Most of the useful pushback was not about whether the stack works. It was about what a real NAS has to do after day one. Several people said the guide is fine as a build recipe but incomplete as an operations guide. A NAS needs disk failure alerts, a clean way to identify the bad drive, a documented replacement flow, and recovery steps you have actually tested. That landed harder than the usual filesystem bikeshedding because multiple comments described the same practical failure mode. Backups existed, but restore details or keys did not.
The technical discussion centered on which
ZFS rules still deserve to be treated as rules. The strongest consensus was that ZFS does not uniquely require
ECC RAM. ECC is better, but the old claim that non-ECC plus ZFS is especially dangerous was called out as folklore from a notorious forum post, not a property of ZFS itself. People made the same distinction on memory sizing and free-space advice. The old "ZFS needs tons of RAM" line got little support for home use, and the 20 percent free-space rule was reframed as a fragmentation and write-performance issue that depends heavily on workload, not a universal cliff.
There was much less tolerance for shaky hardware choices. USB storage enclosures and cheap adapters were repeatedly described as the real footgun, especially under
scrub or sustained load. Used enterprise gear got more respect than consumer gear in this thread.
SAS drives,
LSI host bus adapters, old workstations, MicroServers, and
Proxmox hosts all came up as sensible ways to get reliability without buying a closed appliance. Consumer SSDs in ZFS-heavy roles drew more warnings than non-ECC RAM did.
A second thread running through the comments was that many home setups do not actually need ZFS’s full model. If the goal is cold storage, replaceable media, or a big pile of non-critical files, people pointed to
SnapRAID plus
mergerfs,
Unraid, or even simpler single-disk setups with external backups. The dividing line was not ideology. It was whether you need checksums, snapshots, send and receive, and predictable rebuild behavior badly enough to accept the extra operational complexity.
The mood was practical and broadly pro-DIY. People like owning the stack and avoiding appliance lock-in. They were also blunt that a NAS is not defined by having disks on a network. It is defined by boring failure handling, integrity checks, and restores that work when you are tired and a drive just died.