I run Stash in the standard container (stashapp/stash:latest) on a dedicated linux server.
I have now bought a new better server (with more storage ), and for the migration, I am considering moving to S6 based on what I have read here in Stash Discourse).
For the last 18 months, I have created a process when using the standard container, because I have no intention of letting in run as root. I am also paranoid about letting Stash have write access to my file collection. I am happy to share this process if anyone is interested. Basically, I have a non-production profile in Docker Compose, where I download the standard image, make changes to it, and then save it as a new image in a local registry. I then use that image to create my production container.
What I donāt understand, is why the standard image has so many limitations - and when considering S6 - it seems strange when feederbox826 is such a key part of this community. Why are there so many versions of container images?
Official image was made with ease of use at itās core, so that even people who have no experience with Docker can get it to run. It does lack rootless mode and hardware acceleration as a result.
stash-s6 is meant replace existing the official image at some point once all the to-do tasks are completed.
Itās an open source project, people are welcome to build their own images to fit their needs. Like I mentioned above, rootless mode and hardware acceleration are two reasons for it.
and when considering S6 - it seems strange when feederbox826 is such a key part of this community.
stash-s6 is a repository for staging, Iāve had previous attempts at making container images for stash and s6 allows me to be more volatile, notably with experimenting with uv and ffmpeg-jellyfin and for messing with migrations where I donāt have to worry about putting all the users of the base docker image at risk and instead focus on this smaller group that has more of an idea of the limitations and what might or might not break
Why are there so many versions of container images?
the simple answer is hardware acceleration, most variants of images (past and current) focused on adding hardware acceleration to the image since CUDA doesnāt work out of the box and requires a bit more tweaking and Intel VAAPI requires drivers (free and nonfree). When looking into s6 I also saw that python requirements were much in demand (and being talked about in the discord server at the time) so I took the leap to experiment with venvs, uv and other package managers.
), and for the migration, I am considering moving to S6 based on what I have read here in Stash Discourse).