Just pull the v3.0.0b1 docker tag!

In case someone is not familiar:
docker pull roflcoopter/viseron:3.0.0b1

It should be recorder, not recordings. Typo in the docs, will fix!

The 24/7 recordings are stored as short segments in the segments folder.
Is there any files in there?

Yes, it looks like it's storing it in the segments folder in the container volume instead of the tier 1 I specified which is in /recordings that is linked to an ext4 drive folder. It did make a frontdoor folder like the picture above.

It created another recordings folder in /recordings with all the other cameras except for the frontdoor.

The docs are unclear on this, will try to explain here before I fix the docs.

Since you specified this config for your front door camera:

      recorder:
        continuous:
          max_age:
            days: 14

It will override the storage component config and the path will be set to /recordings

If you want to override the entire storage component config for this specific camera you can do so by using storage under recorder like this:

ffmpeg:
  camera:
    frontdoor:  # This value has to be unique across all cameras
      name: T-FRONTDOOR
      host: 192.168.1.100
      port: 8554
      path: "/T-FRONTDOOR_LOW"

      width: 640
      height: 360
      fps: 30
      codec: h264
      rtsp_transport: tcp

      storage:
        recorder:
          tiers:
            - path: /recordings
              events:
                max_age:
                  days: 14
              continuous:
                max_age:
                  days: 14

I am away and typed this out on my phone so might have messed something up but hopefully you get the idea.

Will think on how to make it a bit more intuitive as well as improve the docs

I think I followed your instructions correctly but I'm getting the error below.

Config:

ffmpeg:
  camera:
    frontdoor:  # This value has to be unique across all cameras
      name: T-FRONTDOOR
      host: 192.168.1.100
      port: 8554
      path: "/T-FRONTDOOR_LOW"
      #username: "admin"
      #password: "1234"
      #path: "/ch0_1.h264"

      width: 640
      height: 360
      fps: 30
      codec: h264
      rtsp_transport: tcp

      storage:
        recorder:
          tiers:
            - path: /recordings
              events:
                max_age:
                  days: 14
              continuous:
                max_age:
                  days: 14

Log:

[2024-04-27 11:01:47] [ERROR   ] [viseron.components] - Error validating config for domain camera and component ffmpeg: extra keys not allowed @ data['frontdoor']['storage']. Got None
Traceback (most recent call last):
  File "/src/viseron/components/__init__.py", line 278, in validate_domain_config
    return domain_module.CONFIG_SCHEMA(config), None
  File "/usr/local/lib/python3.8/dist-packages/voluptuous/schema_builder.py", line 272, in __call__
    return self._compiled([], data)
  File "/usr/local/lib/python3.8/dist-packages/voluptuous/schema_builder.py", line 594, in validate_dict
    return base_validate(path, iteritems(data), out)
  File "/usr/local/lib/python3.8/dist-packages/voluptuous/schema_builder.py", line 432, in validate_mapping
    raise er.MultipleInvalid(errors)
voluptuous.error.MultipleInvalid: extra keys not allowed @ data['frontdoor']['storage']

Sorry i messed it up, should be like this:

ffmpeg:
  camera:
    frontdoor:  # This value has to be unique across all cameras
      name: T-FRONTDOOR
      host: 192.168.1.100
      port: 8554
      path: "/T-FRONTDOOR_LOW"

      width: 640
      height: 360
      fps: 30
      codec: h264
      rtsp_transport: tcp

      recorder:
        storage:
          tiers:
            - path: /recordings
              events:
                max_age:
                  days: 14
              continuous:
                max_age:
                  days: 14

There is no difference in the load on the system from how it worked previously.
In v2, Viseron saved small fragments in mp4 format, which were later concatenated together to create mp4 recordings.

In v3, these segments are stored as m4s fiels (fragmented mp4s), and with create_event_clip they are concatenated together the same way as v2, does it.

The difference tho is that the fragments are kept so double the storage is used.

One question, for what purpose do you want to store the full mp4 files?

Thanks for clarification. So basically, the mp4 fragmented files are still there and if the create_event_clip flag is set they will be combined into an mp4 as it used to be, right?

I am no expert in this but I have been living with mp4 and never heard of HLS until this post. Somehow, the new HLS files must have been created from those fragmented mp4?

Thanks for clarification. So basically, the mp4 fragmented files are still there and if the create_event_clip flag is set they will be combined into an mp4 as it used to be, right?

Correct!

I am no expert in this but I have been living with mp4 and never heard of HLS until this post. Somehow, the new HLS files must have been created from those fragmented mp4?

HLS is just a protocol for streaming and vods.
The underlying videos are still in mp4 format, however they are stored in small 5 second segments, and whn you request a recording the API constructs a playlist containing the relevant segments which forms the full video.
HLS is (one) standard in the streaming industry, you can read more about it here if you are interested: https://www.cloudflare.com/learning/video/what-is-http-live-streaming/

The main reasoning behind using it is to support the timeline with 24/7 recordings.
Would be impossible to store one large mp4 file for that!
It is also good for shorter videos since each segment is downloaded on its own which makes playback start faster

From the reading HLS default storage seems to be 6-second segments. I am just curious if that would be efficient in terms of storage space. My old practice which could be irrelevant these days is storing small fragment files would take more space than storing one large file (having the same total content).

Right now I am running viseron along with another rtsp server which keeps camera feeds stored in mp4 format around the clock, each having 20 minute length (configurable). If I get a spare pc I will set up aother viseron v3 to test run.

By the way, I have a few questions:

1. Where are the the mp4 segment of 5-second files stored? in memory or file storage.
2. Is it correct to say that motion detection including any other processing are performed on the latest 5-second segment generated?

Thank you for your work. I learn new things while setting up viseron.

