I am using HAProxy to load-balance RDP connections between clients and RD Gateway server.
There are two RD Gateway servers( behind HAProxy, the balance method is ROUND_ROBIN, and stick session by SOURCE_IP.

Here is my config:

    user nobody
    group haproxy
    log /dev/log local0 debug alert
    log /dev/log local1 notice alert
    maxconn 2500
    stats socket /var/lib/neutron/lbaas/v2/7ee12684-c45d-4af4-8ae7-381743beb78e/haproxy_stats.sock mode 0666 level user

    log global
    retries 3
    option redispatch
    timeout connect 5000
    timeout client 50000
    timeout server 50000

frontend 377c24a1-1735-4947-a854-0d1fb8d0cd97
    option tcplog
    maxconn 2500
    mode tcp
    default_backend 51d6cead-368e-45a3-bcda-b1fb8cd5f2dd

backend 51d6cead-368e-45a3-bcda-b1fb8cd5f2dd
    mode tcp
    balance roundrobin
    stick-table type ip size 10k
    stick on src
    timeout check 5s
    server 3ccbdfa9-fa69-4388-ad1b-046731825659 weight 1 check inter 5s fall 3
    server 7dde5bd2-ad90-4668-9642-466446646948 weight 1 check inter 5s fall 3

It works but I found that when the backends RDP session is disconnected from RD Gateway(I can see the disconnected logs from Event Viewer), clients connection doesn't stop, I still can see the VM's screen(but can't do anything), it means that the session still keeps connected, why?

What I expect is the frontend should get disconnected info from the backend and try to re-establish the connection again(cause the session has already disconnected from the RD Gateway server).

Is there any miss-configuration? I have no idea about this issue, any help is appreciated.


Strange, The client should get a popup asking to reconnect from the connection broker. Your timeouts are very small, this blog recommends tcp keepalive aswell :

clitimeout 1h
srvtimeout 1h
option tcpka 

I do vaguely recall someone saying RDS had an option to enable/disable the reconnect option?

For more background information and to review the other deployment options you have, take a look at this RDS deployment guide.

| improve this answer | |
  • Thanks, I increase the timeout to 1h. However, option tcpka doesn't work in my case, and I notice that you mentioned about connection broker, I only run terminal service as my RD Gateway, is that the keypoint causes the issue? – Corey Sep 23 at 8:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.