[SOLVED] Tons of disk errors, trouble setting up Unraid

[steberr]

Ryzen 3 3200G

Asrock B450M Pro4 with BIOS 3.90 (latest version)

DDR4 2666 RAM 2x8 GB

450W PSU

 

Originally I had my a parity disk and disk1 disabled, and I had to do a hard reset because the server had locked up.  So, I had no logs to show.

 

I started pre-clearing 2x14TB and 2x12TB drives, and after they were all nearly zeroed (1.5 days or so), the server rebooted.  I looked up the Ryzen fixes, as I started getting reboots around every 20-30 minutes, and had to keep resuming the pre-clears.  Then, after making some of the BIOS changes (typical current for psu and disabling global C-states), I also changed some other setting that was making me reboot every 5 minutes.

 

Finally, I got it all sorted out, and things were stable.  I have no reason to suspect the drives are bad, and they were nearly finished pre-clearing with no errors (and pre-cleared status in Unassigned Devices), so I started to sync parity with a New Config.  2 parity drives (14TB drives) and 2 disks (12 TB drives) with 256GB SSD cache.

 

After running for about 1.5 hours, I received a warning about read errors from the 2 disks.  

Here is a link to a pic from the Main screen with all of the 1 million errors: https://imgur.com/qF8xeef

 

Attached are the diagnostics.  I'd love to get Unraid working, because it seems really great. So many problems so far.

blackburrow-diagnostics-20200617-1927.zip

 

 

EDIT: For anyone else that has a Ryzen build, some or all of the following steps may be necessary to fix similar issues.  

1) If using Marvell SATA controllers--stop.  Use an expansion card with non-Marvell controllers or use other SATA ports on the motherboard without the Marvell controller (if they exist).

 

2) Check BIOS to ensure Global C-States is disabled and Power Supply Idle Control is set to "typical current idle"

 

3) Edit the Flash drive syslinux config to include the following command: rcu_nocbs=0-3 where 0-3 should be the number of CPU cores (if you have 4, 0-3; if you have 8, 0-7).  You can check the number of cores, if you aren't sure, by looking at the Processor information in the Unraid Dashboard.

 

Making all of these changes resolved my issues, and Unraid is now running normally.

Edited June 24, 2020 by steberr
Changed to SOLVED

[steberr]

After re-reading some of the responses in my other post, and some other topics with related issues, it looks like the Marvell SATA controllers on my motherboard may be the culprits at this point.  I am connected the drives into the expansion SATA card, as the SSD cache drive has not had any issues (yet).  I will try pre-clearing again to see if I can get them to complete without a reboot, and if they do, if parity can sync without errors.

 

Is it safe to pre-clear all 4 drives at once, or do I need to pre-clear sequentially?

 

Also purchasing different RAM, as I have 2667 single 2x8 GB, and it looks like according to a Ryzen chart posted in another thread that I should either be using 4x 2133 or 2x 2933.

Edited June 18, 2020 by steberr

[JorgeB]

You should try v6.9 -beta as there are some known issues with the Ryzen SATA controller and older kernels.

 

5 hours ago, steberr said:

as I have 2667 single 2x8 GB

That's fine, up to 2 x 2933 is OK (depending on board)

[steberr]

Ah okay, I see how I misinterpreted the chart.  Well, at least something is okay in my build!

 

Can I pause the pre-clears that are currently in progress to update to v6.9-beta?

Edited June 18, 2020 by steberr

[JorgeB]

23 minutes ago, steberr said:

Can I pause the pre-clears that are currently in progress to update to v6.9-beta?

I believe you can, but it's been a while since I used the plugin.

[steberr]

This appears to have been resolved by changing some settings in the BIOS for Ryzen (global c-states disabled and power supply idle control = typical current idle), adding some code to the syslinux config (rcu_nocbs=0-3), and using an expansion card for all SATA connections.  I have been up for 5+ days and pre-cleared all drives successfully, and I'm currently transferring data to the server, now, with 0 errors.

 

I will be updating the OP with all of the steps I took to resolve the issue.  Thanks for the all the help.