Two drives with the same capacity and the same price class can behave very differently the day a disk dies. The difference sits in how the tracks are written.
The short answer
CMR (conventional magnetic recording) writes tracks side by side. SMR (shingled magnetic recording) overlaps them like roof shingles to fit more data on a platter [3][4]. The catch is slow rewriting. Western Digital says that drive-managed SMR needs idle time for its housekeeping, and that sustained random writes, such as a ZFS resilver, take that idle time away [1]. Our assessment: for a NAS or a ZFS pool, buy CMR and read the data sheet before you buy. SMR is for archives you write once and read many times.
How it works
- Shingled tracks. The write head writes a wider track than the read head needs, so SMR lets new tracks overlap the previous ones and stores more per platter [3]. Seagate describes SMR as increasing areal density "through track overlay" [4].
- Three flavours. SNIA names drive-managed, host-aware and host-managed SMR [5]. Drive-managed SMR hides everything from the computer: the drive caches random writes and reorganises them later. Host-managed SMR is a different device type that the software must drive zone by zone [5]. Seagate says it needs control over zone management, write ordering and filesystem choice [4].
- Where it hurts. A drive-managed SMR disk works well for short bursts with idle time in between, which is what an office or home share looks like [1][5]. WD states that such a drive performs the same as a CMR drive, or slightly slower, when rebuilding a RAID, and that the drive condition and the extent of the rebuild also matter [1]. For ZFS, WD points to its CMR lines, because the sustained random writes of a resilver leave no idle time [1]. A resilver rewrites the data a replaced disk is missing [7], and that is the moment a weak drive shows up.
- You cannot ask the operating system. Linux reports drive-managed SMR disks as regular block devices: the
zonedattribute readsnone, the same as for CMR [6]. The reliable check is the manufacturer's data sheet or label [2]. - Brands and history. In April 2020 Western Digital stated that its WD Red models from 2 to 6 TB used device-managed SMR, while the 8 to 14 TB models were CMR [3]. Models change, so treat that as history and check the current data sheet of the exact model number.
- Where SMR makes sense. Seagate lists large media repositories, backup tiers, compliance archives and object storage as typical SMR workloads, meaning data that is written once and read occasionally [4].
A homelab example
We have not recorded the drive types in Bytearium systems (inventory as of 14 to 16 Oct 2026), so this check is general. List the model numbers of your disks, then look each one up on the maker's data sheet:
lsblk -d -o NAME,MODEL,SIZE
cat /sys/block/sda/queue/zonedA result of none does not tell you CMR from drive-managed SMR [6]. The model number and the data sheet do [2].
Checklist
- Read the data sheet or label of every drive for the recording method [2].
- Choose CMR drives for ZFS pools and RAID arrays [1].
- Do not mix an unknown replacement disk into a pool before you know its type.
- Treat a drive-managed SMR disk as an archive disk with long idle periods [1][4].
- Check the current model, not the memory of an older product generation [3].
- Keep a backup that does not depend on a rebuild.