Hello

I need your help.

 

Setup:

- 2 usb drive boxes, each with 4 mostly Seagate Ironwolf 4 TB drives

- connected to a Ryzen based MiniPC (Minisforum UM350) via USB-C / USB A 3.1

- in operation without major issues since about 2 years

 

Remark:

- I know USB connected drives are not the best solution for a file server, but that's what I have ...

 

Problem:

- After one of my data drives became offline because of i/o errors (read?/write?) I got a new Ironwolf 4TB drive and started the rebuild

- The first rebuild canceled near the end because of to much read errors

- I've started the rebuild a second time and it finished with the following message: "Data rebuild finished (172 errors) warning". The replaced data drive "returned to normal operation"

- the parity drive shows 8 unrecoverable sectors in the smart info

- now I want to buy another hdd to replace the cache drive

 

Questions:

- do read errors mean the data could not be read at all during the rebuild or are these recoverable / recovered read errors?

- can I trust the data on the rebuild data drive?

 

Diagnostics zip is available but I don't want it upload to a public forum, private upload to Limetech would be ok

 

Best regards

Fritz

 

 

 

As nobody answered to this request yet, some of my findings:

 

- I've compared the checksums of my big picture and video library (about 500.000+ files) with GoodSync against a known good copy and found several files with corrupt checksum but same sized & date. I cannot say for sure that this is caused by the rebuild with read errors, but there is a high propability.

- Therefore I think the good state of the drive after a rebuild with read errors cannot be trusted. I think this should be more clearly stated (or is it somewhere??)

- In future file server builds I will move away from the unraid array approach to a file system which keeps checksums of the files for data verification esp. zfs. zfs scrub is your friend, if you are not sure about data integrity.

- I like unraid a lot and will continue to use unraid. I've already started to test unraid 7 with a 2 TB  SSD zfs mirror.

- I'm a little disappointed that nobody answered to my original questions within 7 days. For a paid product I've expected a faster response of a customer help request.

Remark: I've started to use unraid 2+ years ago. In the meantime I've paid for and run 3 unraid instances.

 

I hope this info helps someone looking for the same questions/issues i've had.

 

Please correct me if my comments / findings are wrong.

1 hour ago, DirschedlF said:

I'm a little disappointed that nobody answered to my original questions within 7 days.

Without posting the diags we can't help, and since you mentioned you didn't want to upload them, I didn't ask.

 

Diags are anonymised by default, is there any info there that still concerns you?

 

 

Jorge,

 

thanks for the answer. 

 

Regarding the diagnostic zip file I generally don't like the idea to share information to everyone on the internet. I would feel more comfortable to share it with a limited set of thrustworthy people.

 

In the meantime, I rebuilt a new parity drive without errors and I've reset the error count. Let's see how the system behaves over the next weeks.

 

It might be just me, but one improvement might be to qualify read errors with (recovered) / unrecoverable and make it more clear that a rebuild with read errors cannot be fully trusted. I feel the ok state in the GUI after such a rebuild might mislead someone. Perhaps it's already written somewhere but I couldn't find a clear answer to this situation.

 

Best regards

Fritz