3 ms·
I had multiple corruption on a "server" at my parents, checked the disk and reinstalled from scratch 2 times, then I ran memtest86+ and found out one of the mem
by champtar 6y ago
I had multiple corruption on a "server" at my parents, checked the disk and reinstalled from scratch 2 times, then I ran memtest86+ and found out one of the memory stick was faulty. Btrfs allowed me to detect that and be confident that all data recovered from the disks were ok because of the data checksumming. With ext4 I might have never found out.
- knorker 6y agoI wouldn't count on that though. I'd say you were lucky. If the corruption happens before it gets sent to the disk, then you have checksummed and valid but broken data.
- diegocg 6y agoBtrfs has added some checks to prevent these issues. It can't catch everything, but some people have apparently hit them so it certainly helps https://btrfs.wiki.kernel.org/index.php/Tree-checker https://btrfs.wiki.kernel.org/index.php/Tree-checker
- tremon 6y agoThe parent's point was that if the user data (not the tree metadata) was corrupted in-memory before writing to disk, no amount of checksumming will save you because the checksum was calculated over corrupted data.