Size should not be a concern, i just checked a recording that was around 30 seconds in length.
The total size of all segments is 2564 bytes while the event clip is 2560 bytes.
So there is only a 4 byte difference.

The 5 second segments are stored in the folder /segments by default.
This can be changed by using tiers in the new storage component which allows you to utilize several drives in a tiered manner.
If you want you can put them in memory as well.

Image processing is not done on the segments, Viseron saves the segments to disk while simultaneously reading frames at the appropriate interval to perform image processing.

Before there were 24/7 recordings there was no real use case to run without a detector, overlooked that!
Will fix soon in a coming beta.

Until then you can just enable a motion detector with a low fps value, it wont take much resources.

After fixing my other issue, I wanted to weigh in here and say I don't really care about motion detection, continuous recording is the only thing I need, so +1 for this feature request.

All of my motion detectors were already 1fps, so I just left them

Right, I have made some unrelated changes which should fix that, builds are on the way but I have to rebuild ffmpeg, opencv etc so it will take some time before that is ready to test!

Just release 3.0.0b4, builds should be done in a few minutes.
I think this issue should be fixed then, please test and let me know!

Thanks for the feedback!
That is indeed how it works, it was the easiest solution from a technical standpoint.
I do realize it is a suboptimal approach, will work on that for the next beta

Before I saw that discussion I opened a corresponding issue, as the same happens with segments for events:
#742

But knowing it's not a misconfig, this is probably something that can be ironed out.

v3.0.0b5 is now released which addresses this issue

Seems to be an issue with Docker and Postgresql incompatibility, see this thread: https://stackoverflow.com/questions/76555305/postgres-container-failed-to-start-with-initdb-error-popen-failure-cannot-allo

Can you try updating your docker version?

Worked! Thanks for the tip…
For future travelers, here are the steps I used:
dnf install podman
dnf install conmon - podman was complaining I had an old conmon, so I upgraded mine from 2.0.6 to 2.0.29
podman stop viseron
podman rmi viseron
systemctl restart viseron

1. The frame rate is copied from the source
2. Yes, by using video_filters you can set your desired frame rate. You also have to set fps to match
3. Easiest way is probably to let it store an hour or so and just make a manual calculation since it depends on a lot of variables

create_event_clip: true should be put under recorder for the camera.

If you paste your config i can help out

Heres what it should look like:

ffmpeg:
  camera:
    camera_1:
      name: Front
      host: xxx.xxx.xxx.xxx
      port: 554
      path: /cam/realmonitor?channel=1&subtype=0
      username: admin
      password: computer!
      recorder:
        idle_timeout: 5
        create_event_clip: true
      frame_timeout: 10

    camera_2:
      name: Front2
      host: xxx.xxx.xxx.xxx
      port: 554
      path: /cam/realmonitor?channel=1&subtype=0
      username: admin
      password: computer!
      recorder:
        idle_timeout: 5
        create_event_clip: true
      frame_timeout: 10
    
    camera_3: 
      name: Driveway
      host: xxx.xxx.xxx.xxx
      port: 554
      path: /cam/realmonitor?channel=1&subtype=0
      username: admin
      password: computer!
      recorder:
        idle_timeout: 5
        create_event_clip: true
      frame_timeout: 10


darknet:
  object_detector:
    cameras:
      camera_1:
        fps: 1
        scan_on_motion_only: false
        labels:
          - label: person
            confidence: 0.75
            trigger_recorder: true
          - label: car
            confidence: 0.75
            trigger_recorder: true

      camera_2:
        fps: 1
        scan_on_motion_only: false 
        labels:
          - label: person
            confidence: 0.75
            trigger_recorder: true
          - label: car
            confidence: 0.75
            trigger_recorder: true

      camera_3:
        fps: 1
        scan_on_motion_only: false
        labels:
          - label: person
            confidence: 0.75
            trigger_recorder: true
          - label: car
            confidence: 0.75
            trigger_recorder: true

nvr:
  camera_1:
  camera_2:
  camera_3:

mqtt:
  broker: xxx.xxx.xxx.xxx
  port: 1883

storage:
  recorder:
    tiers:
      - path: /recordings/video
        events:
          max_age:
            days: 30
  snapshots:
    tiers:
      - path: /recordings/pictures
        max_age:
          days: 30
        max_size:
          gb: 100

Cleanup runs every time a file is added, but at most every 10 seconds to not waste resources.
Your config should store continuous for 1 day OR 250gb, so it should never exceed 250gb or 1 day.

Might be a bug, will look into it

Do you by any chance have two cameras configured? The size limit for each tier is per camera, so if you have set 250 gb for max_size and have two cameras it will use up 500gb.

I could have sworn i wrote this in the docs but seems i have forgotten it.

I think that may be it then. I have 6 cameras and thought that the limit was for all the cameras not per camera. So that explains some things.

I think you can do what you want by clicking the blue filter icon at the bottom right. This should popup a menu with a camera icon. You can then click that and select a camera.

Yes that layout does not look great. Will see if i can figure something out.

If you are on iOS you need to be on at least 17.1 (https://webkit.org/blog/14735/webkit-features-in-safari-17-1/)

And yes, @gjong is right, you can choose a camera using the filter icon

Using Go2Rtc is no problem, I have a similar setup. Go2RTC exposes both RTSP and WebRTC to your network. You will have to setup Viseron like this for that:

ffmpeg:
  camera:
    oprit:
      name: Name
      host: go2rtc
      port: 8554
      path: /stream1
      frame_timeout: 10
      username: !secret stream_user
      password: !secret stream_password

Keep in mind that the host should be replaced with either the IP or the hostname of the computer that runs go2rtc.

Thank you that will give me a start