reached during high stress operations (AOT compile, MRP, etc.) on a Hyper-V system with many guests, when each packet is flooded across all virtual switch ports. You will also incur a performance penalty on intra-machine traffic between the AOS NLB cluster and clients and the SQL Server database even without the packet loss. Use Multicast.

Was this reply helpful? Yes No
Tommy Skaue Moderator on at

Like (0)

Report

@Brandon: Have you tried IGMP Multicast? blog.workinghardinit.work/.../windows-nlb-on-windows-server-2012-r2-hyper-v-a-personal-preferred-configuration-using-igmp-with-multicast

@Thomas: Please mark valid answers so the contributors feel your love and gratitude. :-)

Was this reply helpful? Yes No