How do I close connections to my DVR when they get orphaned

[rmcclell]

I see a exponential increase in connections after stopping the docker container and starting it. I have 9 camera and I use g02rtc to reuse the detect and record streams. Not for sure if it is docker issue or Frigate. I quickly exceeds my QSee QC-9916 20 connection max limit.

ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
9
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
9
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
9
ryan@homelab:~$ docker compose down
[+] Running 5/5
✔ Container reverse-proxy        Removed                                            0.5s 
✔ Container signal-cli-rest-api  Removed                                            0.3s 
✔ Container frigate-notify       Removed                                            0.3s 
✔ Container frigate              Removed                                           26.3s 
✔ Network ryan_frigate_network   Removed                                            0.2s 
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
0
ryan@homelab:~$ docker compose up -d
[+] Running 5/5
✔ Network ryan_frigate_network   Created                                            0.1s  
✔ Container frigate              Started                                            0.9s  
✔ Container frigate-notify       Started                                            1.2s  
✔ Container signal-cli-rest-api  Started                                            1.5s 
✔ Container reverse-proxy        Started                                            1.7s 
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
2
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
10
ryan@homelab:~$ docker compose up -d
[+] Running 4/4
✔ Container frigate              Running                                            0.0s  
✔ Container frigate-notify       Running                                            0.0s  
✔ Container signal-cli-rest-api  Running                                            0.0s  
✔ Container reverse-proxy        Running                                            0.0s 
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
16
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
63
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
71
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
71
ryan@homelab:~$ netstat -nt | grep 192.168.1.225:554 | wc -l
80

[hawkeye217]

We can't help without seeing your config and Frigate logs.

[hawkeye217]

Your logs are showing connection timeouts. When your cameras go down, go2rtc will attempt to reconnect since you are restreaming them in your Frigate config. And it looks like a number of your cameras tend to go offline at the same time.

There's generally no reason to change the default timeouts except in very particular circumstances (especially retry_interval). You also shouldn't use manual input_args. Use a preset.

[rmcclell]

A bug report for this issue would get reclassified as a support request anyways, so yes, feel free to post them here.

I think I summited all need but if you can think of any additional info let me know. I would have preferred it to worked the via the bug submit form but I got 2 of the funny github error pages and I gave up. I tried to scrum any sensitive info but I may have missed some. I work as a dev as well so I know how much of pain it is to find bugs without proper info and organized. I think my first attempts might have been cleaner but it would not submit. I seem to see the connections not closing issue whether restart in command line or from the UI. Initial I though the UI might close them better than me doing it manually. It would be nice to add a feature prevent the DVR from exceeding their limits, most consumer grade DVR are pretty limited and mine struggle to recover once it hit that limit. I could get the DVR logs as well if needed. The unit I have is a Q-See QC9916 with a max connections limit of 20.

• docker-compose.yml
• config.yml
• frigate-logs.txt
• go2rtc-log.txt

[rmcclell]

Your logs are showing connection timeouts. When your cameras go down, go2rtc will attempt to reconnect since you are restreaming them in your Frigate config. And it looks like a number of your cameras tend to go offline at the same time.

There's generally no reason to change the default timeouts except in very particular circumstances (especially retry_interval). You also shouldn't use manual input_args. Use a preset.

So my latency on dvr is exceptionally high for a local connection, haven't figure replaced cables and switches, it appears to be dvr hardware is poor quality. it should be sub 1ms but it test from 5ms to 10ms is there recommended preset you suggest? I'll give that a try. It seems stable unless it gets in connection loop if restart Frigate I usually have to restart the camera unit at same time to get them in sync and prevent the connection flood that overwhelms the dvr.

[hawkeye217]

Yes, that seems very odd. The latency is likely a major contributor to your issue.

You should comment out or remove the retry_interval as I suggested above, and you should just remove the global input_args and just let what you've defined at the camera level take precedent.