Comment by adastra22
2 days ago
The oversimplified answer is that ZFS’ in-memory structures are not designed to minimize bitflip risk, as some file systems are. Content is hashed when written to memory cache, but it can be a long time before it then gets to disk. Very little validation is done at that point to protect against writing bad data.
> it can be a long time before it then gets to disk
Well by default the timeout is 5 seconds[1], so not that long.
[1]: https://openzfs.github.io/openzfs-docs/Performance%20and%20T...
That is a long time from this perspective.
There's also a max size, so if you have heavy write traffic it'll flush do disk much earlier. The timeout is just for low-write scenarios.
What file systems offer in-memory protection?