Set OpenZFS manual TRIM rates per device rather than per pool

Calculate the impact of a manual TRIM limit across participating leaf devices.

Source facts

OpenZFS 2.3 manual TRIM reports unallocated pool blocks to underlying devices, allowing supported thin provisioning to reclaim space. It can run independently of the autotrim setting. Its rate limit is expressed in bytes per second and applies per leaf virtual device, rather than as one combined pool-wide ceiling. Without that limit, manual TRIM runs as quickly as possible. The command supports suspension, resumption, cancellation, and waiting for completion as separate controls.

Source: OpenZFS 2.3 — zpool trim.

Applicability

This applies to the OpenZFS 2.3 manual TRIM command on eligible devices. It does not claim that every storage backend honors TRIM or that routine TRIM is a media-sanitization certificate.

DSE recommendation

DSE recommends checking device and storage-backend support before scheduling manual TRIM. Set the expected workload impact and aggregate activity budget with the number of participating leaf devices in mind. Record whether the action is a one-off maintenance task or part of a monitored schedule. Choose a supervised window for the first run and define who can suspend it if service latency changes. Treat secure-erase requirements as a separate reviewed retirement procedure, rather than an assumed consequence of routine space reclamation.

Verification

On representative test storage, observe per-device activity and application latency with the proposed rate. Exercise suspension and resumption, then retain completion status. Confirm the monitoring distinguishes host free space from any capacity reclaimed by the underlying thin-provisioned storage.

Official reference

OpenZFS 2.3 — zpool trim. Reviewed October 15, 2026.