Post

Good morning! We wanted to update you all and say we are 48% of the way through the copy process. Thank you for all your patience and support!

91 reposts639 likes

It came back on hours ago, I understand that they're still working on the site, I'd was already dealing with having to take my computer to the Shop after I'd crashed it Last weekend long story, managed to get it fixed.

Have you considered switching your storage backend from a single server to a distributed file system? Ceph can export namespaces over NFSv4 and it seems like it might be more maintainable.

The problem they had was an overflow in the count of files, (i.e. a filesystem had more files and files that had been deleted but their ID number not reclaimed than the system could handle as a number), making parts of the system think there were less than 0 files.

Once this happens, you can't simply go through and reclaim all those unused ID numbers. The good news is that the fix is simple: Create a new, clean, filesystem to put everything on and move the files there. The bad news is that FA has a _LOT_ of data to move when doing this.

Summarizing and possibly not explained usper well: They're moving every piece uploaded to the site into a new storage partition. So to avoid things getting messy and another error of "hold up, the data supposed to be here and the data being received don't match" they're keeping uploads off for now.

Yeah, FurAffinity is such a good website, I hope it will soon come back with 100% of his functionality. I also hope devs would have a public API for getting posts information

I wish this didn't sound accusatory, but I'm genuinely curious on the mechanics for the copy. Like, ext4 on a PCIe 5 x16 four way RAID 0 over 4 NVMe drives would take hardly any time. BTRFS RAID-1 with 4x consensus on 4x20TB+ HDDs across a gigabit link mounted over NFS could take ages.