DM-SMR drives can turn sustained ZFS writes, scrubs, and resilvers into long stalls. Here is how shingled recording behaves, how to identify a drive before buying it, and how to separate an SMR bottleneck from cabling, controller, firmware, or media problems.

What Shingled Magnetic Recording Actually Does

Shingled magnetic recording (SMR) overlaps tracks like roof shingles. The write head is wider than the final track, so writing a new track partially overwrites the adjacent one. Reading remains straightforward, but writing becomes destructive: to update one shingled track, the drive must read all overlapping downstream tracks, rewrite them, and then write the new data. This is invisible to the host.

Drive-managed SMR (DM-SMR) hides zone management behind firmware and may use cache or conventional zones to absorb writes. Under light, bursty workloads it can resemble a CMR drive. During sustained or fragmented rewriting, background media management can sharply reduce throughput and increase latency. The exact cache design and slowdown are model- and workload-dependent.

Why DM-SMR Can Be a Poor Fit for ZFS

ZFS is copy-on-write: modified blocks are written to new locations, and transaction groups periodically sync dirty data. As pools fill and free space fragments, partial rewrites can become less sequential. OpenZFS explains those costs in its copy-on-write documentation.

That workload can interact badly with a DM-SMR drive’s hidden zone reorganization. Long-latency commands can reduce pool throughput and, on some hardware and timeout configurations, contribute to I/O errors or a device being faulted. This is a performance and operational-risk mismatch, not a universal filesystem incompatibility: results depend on the drive firmware, pool layout and fullness, workload, controller, and operating-system timeouts.

Scrubs and resilvers are I/O-intensive. A scrub reads all allocated data to verify checksums; a resilver reconstructs data known to be out of date. Their duration depends on allocated data, drive behavior, concurrent work, pool topology, and errors, so no single CMR or SMR rebuild duration applies without a defined test setup. Redundancy risk increases while a degraded vdev is rebuilding, but a second timeout is not automatically synonymous with losing the pool; the outcome depends on topology and which devices or data become unavailable.

The WD Red SMR Scandal and Other Offenders

In 2020, Western Digital’s use of DM-SMR in several WD Red NAS models—including WD20EFAX, WD30EFAX, WD40EFAX, and WD60EFAX—drew criticism because the recording method had not been clear to buyers and some sustained RAID/ZFS workloads behaved poorly. Western Digital subsequently separated its NAS branding: its current WD Red recording-technology page describes WD Red as SMR and WD Red Plus/Pro as CMR. Treat historical suffix rules as clues only; verify the exact current model.

SMR also appears in some desktop drive families. Seagate’s current list, for example, identifies particular BarraCuda capacities as SMR while others are CMR. The lesson is that branding, spindle speed, and cache size do not determine recording technology; check the exact model and current data sheet.

How to Check Whether a Drive Is SMR Before You Buy

Manufacturer spec sheets

Sometimes the recording method is listed in fine print. WD’s product page may state “Recording Technology: CMR.” Seagate mentions it in datasheet PDFs. Toshiba N300 explicitly states CMR. If the spec sheet is silent, ask the manufacturer or choose a model whose recording technology is explicitly documented; a missing label is uncertainty, not proof of SMR.

Community-maintained databases

Community-maintained lists can help find reports, but treat them as a lead rather than final authority. Confirm the exact model and revision against a current manufacturer data sheet whenever possible. Western Digital publishes a CMR/SMR identification guide, and Seagate publishes a current CMR/SMR product-family list.

Use current model tables, not a copied shopping list

Drive lineups change faster than this article. Check the exact model number on the listing and drive label against a current manufacturer source. Western Digital’s current WD Red Plus data sheet lists recording technology per model, and Seagate’s CMR/SMR list breaks it out by family and capacity. A static table copied here would eventually become another trap.

If You Already Have SMR Drives in a ZFS Pool

First identify the exact device model:

ls -l /dev/disk/by-id/ | grep -E 'ata-|scsi-' | grep -v part
smartctl -a /dev/sdX | grep -E 'Device Model|Product'

Confirm its recording technology against a manufacturer data sheet. Then use zpool status, kernel logs, SMART data and zpool iostat -v to distinguish a slow workload from link resets, controller errors, or failing media. A long scrub alone does not prove the drive is SMR.

If a DM-SMR drive is the bottleneck, reduce avoidable concurrent writes and keep healthy free space while planning replacement, but do not reduce or skip integrity checks merely to hide the slowdown. OpenZFS’s scrub documentation explains that scrubs find latent corruption while redundancy can still repair it. Monitor a running scrub and schedule it for a low-activity window; do not abandon it as a generic SMR workaround.

There is no universal “safe only as a single-disk ZFS pool” rule. A single-disk pool removes the performance interaction with a stripe but also removes repair capability: ZFS can detect corrupted data without redundancy but cannot reconstruct it. DM-SMR is better reserved for workloads the manufacturer supports and where unpredictable sustained-write latency is acceptable. For an important ZFS array, migrating to documented CMR drives is the conservative choice.

What to Buy Instead: CMR Drive Recommendations for ZFS

Prefer a model whose current manufacturer data sheet explicitly states CMR and whose workload rating, interface, warranty, noise, power, and error-recovery behavior fit your system. Western Digital currently documents WD Red Plus models as CMR, and Seagate’s current recording-technology list documents IronWolf and IronWolf Pro capacities as CMR; verify again when buying because lineups change.

Used enterprise SAS drives can be an option, but verify the exact model, interface, sector format, remaining warranty, SMART history, and recording technology. SAS drives require a compatible HBA and may be louder or draw more power than a home-oriented SATA model. Prices and power figures vary too much by model and seller for a fixed rule.

Even within these lines, verify model numbers before purchase. A quick search for “ CMR SMR” before buying is cheap insurance against a multi-day resilver.

The practical rule is to prefer explicitly documented CMR drives for ZFS pools that must sustain writes or rebuild predictably. DM-SMR can suffer severe and workload-dependent slowdowns, but it is not correct to claim that every SMR drive always breaks ZFS or that one timing number applies to every pool. If an existing device behaves badly under representative load, plan a measured migration rather than relying on a tuning switch to change its recording physics. Check model numbers, consult the databases, and if you have SMR drives now, plan their migration. The time spent verifying recording technology is nothing compared to the time lost waiting for an SMR drive to finish a routine scrub.