Installed seafile and all its dependencies including moviepy via pip installer. Launching the server it works perfectly all images have thumbnails but not videos. Specifically my mp4 videos. I have enabled it using this within seahub_settings.py using the commands at the bottom. My seahub.log shows this error. “2021-04-02 03:21:33,318 [ERROR] seahub.thumbnail.utils:34 Could not find moviepy installed.”
ENABLE_VIDEO_THUMBNAIL = True
THUMBNAIL_VIDEO_FRAME_TIME = 5
THUMBNAIL_ROOT = ‘/home/pi/Server/Server/seahub-data/thumbnail/vthumb/’
THUMBNAIL_SIZE_FOR_ORIGINAL = 1024
Edit: I’m using 8.0.3 on the rpi 4 8gb.
The VIDEO_THUMBNAIL feature is no longer supported.
I know that it is not supported for some time. But why was it removed in the first place? It was working fine, at least on systems I maintain.
I also find that it is a very interesting function, and essential, there is nothing more unpleasant not to see on what you click or which video file you wish to open.
We remove the feature because it is very hard to make it work in all environment. In some systems, if you install moviepy, Seahub will consume a lot of CPU without any user activity.
In a large deployment, generating thumbnails for videos will consume a lot of memory and CPU and make the system unstable, so we don’t recommend to use this feature.
A workaround is that using a dedicated machine for generating video thumbnails, but this requires a lot of effort for system admins. Compared to the additional benefit, it seems not worth to do it.
Thanks for explaining.
Why not continue to provide this feature and write a warning in the manual about the CPU/memory issues?
There are people willing to run a background node or doing it anyways. It would then be required to enable this only on the background node. Other nodes still need to be aware that there are video thumbnails though. Maybe add an option that enables thumbnail generation and another option that enables video thumbnails in general, allowing only to read existing thumbnails.
It would also be okay to only support this feature on amd64 platform and not on others (officially) I think. This feature might burn e.g. a RPI CPU, even though that recent v4 got plenty.