DragonFlyBSD

Environment:
- a storage server with a 20TB /data volume
- a dataset potentially deduplication-friendly (protein sequences)
- deduplication has been enabled with a 5mn per day timelimit for /data

Problem:
- hammer(8) is running most of the day, inducing a high background I/O load on the disks
- the particular command is "hammer -t 300 dedup"

That command is run from the nightly hammer cleanup cron. It starts at 3am an never finishes.

A quick test on a brand new hammer volume confirms the -t argument to hammer dedup has no effect.
hammer dedup will run forever as long as there is still data to deduplicate.