[jitsi-dev] Packet loss on videobridge


#1

I've noticed when running a local video bridge it logs consistent packet
loss around 4 to 5%. This is both with using the stock jitsi meet web
client as well as the one we're developing. I took a wireshark trace on
the machine hosting the bridge and didn't see any missing packets coming
into the server. Is this maybe just an aesthetic bug? Before I started to
dig into ice4j wanted to see if maybe this was a known issue.

Logs I'm seeing:
15:05:27.250 INFO: [108]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.38973897389738976
15:05:30.070 INFO: [39]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.39413489736070384
15:05:32.254 INFO: [134]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.4019088016967126
15:05:35.092 INFO: [39]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.3735408560311284
15:05:37.271 INFO: [108]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.40756302521008403
15:05:42.065 INFO: [39]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.358311345646438
15:05:42.272 INFO: [108]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.41125411553427116
15:05:47.273 INFO: [108]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.4141463414634146
15:05:48.073 INFO: [39]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.34963954685890836
15:05:52.303 INFO: [108]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%:
0.4173100365704998


#2

That's just a bug in ice4j's calculation, it can be safely ignored.

Boris

ยทยทยท

On 25/03/15 23:06, Brian Baldino wrote:

I've noticed when running a local video bridge it logs consistent packet
loss around 4 to 5%. This is both with using the stock jitsi meet web
client as well as the one we're developing. I took a wireshark trace on
the machine hosting the bridge and didn't see any missing packets coming
into the server. Is this maybe just an aesthetic bug? Before I started
to dig into ice4j wanted to see if maybe this was a known issue.

Logs I'm seeing:
15:05:27.250 INFO: [108]
org.ice4j.socket.DelegatingDatagramSocket.logRtpLosses() RTP lost > 5%: