Seeing an extremely bizarre problem over here with long running streams over time. Initially, all is well. We use mediacasters to establish the RTSP connection and pull a multicast stream RTP stream that’s H264 encoded. After some period of time, I won’t receive video when connecting as an rtmp client or it will take minutes to get any video to display. If I’m connected to an affected stream, there’s severe degradation and a packet capture reveals I’m only receiving a trickle of P frames. At the same time, performing a packet capture on the machine running Wowza will indicate a healthy flow of RTP video data including IDR frames spread across fragmentation units. I’ve noticed when this occurs, the RTSP session for the stream is nowhere to be found. I see no RTSP packets with a capture and using a tool like netstat reveals there isn’t an established RTSP connection. I expect OPTIONS requests every 15-60 seconds depending on the keepalive value returned by the camera. Inspecting via the JMX interface shows the mediacaster for the stream is active. If I bounce the mediacaster, all is well for hours or longer. Seems to affect a large majority of the streams on this server.
Wowza version 4.8.19+4