This message was deleted.
# helpdesk
s
This message was deleted.
d
just tried to repro locally, have up to 4 participants and all receiving streams.
will dive into the logs and see what I can spot.
question for you.. what do you use to view the JSON in the CSV logs? it's a bit hard to parse
w
its a dump from kibana šŸ˜• if you run it locally/remotely you can import the logs
d
is it possible to dump a text-only version?
w
we can invite you to the kibana if you want
then you'll have direct log access
d
sure, that works
w
and I can just give you links to queries
cool, email?
d
cool invited
here's a query you can click at anytime that will show: • All livekit server logs • All client side logs from me, you, and furqan
d
nice
w
hope it helps, lmk if you have any q's
d
kk, will have a look
the logs you sent me has a timestamp of
"2021-02-03T22:13:44.216Z"
I'm not seeing those on kibana
w
which logs are you looking at rn
here they are
d
ahhh ok timezone diff then
w
I just manually set the time range to 5pm est -> 6pm est which is around when I tested that
d
the exported ones are GMT
w
ah yes sorry kibana default
d
cool no biggie.. but this is good will take a look
w
sweet
d
am I right to guess that each user has a diff email?
each livekit participant has a diff email
w
yes
we never pass the email to livekit tho, we pass our internal IDs
d
yup, that's fine.. just trying to understand the order of operations
is this consistently reproducible? I see one bug that's in the pion stack.. been meaning to submit a patch to that
we seem to be missing some logs here.. at least for the server
for example.. I see that you have
error getting buffered packet
printed out..
but missing all of the debug logs..
other than the crash.. I do see the error that we all dread NACK causing new tracks not to be added
w
I can enable the debug logs
is that configurable through the env variable?
d
nope.. it's part of the config file
w
šŸ˜• well we started deploying the docker image now
any chance you can make that configurable through env var?
d
k. also if you are building from source
w
I can add a PR too if you want
d
try the dump-sdp branch
w
using the dockerhub image
d
kk
I do have a good clue tho.. but this is the same issue we saw before
has little to do with # of participants.. more with nack causing issues
w
gotcha
d
you can pass --dev to enable debug
w
cool just added that. I'll need to reproduce the bugs for those logs to surface, old debug logs are gone now
also having redis problems so I can't reproduce for maybe an hour or so
I'll send you a timestamp after I reproduce again
d
and which users stream went out?
w
its super odd
I think you should join a call with us and observe it
easier than trying to describe it in text, too lossy
can jump on meet.google.com then jump on livespot call together and observe behavior
lmk if you have a few min for that
d
I'm afk right now
w
no worries
d
will jump on later.. but a description would help regardless.. if it's not completely out.. but lossy.. that's good to know tpo
too
w
if thre's a time that works for you lmk
total loss for some users
they'll get a
TrackUnsubscribed
even tho publisher never unpublished
Copy code
"room event" "trackUnsubscribed" {"_events":{},"_eventsCount":2,"kind":"audio","trackSid":"TR_Pi5iAhEsAPMwd7cC5MxY7H","trackName":"","isMuted":false,"track":{"_events":{},"_eventsCount":0,"kind":"audio","name":"","isStarted":false,"isMuted":false,"attachedElements":[],"mediaStreamTrack":{},"sid":"TR_Pi5iAhEsAPMwd7cC5MxY7H"}} {"_events":{},"_eventsCount":5,"sid":"PA_s93t6vmiUJPvjcachJwhPh","identity":"0x73F8957f89AD035a880d2DacE78937cee9F01502","audioTracks":{},"videoTracks":{},"dataTracks":{},"tracks":{},"participantInfo":{"sid":"PA_s93t6vmiUJPvjcachJwhPh","identity":"0x73F8957f89AD035a880d2DacE78937cee9F01502","state":2,"tracks":[{"sid":"TR_Pi5iAhEsAPMwd7cC5MxY7H","type":0,"name":"","muted":false},{"sid":"TR_TsmF6RxucqEcyHLFxBXeP5","type":1,"name":"c922 Pro Stream Webcam (046d:085c)","muted":false}]}}
for example
different users are getting different track events, its very odd, hard to describe procedurally
I'll have everyone on this test call record their screen and send them to you stitched together
d
yeah I'll hop on later tonight. hanging with my kids for a bit
w
sounds good, have fun!
okay I recorded a video of the problem so we can keep debugging async
top left screen is my computer
bottom right screen is a friends computer
I simulated a third user with a private tab
notice the events that are fired on each window and what happens when publishers start to call in
hope it helps!
d
ok, just getting on, thanks for putting the video together.. free to try it live too
w
ok cool, if you'd like to I'm available now!
d
sure, let's do it
which session?
w
d
wait google meets or livespot?