|
I have mostly good experiences with it. I have more examples of surprising good things it does and scenarios it will survive OK. But then the edge cases, if you hit them, can be brutal. So you have to amp up the backup strategy for important data, just because of the unknown factor. Development is really incredible the 4.2 merge window included almost 2000 additions for Btrfs, and almost as many subtractions. It's really hard for a mortal person to keep track of the bugs, bug fixes, optimizations and regressions. For example right now, a rather major regression in ext4 conversions: The convert goes find, mount is fine, rollback is fine. But if you scrub it, hard panic; even console is lost, let alone all services. But the fs survives. If you balance it, corruption of the fs (and so far not repairable). So right now, which could change at any moment, I'd say avoid ext4 to btrfs conversions. Create a new fs, and keep backups. Usually it's pretty good about at least mounting read only so you can update a backup, even if the fs is beyond repair. Despite piles of kernel features, it still lacks much needed notification to user space of device failures. For multiple device volumes, you probably want to be informed of either flakey or failed devices and right now you have to be watching for it. Some edge cases in degraded operation exists when replacing devices that you have to be careful of, etc. So this whole area needs work, which means more developers interested in such things are needed. But anyway, it's going pretty good. |