Can’t access remote desktop services – server pool contains RD Connection Broker servers that are both clustered and not clustered

rdsremote-desktop-serviceswindows-server-2012-r2

We've had a random error crop up that's preventing us from publishing remote apps. Every time we go to manage Remote Desktop Services in Server Manager, we're greeted with the above error. We only have a single Connection Broker set up, but we set it up in HA with the thought that we might add another down the road. I've run over the configuration, and everything I see points to it being set up correctly. The only weird thing is that we didn't have port 1433 open for SQL on the SQL Server, but that didn't seem to be an issue previously.

This is not a new setup either–we've been publishing remote apps to clients through this gateway for about six months now.

Right now I'm trying to decipher the entries in the ErrorLog table in the SQL DB. Any ideas on where else to look for misconfiguration would be appreciated.

OS: Windows Server 2012R2
SQL: MSSQL Server 2012

Unfortunately we can't pinpoint when this issue cropped up beyond within the last couple weeks, so we can't isolate what changes we made that could've caused this. We had a tech accidentally reboot the SQL Server three weeks ago, and there are a ton of errors in the log on that day at that time. The date modified on the databases lines up with that too. But we suspect that our techs have been able to publish remote apps since.

Best Answer

Turns out someone added the Connection Broker role to one of our Session Hosts. So much facedesk.

Edit: Like I said in the original post, we have our Connection Broker set up in high availability mode (clustered) with the thought that we'll add another one down the road. A tech added the Connection Broker role (unclustered) to one of our Session Hosts (server that hosts the remote app). He then added it to the server pool in Server Manager on our server that manages the collections, resulting in two Connection Brokers as part of the pool, one clustered and one not. RDS didn't like that.

Edit 2: I've been made aware that the reason we couldn't just remove the server from the pool is that the Session Host in question had already been added to the RDS deployment and had Remote Apps published from it. The Connection Broker role was added after the fact in some misguided troubleshooting, and had to be removed from the server for RDS to be accessible again.