All disks on SAS expander have 10k+ UDMA errors

[tucansam]

I realize this could be the host controller, the cable in between, the SAS expander, or the PCIe slot on the server... But I just plugged in 8 disks via a host/SAS expander combo and unraid immediately lost its mind, throwing 10k UDMA errors on some disks and up to 19k UDMA errors on others.  All disks attached to the expander chain have the same issue.

 

I used to get these a lot, using 5-in-3 backplanes, and usually fiddling with the SATA cable and/or ignoring the errors completely caused no major issues.

 

But 100% of the disks, at once, and with such a high error count..... I'm nervous about adding any of them to the array at this point.

 

Comments?

[JorgeB]

14 hours ago, tucansam said:

Comments?

 

Basically of you have to rule out any of these:

14 hours ago, tucansam said:

this could be the host controller, the cable in between, the SAS expander

 

[tucansam]

As in replace each one, one at a time, until I figure out the bad part?  

[JorgeB]

Replace or try with different parts if you have them available, not much else you can do.

[tucansam]

I am re-evaluating the setup and am going to try an 8086 -> 4-port SATA cable directly connected to the external port of the LSI controller.  The expander, the cable from the expander to the drives, and the cable from the LSI to the expander can all be eliminated as variables.  

[tucansam]

OK.

 

I got rid of the expander and all cables.

 

I now have the LSI with four external ports, an 8088 -> 4x SATA cable, and my drives.

 

Just plugged four drives into said cable, and one immediately threw 14000 CRC errors.  In addition, two of the four drives connected show up as "Dev 4" and "Dev 6" and show all SMART attributes.  The other two are "sds" and "sdt" and show no SMART attributes and will not let me run a SMART test.

 

I'm curious if the CRC error reporting is because unraid has noticed the drives... that is to say, I am hoping the error count does not climb.  I'm not going to add them to the array.  Should I mount them individually and move some data around to test them, to see if the error counts rise?

 

Is there a way to reset the CRC error count?

Edited September 12, 2021 by tucansam

[itimpi]

7 hours ago, tucansam said:

s there a way to reset the CRC error count

No.  
 

The CRC error count is stored internally in the drive and never resets.   It typically indicates a connection issue (e.g. power and/or SATA cabling) rather than a problem with the drive, and if that is rectified the best you can do is make it stop increasing.

[tucansam]

Dammit.  I've got a 650w PS with those 5-way SATA splitters, and cables I've used in other servers with no issue.  I suppose its the LSI controller card.  I've eliminated everything else, essentially.  

[tucansam]

Do the LSI cards with the four external ports, and no internals, need to be flashed to IT mode also?  I picked up a Dell branded LSI and it sees no disks attached... But the old card (that was giving CRC errors) saw them.  Is this a firmware issue that is easily fixed?

[JorgeB]

14 hours ago, tucansam said:

Do the LSI cards with the four external ports, and no internals, need to be flashed to IT mode also? 

Depends on the model, usually no.