Originally posted by Redi44
View Post
Lets talk about data integrity. Linus Torvalds complained about one ext4 issue a few years ago and some hacks were put into place to reduce the pains caused by them, but the problem is still there:
ext4 journals only metadata by default, so if something does go wrong, you can lose data. Using fsck.ext4 to repair the filesystem will bring the filesystem into a consistent state, but it only does that. Being consistent doesn't mean that your data is still there. Running fsck.ext4 will do nothing to fix data corruption caused by a crash or even report it.
There are also other issues, such as a lack of checksumming. Recently, an extension was done for inode checksumming, which helps, but it only covers inodes and it doesn't provide the kernel with the general ability to fix things if the checksum fails:
There is no redundancy in ext4 at all. It relies on things like hardware RAID for that. This promotes situations where hardware is used that ignores barriers. ext4 relies on barriers to maintain consistency after a crash provided that fsck.ext4 is used to repair it. Ignoring barriers breaks whatever consistency guarantees ext4 was able to provide. That is not to say that other filesystems are immune (they are not), but ext4 requires other things to provide redundancy while ZFS does not.
I could also talk about performance and features, but if a filesystem can't keep your data safe by design, it isn't worth using.
Originally posted by psycho_driver
View Post
Leave a comment: