Good day all,
You may notice that some images posted after the 23 August are missing. Sorry about that.
What happened is that I updated some packages through apt, which triggered docker to restart. This caused the pictrs database to become corrupted, but only partly because of the way its db (sled) does things.
I did manage to recover about half of the images through abusing image proxies and browser cache, so there’s that at least. The whole of the data is also theoretically recoverable.
Needless to say, I’m going to move pictrs to postgres soon-ish.
Ah, I thought that was just magical internet issues
Thanks for the insight, but more importantly - thanks for looking after this instance
Is the plan to recover the rest of the data? Just wondering if I should reupload some of my older posts’ images.
Thank you for your work keeping the instance going and being open about this sort of stuff.
I’ll try to, but I doubt I’ll be able to (I already tried for a few hours before posting this) so reuploading is probably the better call.
I must ask the big B question!
Backups!
Do we have them?!? (How bis is the total DB?)
The pictrs db is at about 12 GB on disk, with < 300 GB of image data. It’s not safe to back it up while it’s running however, so it only gets done every update (unlike the main db that I try to keep at least weekly backups of). This is part of the motivation of moving pictrs to postgres, as that can safely be backed up while running.