During the first window I'll be migrating us from OVH to our new dedicated hardware. After this migration there will likely be some temporarily broken images, as it takes approximately 8 hours to resync our object storage from OVH.

This is a major change and despite my testing, may have some unintended side effects. If you run into any problems that aren't just a broken image, please let us know.

The second maintenance window is to migrate our pict-rs database from it's local sled-db into our primary postgres DB. This is a much smaller change but since pict-rs checks every image as it goes through them, it takes about 1.5 hours.

As usual, you can check https://status.lemmy.ca/ for updates.

you are viewing a single comment's thread
view the rest of the comments

[–] Shadow@lemmy.ca 2 points 6 days ago (1 children)

I think this is fixed now. I'm not sure why but pictrs was expecting a bunch of resized images that don't exist in either the old or new object stores. I've manually cleaned up the DB.

[–] Darkassassin07@lemmy.ca 2 points 6 days ago* (last edited 6 days ago) (2 children)

Many of them are, but scrolling through this thread and checking the profiles of various lemmy.ca accounts; some are still broken.

@sunshine@lemmy.ca for example

[–] Shadow@lemmy.ca 1 points 6 days ago (1 children)

I've hacked in a dirty dirty fix for now. It looks like the pictrs database migration failed horribly and cloudflare was caching the images, so nobody noticed. Now I've got two copies of pictrs running (old + new) and some code in place to try both of them.

Sigh.

[–] Darkassassin07@lemmy.ca 1 points 6 days ago

Oh shit.... Well at least you found it before everything just stopped working? :(

Thanks for your efforts m8; I appreciate the hard work. Best of luck fixing that.

[–] Shadow@lemmy.ca 1 points 6 days ago

Well shit, cloudflare's been caching all our images and deceiving me into thinking stuff was working properly. Digging, but I made a mess =(