[curtis3389]

Regardless of everything else, most people should not be using bcachefs yet. Kent has even stated that unless you're okay not being able to access your data for chunks of time while bugs are being fixed, you shouldn't be using it. The conventional wisdom would be to wait 10 years after a new filesystem is introduced for it to stabilize before switching, so we're looking at summer next year at the earliest.

[Filligree]

Apart from that, there are (or were, last I tried it six months ago) some performance bugs in the code.

Nothing that completely breaks it, but I found at the time that the high variance on read requests for Samsung 970 series NVMe causes the filesystem to also dispatch reads of cached data to the HDDs, even when it’s fully cached.

Which predictably increases latency a lot.

Really I should make another stab at fixing that, but the whole driver is C, and I’m not good at writing flawless C. Combine that with the problem actually being hard…

(“Always read from SSD” isn’t a valid solution here.)

[koverstreet]

"Always read from SSD" seems like what you'd want, no?

I have something on the back burner to start benchmarking devices at format time, that would let us start making better decisions in these situations.

[mixmastamyk]

Is there not some sort of standardized, stringent filesystem test yet? Like Jepsen is for databases? If passed, one can be sure it is reasonably free from bugs? Guess not.

[pantalaimon]

The thing is that filesystems are inherently statful, so the same test might trigger different edge cases depending on the state of the fs.