I built a new core ultra based server back in November of last year and noticed that it would report 5 or 10 sync errors every time it did a parity check. At first I thought it might be because the kernel that Unraid was using didn't have the drivers for the latest intel CPU but now with the release of 7.1 that is no longer the case so I figured I should spend some time troubleshooting this.

 

I have replaced all the SATA cables, the drives were originally plugged into 2 asm1166 m.2 cards, and are now plugged into 2 asm1166 pcie cards that I had used for the last few years in my previous server without issue and ran a memtest which showed no errors. Also updated the mainboard BIOS, so kind of stumped as to where to go next.

 

has anyone else had these sorts of issues or is there anything obvious in the diagnostics?

 

 

 

phonebox-diagnostics-20250506-1054.zip

Do you see the errors if you do another check without rebooting, or only after a reboot?

That's a good question, it usually errors within the first hour of starting a parity check so I'll start another and see what happens and report back later.

I remember I was facing the same problem. Since I only used the onboard SATA 1-8 the errors are gone.

But the 2 cheap SATA controllers in the slots (dunno which chiptype, too long ago) had the problem on server shutdown.

They seemed to modify some sectors before power turned off. The sector numbers were all quite "early" on the disk and mostly the same each time.

(It was harmless, did never kill any data. Maybe a lazy flush of some cached directory or something.)

Its only noticeable with array drives.

So I moved the drives "to the front" and everything is fine again.

(But this does not mean that this is the same for you, just give it a try if you can)

 

The second run of parity so far has not had any errors yet and is well past the point it usually has the 5 errors by now. I'll let it complete and try running it again in a weeks time without rebooting to confirm. 

 

The mainboard is a Asrock Z890 Pro-A which unfortunately only has 4 onboard sata ports so not possible to run the whole array from them. My case currently has space for 16 HDD`s and I was hoping to be able to use 2 6 port m.2 and a 6 port pcie asm1166 based cards due to the power draw (a 9300-16i would cost around £50 a year more than the 3 asm1166 cards)

I would propose that you do another check very soon (should give 0 errors again), then reboot and recheck.

If there are 5 errors again after the reboot, you sadly have the wrong SATA adapters.

 

You have to consider then if you can live with those errors (that should be harmless, but of course there is no guarantee for this) or throw out the sata controllers.

 

But in any case, stay away from those >4port cards! They use a multiplexer intern this halfs the speed instantly. You slow down your whole system and even risk blocking (accessing 2 drives on the different side of the multiplexer is not possible, a write to the array will always need this).

 

Quote

I would propose that you do another check very soon (should give 0 errors again), then reboot and recheck.

If there are 5 errors again after the reboot, you sadly have the wrong SATA adapters.

 

You have to consider then if you can live with those errors (that should be harmless, but of course there is no guarantee for this) or throw out the sata controllers.

That's my plan now, just waiting for the current parity check to finish. It's at 50% now with no errors yet.

Quote

But in any case, stay away from those >4port cards! They use a multiplexer intern this halfs the speed instantly. You slow down your whole system and even risk blocking (accessing 2 drives on the different side of the multiplexer is not possible, a write to the array will always need this).

I don't think that is correct for the asm1166 as it natively supports 6 sata ports, @JorgeB would know more though as he has done extensive testing with them.

 

35 minutes ago, TenThirty said:

the asm1166 as it natively supports 6 sata ports,

That is correct, but could still be related to the controllers, either an issue with the specific models you are using, or some other hardware combination/compaibility issue.

 

 

2 hours ago, TenThirty said:

That's my plan now, just waiting for the current parity check to finish. It's at 50% now with no errors yet.

no need to wait. if the errors are always in the front (this was the same here too), you can safely cancel, reboot and start a new check.

If it shows the usual errors again then, you know what to blame and to do.

Just an hour run should be enough to know.

(The check does not change anything normally, it just reads the data and marks the problematic sectors. If you set the "correct me" hook it then "repairs" these faults by overwriting them with the supposly correct value)

Write down the sector numbers that it complains about, if they are always the same (or at least very close to each other from the last run) its very likely that you have a controller problem.