correct way to fix disk errors

[Agamemnon]

I just want to make sure what the best course of action is for my situation.

Currently my parity drive is 18TB. I have a new 20TB drive ready and was planning to update the parity with it.

 

Now one of my 14TB data drives threw 12,310 Errors and disk log shows a lot of the following errors:

Tower kernel: I/O error, dev sdh, sector 26866835136 op 0x0:(READ) flags 0x4000 phys_seg 128 prio class 2

At the same time UDMA CRC error count jumped to 1813. So far it hasn't gone higher after switching to a different SATA cable.

 

When I switch out the drive with a new one the errors would get fixed from parity?

Problem is, I can't use the 20TB on hand unless I do a procedure I found that's called a parity swap, but I am unsure if this is advisable or even still supported.

Alternative would be to get another 18TB to replace the data disk first and then upgrade parity later.

[JorgeB]

16 minutes ago, Agamemnon said:

So far it hasn't gone higher after switching to a different SATA cable.

CRC errors don't return to 0, important is that they don't keep increasing after replacing the cable, also please post the diagnostics.

[Agamemnon]

Here are the diagnostics.

 

EDIT: Attachment removed

 

Edited January 2 by Agamemnon

[JorgeB]

Looks like a power/connection problem, and since there were CRC errors the SATA cables should be the first one to be replaced, then keep monitoring, the read errors will disappear with a reboot.

[Agamemnon]

Ok, thank you for looking at the diagnostics. I thought I would have to do a parity check to get rid of the disk errors.

 

What about the parity swap to use the new 20TB as parity and the parity drive to replace the data disk with issues?

[JorgeB]

33 minutes ago, Agamemnon said:

I thought I would have to do a parity check to get rid of the disk errors.

That won't do it, reboot instead, but like mentioned replace the SATA cable also, I don't the need to replace the disk for now.

[Agamemnon]

On 12/27/2023 at 7:38 PM, JorgeB said:

That won't do it, reboot instead, but like mentioned replace the SATA cable also, I don't the need to replace the disk for now.

So far no more UDMA CRC Errors occured.

Should I run a parity check (without writing corrections) before updating Parity drive to 20TB? I am currently preclearing the disk and it will be done tomorrow.

Last parity check was 2 weeks ago before the CRC Errors came up.

[trurl]

Click on the SMART warning for that disk on the Dashboard page to acknowledge the current CRC count and it will warn again if it increases.

 

Do you have Notifications setup to alert by email or other agent as soon as a problem is detected?

[Agamemnon]

On 12/27/2023 at 4:59 PM, Agamemnon said:

Tower kernel: I/O error, dev sdh, sector 26866835136 op 0x0:(READ) flags 0x4000 phys_seg 128 prio class 2

I am just unsure if I need to run a parity check before upgrading the parity disk because of the I/O errors the data disk showed when the CRC errors occured.

 

I acknowledged the SMART warning and will set up email notifications.

[trurl]

The current contents of parity will have no impact on building parity to a new larger disk. All of the parity would be recalculated based on the contents of all the data disks in the array. And rebuild would basically be exercising the same hardware, but would include the new parity disk and not include the old parity disk you are removing.

[Agamemnon]

Thank you, I appreciate both of your help on this.

I somehow was thinking about data corruption protection like ZFS or BTRFS to fix potentially broken files, which parity doesn't give me.