Bcachefs Sends In More Fixes For Linux 6.9: Recovery & Repair Issues Settling Down

[phoronix]

Phoronix: Bcachefs Sends In More Fixes For Linux 6.9: Recovery & Repair Issues Settling Down

Another week, another round of Bcachefs file-system fixes for the mainline Linux kernel...

https://www.phoronix.com/news/Bcachefs-Yet-More-Linux-6.9

[andyprough]

Something I didn't previously know - Torvalds is listed as having the second most number of commits behind Overstreet, with about 3X more commits than anyone else. And Greg KH has the sixth most commits. Kind of lends some extra legitimacy, to see that kernel hierearchy are also invested in the file system's success. https://github.com/koverstreet/bcach...s/contributors

[timofonic]

Nice! I see this as good progress. The more fixing, the more robust will be

[dralley]

Something I didn't previously know - Torvalds is listed as having the second most number of commits behind Overstreet, with about 3X more commits than anyone else. And Greg KH has the sixth most commits. Kind of lends some extra legitimacy, to see that kernel hierearchy are also invested in the file system's success. https://github.com/koverstreet/bcach...s/contributors

I'm pretty sure that only considers commits that are easily tied to a specific github account. There's lots of kernel devs that don't have a github account at all, or use a different email for kernel development than for other contributions.

But Kent has definitely been a prolific kernel developer for many years.

[swastika]

Something I didn't previously know - Torvalds is listed as having the second most number of commits behind Overstreet, with about 3X more commits than anyone else. And Greg KH has the sixth most commits. Kind of lends some extra legitimacy, to see that kernel hierearchy are also invested in the file system's success. https://github.com/koverstreet/bcach...s/contributors

Don't use GitHub for these stats. Kernel development does not use GitHub and it both overcounts and undercounts commits because of how it does it. You should read the Linux Foundation reports or LWN if you want more accurate data on that.

[varikonniemi]

Gotta love bcachefs still being experimental, while i have been running it on RAID with only one hiccup since first release, the on disk format upgrade mess. But also that bug was gracefully handled automatically, only found out about it by checking kernel logs.

I guess this level of stability is a result of the root algorithms being tested and true from bcache and bcachefs is just the plumbing around it to make it a filesystem.

[Developer12]

Gotta love bcachefs still being experimental, while i have been running it on RAID with only one hiccup since first release, the on disk format upgrade mess. But also that bug was gracefully handled automatically, only found out about it by checking kernel logs.

I guess this level of stability is a result of the root algorithms being tested and true from bcache and bcachefs is just the plumbing around it to make it a filesystem.

At this stage this is mostly dumb luck. Maybe it's well written, or maybe there'll be another big bug that'll bite people every 6 months for the next 3-4 years. We just don't know yet. It was very lucky in the first place that this bug tended to leave enough data on disk to attempt a repair.

I wish the bcachefs people the best of luck and godspeed, but it's really not worth getting hung up on one way or the other for at least a few years. As overstreet said, they've barely even started ramping the testing and fuzzing, and there aren't a lot of bcachefs users yet.

[varikonniemi]

At this stage this is mostly dumb luck. Maybe it's well written, or maybe there'll be another big bug that'll bite people every 6 months for the next 3-4 years. We just don't know yet. It was very lucky in the first place that this bug tended to leave enough data on disk to attempt a repair.

I wish the bcachefs people the best of luck and godspeed, but it's really not worth getting hung up on one way or the other for at least a few years. As overstreet said, they've barely even started ramping the testing and fuzzing, and there aren't a lot of bcachefs users yet.

I partially agree, but to be fair he also indirectly said that it's not expected to have data eating bugs, just unavailability (due to the details of the implementation, the data will be found)

[janos666]

"but has shown much promise around its features and functionality"
How about online defragmentation...? It's mainly for HDD with flash cache, so defragmenting the HDD space is not negligible in my opinion.

[Developer12]

I partially agree, but to be fair he also indirectly said that it's not expected to have data eating bugs, just unavailability (due to the details of the implementation, the data will be found)

The data being found is contingent on the data still being there. ZFS for example is capable of rolling back the filesystem by rolling back recent transactions, but past a certain point the old transaction record and the data referred to at that point in time isn't there anymore.