Jitsi-meet-torture - Can it run on a secure domain ? (also, what does it do?)

I would like to know if jitsi-meet-torture can do it’s magik on a secure domain?

I tried it (after a lot of work to make it run!) and looked at the screenshots… They are trying to log into a room and can’t because there is no host…!
org.jitsi.meet.test.ActiveSpeakerTest-web.participant1|620x344

Can it run on a secure domain?

Also, what does it test? What useful data can I pull-out of a test run? Does it stop by itself or do I need to stop it myself?

This module is not documented at all appart from a basic line of code to run it!

@damencho, sorry to call you out again, but I think you are the most susceptible person to be able to answer this… :slight_smile:

Also, if it can’t run on secure domain, what is the best way to temporarily disable secure domain during the test?

I tried removing the line in jicofo and it seems to work, but I am not certain

Torture is not designed to run on secure domain.

It has the logic to stop itself. We use this as integration tests, run on every PR across projects.

Not sure about this one. Maybe you can probably make it work with secure domain providing username and password the same way jibri uses those, by adding localstorage values for username and password:

Thank you @damencho!

I was able to make it run by simply removing my domain from sip-communicator.properties during the test…

Although, I have no Idea what I was testing and what the results meant! It ran 275 tests and 49 failed…

What I am after, is a way to tell how many conferences and participants our server can handle at once… Am I doing the right thing by running this module?

Also, I am running it on the same instance, is this the way to go or should I use a client computer/server?

Spec wise, our server is an 8 core Xeon with Hyperthread running at 4.2ghz with 32gb of RAM on a gigabit link. (dedicated server at OVH)…

Our concern is that we are running a big event next week (a virtual trade show) and the participant’s response far exceeded our initial prediction… We are now certain to have 1000 people minimum, in around 100-150 rooms but it might grow a lot bigger (same number of rooms, more participants!)… We are able to mitigate the effects by offloading the keynote presentations to youtube, but the sub conferences will still pump a lot of “juice” !

What do you think… Should I plan on scaling-up already?

regards, Normand

There is a malleus script you can use to create multiple participants.

You need to create a selenium grid with multiple grid nodes, maybe you need 4GB Ram per chrome instance per node, this is for one participant.

So you will need a lot of resources to stress it with over 1000 participants.

How do you test your instances against large loads at meet.jit.si ?

Also, in your experience, what can I expect from our setup in terms of max participants and rooms?

Make sure you are using epoll as network backend for prosody, nake sure you don’t have filehandles limitations on prosody, nginx …
And if you have just one bridge … I would say add at least 3 or 4…

Okay, I will need to google that as I have no idea what you are talking about!
For those applications, we used the default setup…

That brings me to another question… Would we have been better off splitting the server in 4 (via VM or dockers) instead of having the machine on a single overpowered server? Would jitsi ever use our 8 cores 16 threads… Does it scale well on multiple cores?

Jvb yes, it checks cpus and spread the work, I think. For bigger scale it is recommended to have multiple bridges, to spread the load and the bandwidth. Where the bandwidth normally is the first problem to hit.

Okay, @damencho, I googled about filehadle limitations and epoll, but I am so stressed-out by this event that I can’t think straight… I do not understand what I read… Can you rephrase this in a way my dog could understand?

Also, we could potentially have a 10gbps symmetric link on our server for one day… How many user do you feel our setup could handle with that? 1000? more? (not asking a definitive number here, just your gut feeling.

Hard to say … but its always good to have in backup few bridges if you don’t have autoscaling.

That is what we are looking into… our main instance is hosted at OVH.
The hardware, internally, is really powerful for cheap pricing, but public bandwidth is not great!
Their scalable solutions are not that great neither…

I am looking if I can get extra video bridges on amazon cloud to backup the main instance… Could that work? Would jitsi offload automatically to those bridges if I follow the guides on jitsi scaling?

Yep, adding more bridges and jicofo will do the job and will spread conferences across bridges

Okay, and if I followed the guides, am I okay for the epoll thing and filehandles limitations? What manipulations do I have to do to check that and fix it if needed?

Sorry @damencho if I ask so much, I am really in needs right now, the learning curve is steep and the event is in 7 days! Thanks a lot for your time!
After all this will be done, I will keep contributing to the community (I already do) and mostly try to make some French Guides because there are a lot of french speaking people on this forum :slight_smile: