Hacker News Reader

Somewhat depends. Scrubbing is about detecting bit rot. You need to make sure that all of your data is periodically accessed so that zfs can check for errors. You can accomplish this either by having a workload which touches all of your data on a regular basis or by scrubbing. Depends on your setup.

View on HN

Scrubbing is still preferable though, as normal reads won't cause every copy of the metadata to be read. Likewise, on a mirrored pair of disks, a given block of file content will only be read from one disk. (That is, since the file content is verified using checksums contained in the metadata, we don't need to read the file data back from both disks, so they can be used to service read requests independently of one another.)

All good points, especially about mirrors.

I added `zpool scrub tank` to cron once a week.

Reply on news.ycombinator.com