2 ms·
it's even worse: there's no guarantee that if you don't touch something on the filesystem it will not be corrupted. It used to be almost the case with HDDs but
by toolslive 5y ago
it's even worse: there's no guarantee that if you don't touch something on the filesystem it will not be corrupted. It used to be almost the case with HDDs but you're using SSD and NVMe devices. Basically, it can happen that when there's a sudden power off, they will do weird stuff like "set the 6th bit of every byte in an entire region". In essence, whatever your strategy for making/keeping a summary of the source side of your backup, you will always need to be able to figure out the state/diff from scratch. This needs to be work. Everything else is merely there to speed things up.
- jiggawatts 5y agoChecksums are reasonably robust against this kind of thing, but there are storage systems out there (most SSDs!) that can even reorder blocks! So you can have valid checksums but still get corrupt data…
- toolslive 5y ago| So you can have valid checksums but still get corrupt data… Yup, therefore: | you will always need to be able to figure out the state/diff from scratch. So if you evaluate a backup system, this needs to be the first thing to check. "what if I backup and accidentally lose my log/summary/sync-state ?"