Jump to content

reiserfsck --rebuild-tree, how to procede after


Recommended Posts

Running Unraid 5rc3

Im in the process of running "reiserfsck --rebuild-tree"

 

I unassigned the physical drive that Im running reiserfsck on (and eventually the --rebuild-tree too), then restarted the array.  Now that its almost finished.  How do I get the repaired disk back in the array?  If I reassign it, will I need to run the new config, since it will be different than the logical version of the drive (with errors) Ive been using while rebuilding the tree?  Also do I need to do a parity check (trusting the array) as soon as I reassign it?

 

I figured if I unassigned the disk (keeping it from unmounting), versus starting array in maint mode, that I would be able to use the array while rebuilding the tree.  Now I'm wondering if I made more problems for myself.

 

BTW, I have not written any files to the array since the disk has been unassigned.

Link to comment

No, I ran it on md.  Do I trust array when finished or parity?

 

Thanks!

I think from the way you described your recovery actions we assumed that you had run against the raw device rather than putting the array into Maintenance mode and running against the md device (which is the correct way to do this).

 

If you ran it on the md device then you should be able to start the array on completion and see all your data (unRAID should be emulating the missing drive in such a case).  Assuming that is the case and the data looks fine, then when you stop the array and assign the physical drive unRAID should offer to rebuild it - let it do that and then when that completes you should be back in a protected mode.

Link to comment

Is this correct?

1. Array was stopped and disk unassigned.

2. Array started

3. Reiserfsck was performed on /dev/mdX

 

If so, stop array and reassign the disk. Start the array and the disk will be rebuilt with the corrected file system.

 

With disk unassigned and operations performed on the dev/mdX only the simulated disk was corrected and the unassigned disk has invalid content. Reassign the disk and the corrected content will be written.

 

Why did you unassigned the disk? The rebuild that must be performed now was avoidable.

 

Check Disk Filesystems in my sig has the correct procedure for next time...

Link to comment

Is this correct?

1. Array was stopped and disk unassigned.

2. Array started

3. Reiserfsck was performed on /dev/mdX

 

If so, stop array and reassign the disk. Start the array and the disk will be rebuilt with the corrected file system.

 

With disk unassigned and operations performed on the dev/mdX only the simulated disk was corrected and the unassigned disk has invalid content. Reassign the disk and the corrected content will be written.

 

Why did you unassigned the disk? The rebuild that must be performed now was avoidable.

 

Check Disk Filesystems in my sig has the correct procedure for next time...

If the recovery on the emulated content had failed, he would still have the original drive unaltered for another bite at the apple. As long as the rest of the drives are 100% healthy, it's a pretty low risk strategy, albeit one that requires a fair bit of background knowledge to accomplish safely. Perhaps a better strategy would be to pull the original drive and replace it, allow the rebuild to take place, then operate on the newly rebuilt drive. That way fault tolerance on the rest of the drives would be compromised for a shorter period of time, while still allowing for other recovery methods to be tried on the original drive.
Link to comment

Yes dgaschk, I did it like you described.. My thought process was maint mode was to ensure disk was not mounted,, and I could still use the emmulated disk during rebuild. I thought I was correcting the physical disk, but seems like It did the opposite (corrected the emmulated version). Upon completion, I will reassign then perform rebuild of the disk.

 

Thanks everyone, Ill post my success after the rebuild. Hopefully I won't create any more issues and headaches.

Link to comment

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...