Add healthcheck

[felipecrs]

This is work in progress, I will write more details before I mark it as ready for review.

Benefits:

• Hass Supervisor watchdog (if enabled) will restart the container once it is unhealthy

For now, this is blocked by:

• --start-interval does not do anything moby/buildkit#4355

[AlexxIT]

API port may be closed or using another number

[felipecrs]

Right, I have an idea to fix that.

[bonuzzz]

@felipecrs as go2rtc has static config path in docker, you could parse config file and use it in healthcheck sh script. It's possibility describes here https://stackoverflow.com/questions/55375371/using-a-sh-script-for-docker-healthchecks
Anyway I've never seen go2rtc hanging...
As variant, you could parse output existence of "go2rtc --version"

[felipecrs]

Anyway I've never seen go2rtc hanging...

People in Frigate's world have seen go2rtc hanging many times. That's why Frigate (myself) built its own healthcheck for go2rtc to restart it when it hangs.

I, for one, have an issue open that describes a hanging situation happening in latest version 1.8.4:

• go2rtc does not recover connection to camera unless I restart go2rtc #762 (read last two paragraphs)

As variant, you could parse output existence of "go2rtc --version"

I don't see how invoking a new go2rtc process would help to assess the functionality of the existing one.

[felipecrs]

@felipecrs as go2rtc has static config path in docker, you could parse config file and use it in healthcheck sh script. It's possibility describes here https://stackoverflow.com/questions/55375371/using-a-sh-script-for-docker-healthchecks

That's a very good idea, and it was what I was intending to do too. But then I forgot about this PR. lol

[bonuzzz]

I mean I had never seen hanging go2rtc myself and I don't know what exactly needs to check like rtsp port or web port. What is staying accessed when process hangs.

[bonuzzz]

Probably it could be related to this pr #440 to track for example sighup signal

[felipecrs]

@bonuzzz, can you elaborate?

[bonuzzz]

@felipecrs unfortunately I don't know how to reproduce go2rtc hanging in my environment. I think more complex solution is needed here than just to check open ports, because we don't know how the process behaves itself in hanging state.
I saw your homeassistant automation to check stream status and in my opinion this is the best solution at the moment.

[bonuzzz]

Very dirty solution for frigate is parsing /dev/shm/logs/go2rtc/current for phrase "sync.(*Mutex).Lock(...)", "goroutine" of "WRN [streams] json locked" like tail -f /dev/shm/logs/go2rtc/current | grep "json locked"

[felipecrs]

I think the current solution Frigate uses is already very good. :)