Skip to main content

My Experience with NCQ Errors

· 3 min read
Milioto Pietro
Milioto Pietro
Sys Admin

What started as a normal day in my homelab turned into hours of debugging, unexpected freezes, and a long chain of NCQ timeout errors coming from my Seagate Exos 12TB drive. I swapped cables, changed power supplies, rebooted the system dozens of times — nothing seemed to fix it. Eventually, I learned the hard way how sensitive enterprise HDDs can be to power stability and controller behavior.

The Beginning of the Problem​

The issue began when I noticed strange freezes while accessing my 12TB Seagate Exos (SATA, model Exos X18/X20, 7200 RPM).
The logs were full of lines like:

ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
ata1.00: failed command: READ FPDMA QUEUED
ata1.00: cmd 60/08

This is the infamous NCQ (Native Command Queuing) error, a sign that the drive was stuck executing a queued command.

Sometimes the disk would disappear for a second, sometimes the entire OS froze for minutes.
It looked like either:

  • the drive was failing,
  • the controller was unstable,
  • or power delivery wasn’t sufficient.

At the beginning, I obviously feared the worst: a dying HDD.

My First Attempts (and Failures)​

Like anyone who works with Linux storage long enough, I tried all the usual steps:

  • Changing the SATA cable
  • Changing the SATA port
  • Checking SMART long tests
  • Reseating the drive
  • Checked PSU rails

Nothing helped. The NCQ errors kept coming back,

The Real Cause: NCQ on Enterprise Drives​

Eventually I learned a painful truth:
Some Seagate Exos drives behave badly with NCQ on certain controllers.

Enterprise disks have aggressive queueing behavior, and cheaper onboard SATA controllers (especially in mini-PCs or consumer motherboards) can't always keep up.

The result:
NCQ commands get stuck → the kernel marks the command as frozen → ATA timeouts explode.

So I decided to test the one thing I hadn't tried yet: disabling NCQ.

The Fix: Disable NCQ Completely​

I disabled NCQ using:

echo 1 > /sys/block/sdX/device/queue_depth

and the effect was immediate:

  • no more freezes
  • no more ATA errors
  • no more I/O stalls
  • disk running perfectly stable

After hours of debugging, that one line fixed everything.
To make it permanent, I added a kernel parameter:

libata.force=noncq

Updated GRUB:

update-grub

Rebooted.
And from that moment on, the drive has been flawless.

Why Disabling NCQ Works​

NCQ is designed to improve performance by sending multiple commands at once.
But with certain combinations of:

  • enterprise drives (Exos, IronWolf Pro)
  • consumer SATA controllers
  • high queue depths
  • specific firmware versions

NCQ can cause more harm than good.
Disabling it reduces performance slightly, but guarantees:

  • stable read/write behavior
  • zero freezes
  • no timeouts
  • consistent operation under load

And on a 12TB drive used for bulk storage, consistency matters more than a small NCQ performance gain.

Final Thoughts​

In the end, the disk was never faulty.
The problem was NCQ itself.

If you own a Seagate Exos drive and are facing:

  • READ FPDMA QUEUED
  • frozen command
  • NCQ errors
  • ata1.00: exception Emask...

Try disabling NCQ before assuming your HDD is failing.
It saved my drive — and hours of frustration.