I’m running on 0.19.3 without any issues on Linux arm64. I built my own docker image though.

The web is built on hot linking hypermedia. It is more fragile obviously, but it distributes the bandwidth and storage load. If nobody hotlinked, then small forum admins/Lemmy admins/etc. have considerably more cost to bear.

Yes, there are a few issues in the lemmy-ui including this PR with a temp fix: https://github.com/LemmyNet/lemmy-ui/pull/2058

Haha yes, that configuration flag PR is mine

100% agreed. I don’t have the time to make a change that complex right now, so I did a fairly blunt approach with the hope that larger instances will keep caching on to reduce load.

Agreed, I sadly don’t have the time to implement that.

I don’t think anything in lemmy is currently clearing that. There are community scripts around that do some clearing but I have not tried them.

go to https://campfyre.nickwebster.dev, sort by “new”

Thumbnails still seem to work.

That’s why I made it a config option that defaults to true (defaults to caching on).

I think big instances should cache, but for smaller instances with less funding and resources it makes sense to skip the caching.

Yep! There’s a pretty rapid growth of pictrs data that’s never going to go away from all the images being cached for thumbnails on my instance.

It’s starting to get to ~1GB per week at this point.

I pulled this PR into the version of lemmy I run on my instance and it looks promising.

Oh wow, this is tragic

That was my understanding too but I just looked through the storage bucket I have backing pict-rs on my instance and there is loads of stuff that is not mine on there.

I made a kubernetes deployment for my lemmy service + use object storage for image hosting. Everything lemmy-side looks like it should scale fine. I’m not doing open registrations though, so it won’t impact me. The key bottleneck in my case is the database, but if need be a larger node can be provisioned to let the DB expand a bit.

I think for larger instances some form of server-abstraction will be useful for scaling (i.e. k8s, cloud run, EKS, etc.)

Might be taking some time to backfill, I can pull down posts when I explicitly go to the community (see edit 2), but no comments are showing up as of now https://campfyre.nickwebster.dev/post/21