[bravetraveler]

Here to also suggest alternatives

BTRFS is the only 'RAID' that I've encountered that will almost predictably fail if forcefully powered off

RAID10 on gen4 NVMe drives, should sync pretty quick, you'd think.

LVM RAID, ZFS, mdadm, all of them are considerably more reliable

I can hear lamenting already: "that's not normal! you shouldn't expect consistency here!"

I tested this because reality says we do unusual things all of the time, don't hate me - hate the results.

Other implementations are demonstrably more robust

[rektide]

This is a you problem, that no one else reproduces, and which you have reported only here. It's degrading to the conversation to have this kind of casual slander.

Btrfs raid is stupid reliable, if you dont f it up & dont put metadata in raid5/6.

Having checksum consistency & checkpoints is a superpower most of the alternatives dont have. Btrfs is much more flexible about adding/removing drives, of various sizes even, to a raid pool, as compared to zfs, & is imo ridiculously easier to operate, from not needing out of tree modules to just requiring a lot less specialist knowledge in general of different zfs caches for tuning/setup. There's no alternative close & it runs great across millions of systems, with operators such as Facebook/Meta.

[bravetraveler]

Everyone in production reproduces this. Intentionally or not, everything on a long enough timeline will suddenly lose power.

I'm sorry to offend what is otherwise a decent filesystem, but I, and the systems I'm responsible for, will not use BTRFS with RAID and this characteristic.

There's a lot of qualification there, read it again.

I would much rather let the journal replay and do its thing, than rebuild the array. Something every alternative offers, consistently.

BTRFS RAID has routinely burned me every time I have given it the opportunity. It hasn't survived nearly as much as the other test arrays.

I'm fine discussing filesystems, benefits, and all of that - but I can't entertain it when you try to make this so heated.

These are anecdotes from reliability tests I've been doing... I'm sharing them so that they can be considered.

You accuse me, while opening like this? At least my anecdotes weren't insulting.

edit: To add, this isn't something I've kept secret. I don't know why you say that. Have you read every communication I've written?

I can show you screenshots that span the last three months discussing this exact situation

I haven't reported it upstream because I have better things to do; ship.

You might see this, but hopes aren't high. This is about the response I expected. It's an extreme edge case, but it consistently cuts.

The filesystem we use is incredibly uninteresting. By design.

I find it utterly hilarious you think I'm the only person who has had issues with BTRFS RAID