Disk write errors

[Kaltar]

I have a disk that is really making me look like a fool.

Every other day, it will generate a lot of write errors, and unraid disables it.

I then take it out of the array, and back in, let the parity rebuild without errors. Do a full smart test which completes with succes, thinking all is good.

Then a few days later, it generates write errors again, and unraid disables the disk again.

 

I have now taken it out of my array, and run almost a full erase and clear preclear cycle. It is 78% done at step 6 of 6.

With no errors whatsoever.

 

What logs do you guys need to help figuring this out ?

[JorgeB]

4 hours ago, Kaltar said:

What logs do you guys need to help figuring this out ?

Tools -> Diagnostics

 

Better if they cover when the errors happened.

[Kaltar]

I have attached the diagnostics file.

The problem seems to start around nov 29th 11:16
Nov 29 11:16:26 Tower kernel: sd 33:0:8:0: [sdj] Synchronize Cache(10) failed: Result: hostbyte=0x01 driverbyte=0x00

 

And then a huge number of errors some hours later

Nov 29 22:08:25 Tower kernel: md: disk4 read error, sector=1955073552

tower-diagnostics-20181203-2122.zip

[JorgeB]

Twice that disk got disconnected and reconnect with a different letter, unless the 2nd time it was you who disconnected it, in any case the disk does look fine so I would suggest swapping both cables/backplane with a different disk, and if it happens again to the same disk there could be a problem with it despite the healthy looking SMART.

[Kaltar]

I must confess I had not noticed the change of drive letter.

The first time the drive was disabled, I ran a quick smart test, when nothing showed, I thought it was just a fluke.

In order to add the disk back to the array I did the following

1) Stop the array

2) Remove the disk from the array

3) Start the array

4) Stop the array again

5) Add the disk back to the array

6) Start the array and let parity rebuild/resync

 

Is that the right way of doing it ?

 

Anyway, after the complete preclear of the disk, I have added it back to the array, this time it is named "sdh" I will try swapping it around if the error occurs again.

[JorgeB]

14 minutes ago, Kaltar said:

Is that the right way of doing it ?

Yes.