Removing a data drive.. what happens to disk numbers?

[jbuszkie]

I'm consolidating a couple disks and will remove them from my system to save a little power.

I'm following this to remove the drive.  The drive I'm removing is disk9  I have a disk 10 that's staying.

Will disk10 become disk9?  or will I have a hole where disk9 used to be?  will it go disk7, disk8, disk10?

 

Also..  If I don't zero out the drive first and do the rebuild..  Will I still be protected if I have dual parity?  I'm thinking no

since it "resets the array configuration" and has to rebuild parity from scratch?

 

Thanks,

 

Jim

[JorgeB]

34 minutes ago, jbuszkie said:

Will disk10 become disk9?

Nope, it will be an unassigned disk.

[jbuszkie]

Is there a way to remove the hole?  and make disk10 -> disk9?

 

 

[JonathanM]

Not without rebuilding the parity2 slot.

[jbuszkie]

15 minutes ago, JonathanM said:

Not without rebuilding the parity2 slot.

Will the array be protected while it rebuilds parity2?

 

Is this documented somewhere?

 

Thanks!

[JonathanM]

3 minutes ago, jbuszkie said:

Will the array be protected while it rebuilds parity2?

 

Is this documented somewhere?

 

Thanks!

It's a little complicated. