It looks like if I mount local host folder for clips, I don’t get the database locked error but if I use network shared folder, I get the error even though frigate can write the clips (version 0.7.0) on the network shared drive.
Local folder without error:
drwxrwxrwx 2 user user 4096 Dec 9 15:34 beta
drwxr-xr-x 2 root root 4096 Dec 9 19:51 clips
drwxrwxrwx 2 user user 4096 Dec 9 11:10 config
drwxrwxrwx 2 root root 4096 Dec 9 11:10 config7
drwxr-xr-x 2 root root 4096 Dec 9 19:50 recordings
root@hp:/home/user/data/frigate# cd clips/
root@hp:/home/user/data/frigate/clips# ls -l
total 20
-rw-r--r-- 1 root root 20480 Dec 9 19:51 frigate.db
Network folder with database locked error but version 0.7.0 can write clips
drwxr-xr-x 2 root root 0 Dec 9 19:07 clips
-rwxr-xr-x 1 root root 0 Dec 9 19:03 frigate.db
drwxr-xr-x 2 root root 0 Dec 9 19:01 recordings
drwxr-xr-x 2 root root 0 Dec 9 06:11 snapshot
drwxr-xr-x 2 root root 0 Dec 9 19:19 webcam
I’m running proxmox with Hassio VM + LXC Container for Frigate.
Currently using Hassio Samba addon to share media directory and Frigate LXC container mount it using fstab.
Now i’m trying to use NFS to share the clips folder from LXC to Hassio VM. I’m getting connection refused while trying to mount the directory from Hassio VM.
I’m thinking to abandon my proxmox setup and go for debian supervised instead. Less headache running 2 separate VM/Container
If it helps, I’m running home assistant container in docker on bare metal Ubuntu, in the same stack with frigate. Not sure it should matter, but the smb share is hosted by an Ubuntu VM on a separate proxmox server running next to the home assistant server.
Ok, installed the Frigate NVR Beta 1.0 in HA. Still don’t see where I can add my actual Frigate host that is running on the NUC. On HA in Frigate NVR Beta -> Configuration I can see where I can change ports for the RTMP streams and Web Interface.
Am I missing a tutorial or something from somewhere? I feel like I’ve having a hard time getting this figured out for some reason and I’m no slouch when it comes to software.
You can continue to drill down until you see an “All” folder. That folder should have a limit of 10,000. The limit of 50 is by design. HomeAssistant does not support lazy loading of items nor does it offer caching of thumbnails. Over slow/mobile networks, showing lots of items means you have to load every thumbnail with each click and it ends up being way too slow. My assumption was that most of the time you are looking for what happened recently with a few quick filters, but that you can still get to everything if needed.
So it looks like sqlite databases aren’t reliable when run from a network share. I can add a configuration option for the database to allow you to store it elsewhere. This database does have all of the metadata for clips, so if it is lost, HomeAssistant won’t show them in the media browser anymore.
I wasn’t aware of the “all” folder. I was actually expecting something like a page by page browsing, but I’m not that familiar with the media browser.
But yes, it makes sense to limit the number reduce the i/o.
Anyways, selecting the "All, is in my case to much for the PI to handle. That is to say it eventually gets there, but it takes a few minutes (which looks like a freeze, but isn’t).
After looking at doods I thought the USB Coral Edge TPU on my j4105 board just wasn’t capable of doing the job with my 3 Foscam + 1 reolink HighRes cameras. Detection rate was rather bad, from comments I thought the TensorFlow Lite models were the root cause. Frigate just worked (Software was key?) for >3 month now & WAF is high due to cat door tracking
Your latest Beta rocks. 24x7 recoding, image cropping, Media Browser, lower CPU usage, HA Integration, Dokumentation style - just what i was looking for and didn’t find in Milestone, Shinobi, iSpy Agent DVR, Zoneminder, …
The professional, unagitated way you appear (to me) in this forum. Great style!
I’m a (scotish, sorry) sponsor of yours now - hope to vanish in the crowd soon (come on guys - we could be more…)
Thanks!
Alex
PS: Sorry, but:
+1 for the ‘turn of detection when person at home’
Would love to have /media/frigate/[clips|recordings] on an NFS / SMB share instead of the docker host. I do that with a number of containers, currently frigate doesn’t seem to like it. Even with clips + db locally recodings don’t play out in media browser
Same here for the thought that tensorflow lite is giving less than stellar results. Also using Doods and it works well, but using the heavier models.
This is why I never bought a coral. I thought it was limited to the lite version.
Does frigate do more than tensorflow lite? Or is that not an option? Looks interesting and I have got some time off coming up. Maybe I’ll try it just to try it and see.
@blakeblackshear I’ve started having a weird issue with the HA addon.
2 of my cameras are bouncing back and forth with each other when I view the feed.
One called hot_tub with a feed from rtsp://xyz:[email protected]:554/h264Preview_05_sub
The other called pool with a feed from rtsp://xyz:[email protected]:554/h264Preview_06_sub
When I go to 192.168.XX.XX:5000/hot_tub I see the feed from hot_tub, then it jumps to the feed from pool, then back to hot_tub and it flips back and forth. Same happens when I view the feed from pool. My side_yard, side_door, back_door cameras are all working fine.
When I view the individual rtsp feeds there are no issues. I’ve rebooted the camera feeds, frigate, HA. Renamed the cameras. All was working fine (or at least it was last time I looked at the frigate feed). This morning I added a poly mask for the pool camera and that’s when I noticed this strange issue.
Looking at the startup log you can see it’s starting 2 processes with different feeds for pool
Camera process started for hot_tub: 20
Camera capture process started for pool: 21Creating ffmpeg process...
ffmpeg -hide_banner -loglevel panic -hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -hwaccel_output_format yuv420p -avoid_negative_ts make_zero -fflags nobuffer -flags low_delay -strict experimental -fflags +genpts+discardcorrupt -rtsp_transport tcp -stimeout 5000000 -use_wallclock_as_timestamps 1 -i rtsp://xyz:[email protected]:554/h264Preview_05_sub -r 5 -f rawvideo -pix_fmt yuv420p pipe:
Camera process started for pool: 25Creating ffmpeg process...
ffmpeg -hide_banner -loglevel panic -hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -hwaccel_output_format yuv420p -avoid_negative_ts make_zero -fflags nobuffer -flags low_delay -strict experimental -fflags +genpts+discardcorrupt -rtsp_transport tcp -stimeout 5000000 -use_wallclock_as_timestamps 1 -i rtsp://xyz:[email protected]:554/h264Preview_06_sub -r 5 -f rawvideo -pix_fmt yuv420p pipe:
EDIT: I uninstalled/re-installed and same thing. Then I moved to the beta and now ALL my camera’s are doing the same thing. If I re-order the order of the cameras in the config it changes which camera it jumps back and forth with.
EDIT2: I was able to get the cameras to stop bouncing. I changed all the feeds to the 4k feed, restarted frigate, then changed them back to the substream and that seems to have fixed it. But my log is still loaded with the error above.
Sorry, I don’t use HASS. Frigate runs on a Docker system on top of ubuntu.
I NFS to a Synology system, using a simple mount command after modifying the /etc/fstab file.
