Recovering from corruption

[mortar]

When I set up my freenas experiment, I had some corruption issues rather sooner than expected:
https://forums.freenas.org/index.php?threads/asrock-c2550cd4i-corrupting-files.39879/

It basically looked something like:

Code:

NAME  STATE  READ WRITE CKSUM  
  pool-a  ONLINE  0  0  1  
  raidz1-0  ONLINE  0  0  2  
  gptid/xxx  ONLINE  0  0  0  
  gptid/yyy  ONLINE  0  0  0  
  gptid/zzz  ONLINE  0  0  0  

errors: Permanent errors have been detected in the following files:  

  pool-a/mydataset:myfile

Months passed and I didn't bother to do anything about the above corruption in just one file since things were otherwise running fine. Now I decided to clean up the setup a bit and deleted the offending file as a first measure. The following scrub continued reporting a similar error with the only difference that the filename has changed:

Code:

errors: Permanent errors have been detected in the following files:  
 
  pool-a/mydataset:<0x63>

Since there's no file named <0x63>, it obviously refers to some other type of filesystem data, which I find somewhat disturbing. How do I recover from this without rebuilding the whole thing? It was difficult enough to get the smb shares to behave themselves and I wouldn't like to start from scratch.

That the error can migrate from a file to somewhere else looks like a bug in ZFS code to me.

This thing is running FreeNAS-9.3-STABLE-201602031011. Would upgrading to 9.10 likely just increase the amount of problems?

 

[Ericloewe]

It's probably going to be metadata associated with that file, possibly in a snapshot.

This thing is running FreeNAS-9.3-STABLE-201602031011. Would upgrading to 9.10 likely just increase the amount of problems?

No, It's going to be neutral in that regard.

 

[mortar]

Ok, Fixed by running another scrub on it. It was a large file so maybe the delete was half-way through during the first scrub or something.