I setup a remote desktop farm to test it out before going to a production environment. Current setup: two remote desktop servers (TSFarm1 and TSFarm2) with a third server that has the Broker role installed (TSFarmBroker). I am using Network Load Balancing to balance the load.
Farm2 is set to not accept any new connections.
I have 4 users (Uno 1-4) logged in, two on Farm 1 (Uno1 and 2) and two on Farm2 (Uno3 and 4)
I logged out Uno3 from Farm 2, it logged back into Farm1. No problem.
I then closed the RD session for Uno4, so it went into a disconnected state on Farm2.
When I created a new RD session for User 4, it sometimes reconnects to Farm 2 like it should, but there are also times when it will create a new session on Farm 1 leaving the Farm 2 session in the disconnected state.
How can I prevent this from happening?
I cannot have a user with multiple RD sessions. If they left a window or a program running on the disconnected session, they would not get back to it on the active session on the other server.
Here is a screen shot from the Broker showing Uno4 connected to multiple servers:
Todas las respuestas
Have you restricted users to a single session?
When user4 is able to resume the session on Farm2 was he redirected to Farm2 from Farm1 or did NLB push him to Farm2? Just to be sure we have our bases covered, would you verify that tsfarm1 and tsfarm2 in the "Session Directory Computers" group on the TSFarmbroker server? http://technet.microsoft.com/en-us/library/cc753630.aspx
Looks like there might be part of my problem here. I double checked the Session Broker Computers group on TSFarmBroker, and all I saw were SID's that didn't resolve to the server names.
I have removed the entries and readded the TSFarm1 and 2 to the list. I will reboot and test again.
Hope this solves it. I'll post back either way.
So TSFarm1 and TSFarm2 are both listed in the Session Broker Computers group in TSFarmBroker.
All three servers were rebooted just to be sure things were "clean".
I logged in all 4 users, the load balancing worked well.
Started doing the same scenario of a disconnected user (Uno1) on a server marked for no new logins (TSFarm1).
In the first 5 or 6 reconnects, I could see that Uno1 was connecting to TSFarm2 initally, then redirected to TSFarm1 because that is where the disconnected session was.
However, on the 6th or 7th reconnect (I was interrupted and lost count), the same thing happened. Uno1 is now logged in twice.
Does TSFarmBroker also need to be in the Session Broker Computers group? I don't recall that specifically.
I think I may have found (and fixed) another issue with my setup.
I forgot that I had setup these virtual servers in another domain and then moved them to a test domain.
The the farm server name was still pointing back to the old domain information.
I had already taken care of the Hyper-V setting to allow network load balancing.
More testing. I'll report back what I find
It sounds like for some reason the Connection Broker is not working at all. I think it seems like it is working some of the time is because NLB is redirecting you to the server the user's session is disconnected from.
I would start to check settings:
1. Verify firewall setting on TSFarmBroker
2. Verify both TSFarm1 & 2 are using the same Connection Broker farm name
3. Use IP Address Redirection is enabled (This is the default).
In my environment we set all connection broker settings in GPO:
Thanks for the suggestion to set things through GP. I went ahead and did that just to make sure both servers had the same settings.
Now that it's a day later, I checked the settings on TSFarm1 and 2 to verify GP has applied, so on to more testing.
This time I left the setting for accepting connections alone, so both can accept, and I was able to duplicate this issue again.
I have noticed that I reconnect correctly if I try it right away, but after about 10 minutes, then I am able to login twice.
So now I'm wondering, is there a conflict with the RD Connection Broker load balancing and the Network Load Balancing?
Is the one taking precedence over the other and not acknowledging the disconnected sesstion?
The RD Connection Broker load balancing setting isn't listed in your screen shot, so I presume it's not set.
How are you load balancing your servers?
Thanks for the help!
I am using a hardware based load balancer for my terminal servers. The one I am using is made by Kemp Technologies, but this dose not sound like an issue with NLB. If you would like to try something other that NLB to see if it will fix your issue you can download a trial of of the Kemp load balancer in the form of a VM: http://www.kemptechnologies.com/us/server-load-balancing-appliances/virtual-loadbalancer/vlm-download.html
or you can use an open source solution like Crossroads Load Balancer:
One other thing I would check is to see if you are getting the following message in your system log of your session broker server when you reboot one of the terminal servers:
- Editado Joel VanOs jueves, 10 de enero de 2013 2:40
I to am seeing this issue, on multiple farms.
I can assure you it's not a function of wndows nlb. I've seen it using both our F5 load balancers and Windows NLB. (yes, I switched from ip to token based redirection and back as required here...)
It seems to occur more frequently on farms that allow longer connections, i.e. you don't force logoff of disconnected sessions or allow them to remain for multiple days. It does occur on farms that we have shorter disconnected timeout limites on as well. Some farms allow connections to remain for as much as 7 days, others as little as 6 hours, all display this behavior.
These are Win2K8 R2 RDS Hosts and Connection broker/s.
Has anyone found a cuase or solution yet???