HPB problems with lost janus packets

Hi forum!

I’ve problems to get Nextcloud Talk running smoothly with High Performance Backend.
First my setup:

  • Nextcloud 19.0.4, Talk 9.0.5 (also tested 20.0.1 Talk 10.0.1)
  • Ubuntu 20.04.1 LTS
  • janus gateway 0.10.5 (also tested 0.10.4)
  • nextcloud-spreed-signaling version 9206a56113064b8e44b0de12d5128f8e88617679/go1.13.8
  • nats docker latest
  • coturn 4.5.1.1-1
  • libSRTP2-1_2.3.0-4

I’ve setup the HPB with the tutorial from https://decatec.de/home-server/nextcloud-talk-mit-eigenem-signaling-server-high-performance-backend/

Here my problem:
The call is running fine until someone is starting to share a screen. From there all video streams are freezing till the screenshare is not stopped.
In signaling log i can see many entries like:

Nov 12 12:37:58 myhostname signaling[1789]: mcu_janus.go:1012: Subscriber _Tx-QfEhFjvhSFiOnUI2LZnFXscJY7RlRZiIkLepDgZ8aHVDRVotaG5SWkwwaVdYRkhDU0UyVVhXQ2ZhS0phcnFaVVFLdWcteU9maEN5S1VyRnJOeW9pd1l1Y0hPR0QwZlIzbEszd2Z6aUEweXh1d05YWXBGYVFoZlFkalN5b2x3ZHBkM2E5X1cwdkNBdDdaVHJ3ZHZMWTNMaWdlTXpkM01ZQnBJZkJNcTc4VVBWek1qWkdrV0ZQelJwcnpWeTZUYTRTQl95cDl4fDEwNDI4MTUwNjE= (5814447515383452) is reporting 39 lost packets on the uplink (Janus -> client)
Nov 12 12:37:58 myhostname signaling[1789]: mcu_janus.go:725: Publisher _Tx-QfEhFjvhSFiOnUI2LZnFXscJY7RlRZiIkLepDgZ8aHVDRVotaG5SWkwwaVdYRkhDU0UyVVhXQ2ZhS0phcnFaVVFLdWcteU9maEN5S1VyRnJOeW9pd1l1Y0hPR0QwZlIzbEszd2Z6aUEweXh1d05YWXBGYVFoZlFkalN5b2x3ZHBkM2E5X1cwdkNBdDdaVHJ3ZHZMWTNMaWdlTXpkM01ZQnBJZkJNcTc4VVBWek1qWkdrV0ZQelJwcnpWeTZUYTRTQl95cDl4fDEwNDI4MTUwNjE= (7157238091411248) is reporting 14 lost packets on the downlink (client -> Janus)

All other log files are fine.
I tried a lot of stuff:

  • disable ufw on signaling server
  • allow all traffic between signaling server an internet (DMZ), no firewalling between
  • test multiple browsers / clients
  • client’s and server’s CPU/Network/Ram is easy going

Here my configs:
/etc/turnserver.conf

listening-port=3478
tls-listening-port=5349
min-port=49152
max-port=65535
fingerprint
lt-cred-mech
use-auth-secret
static-auth-secret=secret
realm=mydomain.de
total-quota=100
bps-capacity=0
stale-nonce=600
cert=/etc/ssl/certs/cer.cer
pkey=/etc/ssl/private/key.key
cipher-list="ECDHE-RSA-AES256-GCM-SHA512:DHE-RSA-AES256-GCM-SHA512:ECDHE-RSA-AES256-GCM-SHA384:ECDHE-RSA-AES256-SHA384″
no-stdout-log
syslog
no-multicast-peers
no-tlsv1
no-tlsv1_1

/etc/janus/janus.jcfg (shortened)

media: {
        ipv6 = false
        rtp_port_range = "20000-40000"
}
nat: {
        stun_server = "signaling.mydomain.de"
        stun_port = 5349
        nice_debug = false
        full_trickle = true
        turn_server = "signaling.mydomain.de"
        turn_port = 5349
        turn_rest_api_key = "secret"
}

/etc/signaling/server.conf

[http]
listen = 127.0.0.1:8080
[https]
listen = 127.0.0.1:8443
certificate = /etc/ssl/certs/cer.cer
key = /etc/ssl/private/key.key
[app]
debug = false
[sessions]
hashkey = secret
blockkey = secret
[clients]
internalsecret = secret
[backend]
backends = cloud, testcloud
allowall = false
secret = secret
timeout = 10
connectionsperhost = 8
[cloud]
url = https://cloud.mydomain.de
secret = secret
[testcloud]
url = https://testcloud.mydomain.de
secret = secret
[nats]
url = nats://localhost:4222
[mcu]
type = janus
url = ws://127.0.0.1:8188
[turn]
apikey = secret
secret = secret
servers = turn:signaling.mydomain.de:5349?transport=udp,turn:signaling.mydomain.de:5349?transport=tcp
[geoip]
[geoip-overrides]
[stats]

Someone similar problems, or ideas?
Best regards
Michael

Note: Without HPB (no signaling server configured) it work’s perfectly till there are not more than 4 partipicants…

Problem also tested with janus-gateway versions:

  • 0.10.9
  • 0.10.7
  • 0.10.4
  • 0.9.5

Hi michi,

have you considered opening an issue in the nextcloud-spreed-signaling repository on Github?

https://github.com/strukturag/nextcloud-spreed-signaling/issues/new/choose

Hi Chartman,

not yet - i will do this today (after not much feedback here :wink: )

I will try to reproduce it later today :slight_smile:

Did someone find a solution to this problem? I’m facing the same behavior right now. @michi have you opened another issue in the janus-gatway project as recommended here: mcu_janus.go: many lost Janus packages · Issue #59 · strukturag/nextcloud-spreed-signaling · GitHub i was not able to find any.

@Rene_Krick i couldn’t find a solution - i disabled HPB.