-
-
Notifications
You must be signed in to change notification settings - Fork 1.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Support]: Camera feeds keep crashing #7642
Comments
|
I am still able to access that camera through Synology DS Cam, which I set
up as a backup. So the camera is still online.
Why would frigate be unable to reach it?
If it is helpful/diagnostic, when I reboot frigate, all the cameras work
again.
…On Thu, Aug 31, 2023 at 3:51 PM Nicolas Mowen ***@***.***> wrote:
tcp:https://192.168.7.166:554?timeout=5000000 failed: No route to host means
the camera is not on the network at that ip address
—
Reply to this email directly, view it on GitHub
<#7642 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AOX6NXHMPNOQZAK6U65B7OLXYDTKXANCNFSM6AAAAAA4GSA2QY>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
potentially because the camera is struggling to provide multiple streams at once. Ffmpeg makes no mistake on that, it is not seeing the device on the network or the stream is timing out and stops sending data |
So that error about a CUDA-compatible device is irrelevant? I only spun up the Synology system after these problems started happening with Frigate. The whole point of using frigate was that I had minimal connections to the camera itself. |
it is its own problem, but the logs are clear about the camera
you're not using go2rtc, so frigate is not restreaming in any case, that is irrelevant, I was only offering an example, not saying that was definitely the reason |
Entirely possible I have multiple issues here.
|
frigate is "trying again" hence the repeating logs
if you don't have go2rtc setup then the home assistant entity is just a snapshot that updates once per second
for one I would start by using the recommended ffmpeg preset instead of manual args. Other then that, need to look up the error and see what others say on it |
It seems like my setup is pretty out-dated. Can you point me to a setup page where I can learn more about this? I remember battling for a while to find the proper manual arguments to make these cameras work, but like I said earlier that was several versions of frigate ago. |
So I should
to this:
OK I'll make those changes and try again. Thank you! |
After making your recommended changes, I went back to HomeAssistant and I'm definitely getting more than one frame per second on my dashboard through frigate. Is there some other way (perhaps an old, abandon system) that was able to make that happen without go2rtc? |
yes, it used to use rtmp, that has been deprecated, as is explained in the 0.12 release notes |
Thank you very much for pointing out those Release Notes to me, I had not seen those before. That was the first I saw your involvement. Let me take a second to say THANK YOU for all of these changes. It looks like I was not paying attention for too long and you've accomplished much of what I was disappointed by not having previously. I have some tinkering to do. I'm not sure I understand how my frontdoor feed is still working if rtmp is deprecated and i have rtmp enabled: false in my current config. Is there something more I would need to investigate/do to ensure that is truly turned off as I look into enabling the built-in go2rtc? |
Still working where? as I said previously, the HA feed is just a snapshot that updates once per second if go2rtc is not setup |
I wish you were sitting next to me. Here I have the pro telling me there's no way I'm getting more than 1fps, but I'm sitting at my computer looking at HomeAssistant and watching a bird smoothly fly by and trees smoothly waving in the wind. When I first navigate to that page in HA, you are 100% right. It's not great quality, and its 1fps. But after a few seconds, the "high quality mode" kicks in and I get HD camera quality and a pretty smooth fps. Is there a log I can share with you that would show you what is happening? |
what is your current config? |
In HA or in frigate? The only change I've made in frigate is the codec thing u suggested a minute ago. In HA, I am using a custom frigate card. I see the little bird-like icon in the topleft corner of the stream I am talking about. My yaml looks like this:
|
oh, you didn't say you were using the frigate card, in that case it is using the jsmpeg feed from frigate which is 5 fps |
5fps is plenty for my re-streaming purposes. So then I wont bother with the go2rtc since I can use HA to get 5fps on all my cameras using the frigate card. That works for me. |
How long did it stay online before? |
Not long. Maybe a day or two. |
I am trying to set up FrontDoor with separate streams, but its not working. Any idea why? Does that "Roles" stuff not work anymore?
