This message was deleted.
# helpdesk
s
This message was deleted.
f
How does your CLI command line look? It is possible that participants in the load test are hitting the set time and leaving? Like you said, that is a proper leave.
cc: @little-nest-1933
l
livekit-cli load-test --url ws://10.184.17.33:7880 --api-key devkey --api-secret secret --room load-test --video-publishers 30 --audio-publishers 40 --subscribers 50 --duration 10m
I could not find any set time, but they do leave in between the call. Thats for sure.
f
You can look for logs •
new participant joined
- this is when participant joins •
participant closing
- when participant is closed (the reason in that log tells why) That should be 10m for the participants in the load test. If not and if they are leaving with
CLIENT_REQUEST_LEAVE
, seems like the load test is closing the participant early. If you find it is not 10m, can you please file a bug in
livekit-cli
repo with details?
FWIW, I just tried the same command against latest server code in
master
branch and no participant left before 10m. Also, forgot that the 10m timer starts only after all participants join. So, the least a participant is expected to be in the session is 10m and earlier joining participants could be there for a bit longer. In my run, took 18 seconds for all participants (90) to join. So, the first participant who joined ran for 10m18s roughly.
l
@fancy-wire-61616 There is another issue. Irrespective the permissions I specify by default can publish is true for all participants. This is not ideal.
f
Can you explain more please? In what context are you referring to? In load test? Or elsewhere? In code, it is a pointer which is nil (and hence false) by default.
l
In this load test case
all participants are being given canPublish=true by default.
Be it a suscriber or publisher
I check it, every participant is getting true by default
f
Is that affecting your test? Even though it can publish, only given number of publishers will publish from load test.
l
I was hoping to filter out the publishers in a code for a use case, that’s not happening.
f
Got it. Can you please file a feature request with details when you get a chance please?
l
Ok
f
Actually, the read is returning default
true
here due to backwards compatibility reasons IIRC. We will have to look at subscriber only participants explicitly setting to false.
l
Exactly