Talk + HPB : bad performance even with 2-3 users only

We have set up Nextcloud with separate HPB signaling server and we are trying to observe a significant performance amelioration, but… with only 2-3 users in a video call, video is slow, freeze or simply doesn’t work. On the same network, we can have hundred of video calls with zoom with a pretty good quality.
Configuration seems to be ok, Netxcloud talk settings page says “OK : current version …” so it should work…
We also tried to install the signaling server and Nextcloud server elsewhere on dedicated servers with >1 Tb of bandwidth, using only 4G phones to make a call, so our network/proxy/firewall/whatever could not be responsible of the problem, but same results : even 2 participants cannot talk properly…

Is there somewhere an organization using Nextcloud talk in a production environment ? How many simultaneous video calls can they have ?
What are we doing wrong ?

Is the HPB working? check the browser console if you go into a room and open the logs on the HPB Server.

Finally we get it to work, it was a firewall issue

Let me know how it goes with the HPB. I was considering setting it up but it’s kinda expensive for the server specs required and it’s for non commercial use a privately owned chat for friends, family & who ever feels like they want to register i geuss.

Well, we now have a HPB running without apparent error, Nextcloud says it’s OK, we can launch audio and video calls,… BUT :

  • Video calls with more than 4 people are not possible (lags, failed connection…)
  • No perceptible difference between with or without HPB
  • CPU charge on HPB is near to 0.0
  • Stopping the HPB services during a call doesn’t interrupt the call (maybe it’s normal ?)
  • 4 calls with 2 people each on LAN is correct and usable. But 1 call with 8 people is impossible.

So we wonder if the HPB is effectively used by Talk. How can we ensure that it’s the case ?
I saw somewhere on the net a guy describing a call with 32 people simultaneously so I think it’s possible and we certainly fail at some point. But where ?

Sounds like your nextcloud is NOT using your HPB. If I stop the backend service the call ends… And I had calls with about 10 people (including video) withaout any problems…

Why the Nextcloud Talk preferences page is all OK if the HPB is not used ? How can we confirm and resolve that ? Any clue ?

We have found the solution to our problem. In the HPB control panel, our url was in the following format:

The url must be in the format wss:// to work and use the wss protocol. The right url is :


I anyone using sophos utm 9 as a reverse proxy?
I have problems with websockets since update to sophos utm 9.709-3.
Nextcloud - Settings Talk - HBP: wss://mydomain/standalone-signaling test says “ok running version…” but in talk i have problems while connecting to high performance backend.
Signaling server on my Ubungu machine throws errors like this: “websocket: the client is not using the websocket protocol:” and proxy protocol is throwing 400 errors while accessing standalone-signaling/spreed.

i did several tests with the trailing slashes, but i thing sophos did something special with the security fixes.

can somebody help?

thank u best regards