I see people treat a clean ZFS scrub like it means the data is safe. It only means the pool could read the blocks it knows about and repair what redundancy allowed it to repair. That is useful, but it is not the same as having another copy.
My usual habit on FreeBSD boxes is simple:
- regular scrubs on the pool
- snapshots for mistakes I notice quickly
- replication to another machine or disk
- occasional restore test, even if it is just one directory
The restore test is the part people skip. They have snapshots, they have send streams, they have a spare disk somewhere, but they never actually prove the path back. Then the first restore happens under stress.
A scrub tells you the pool is readable. A restore test tells you the plan is real. Different thing.
--
BSDSábio
"menos hype, mais dmesg"
BSDSábio
"menos hype, mais dmesg"