|
what are the logs? |
I don't get any. Frigate won't event start. I'm gonna shut down for the night. Thanks again. |
if frigate is failing to start, there will be logs showing the error in the config |
For what it's worth, I had the same problem with cameras and streams getting disconnecting and giving error messages. Here's how the problem went away (for a couple of days now, I may need to update that in the future if it reappears). Solution 1: Do not use the Coral at all; use OpenVino instead. CPU is "Intel(R) Core(TM) i3-4150T CPU @ 3.00GHz". Coral's inference speed is slightly over 10ms. OpenVino's inference in this CPU is 30-40ms. If that is acceptable, try it. Solution 2: Bind mount the standard edgetpu library to the docker container. The container has preinstalled the package "libedgetpu1-max" which has the Coral running at the maximum frequency (500 MHz). |
@s482dcaw sounds like your device is struggling to power both USB devices, that's when a USB powered hub is recommended. |
That is probable, and since the error messages are pretty similar, others may have the same problem. That's why I wrote down the details. |
it did not... :-( (base) josh@PowerSpec:~/Docker/frigate$ sudo vainfo |
No, I run my integrated GPU alongside nvidia GPU without issues. Okay but what happens when you do |
3 options: |
and that was run on the host right? |
I'm using ssh to do all this from a different machine. So... yes? |
okay, was the machine restarted after the guide you followed? |
No. I will try a reboot. |
rebooted and tried vainfo: |
after reboot what is in /dev/dri ? |
Same three options |
what about if you run |
|
okay try running
|
|
I tried that twice. Once combined as you gave it, then separately. It gave same results both times. |
Not sure then, seems it doesn't like the driver or something else is weird |
Frigate is the core function for this machine. Is there a better OS I should start over with? |
debian is what the docs recommend |
I decided to leave the current machine running while I set up Debian on a different one (another i7 workstation with a 30-series GPU). Question for you - if I leave the config.yaml with no hwaccel arguments in the current build, what exactly is happening? Is frigate asking the CPU to do the decoding without using the CPU's built-in GPU? So it's still the CPU doing the work, its just like asking the general-purpose CPU to work harder than if it could utilize the GPU parts which would be more efficient at doing the same thing? Do I understand that correctly? If so, am I over-tasking the CPU to handle 5 video feeds in this inefficient method for the time being? |
yes, it just uses the software decoder. Depends on the CPU and resolution of the stream that is being decoded |
Its an i7-10700. |
yes |
Question - does the stuff under each camera on the System page in frigate always add up to 100% for CPU? Like, it's giving you a breakdown of the total demand being made by Frigate (regardless of the total potential of that CPU)? Or is the System page a good way to gauge how hard the CPU is working overall? |
each CPU % field is the % of a single CPU core that that specific process is using |
That's very useful, thanks. Looks like my frontdoor is consistently eating up between 30 and 50% of that CPU. The ffprobe on that System page lets me see that the low quality stream is also running at full HD. I'll go into that camera's settings and change the quality of the backup stream. Everything else on here is typically below 5%, which seems like not a bad way to let it run long-term. |
another thing to consider is that the software decoder is slower than a hardware decoder, so there will be more latency in the detections |
For the sake of completing the record. The camera was set up fine. My config.yml erroneously had the same rtsp path for both detect and record. I corrected the config file. |
Update: the new machine has been running smoothly. New hardware, running on Debian, but all my yaml files are the same. Must have been something in the hardware causing the original problems. One last question for you before I mark this closed... |
Yes, multiple containers can access the nvidia gpu at the same |
Describe the problem you are having
I have had frigate for a while. It started intermittently losing connection to my cameras. I upgraded to latest version, which included upgrading all my HomeAssistant Notifications. Now that I am on the latest version, I am still having intermittent issues where one (or more) of my cameras start showing "no frames have been received, check error logs"
Rightnow its affecting only 1 of my cameras. But it does sometimes affect others. Logs below reflect current state, where only the "Frontdoor" camera is offline.
Version
0.12.1-367D724
Frigate config file
Relevant log output
FFprobe output from your camera
Frigate stats
Operating system
Other Linux
Install method
Docker Compose
Coral version
USB
Network connection
Wired
Camera make and model
Amcrest AD110
Any other information that may be helpful
No response
The text was updated successfully, but these errors were encountered: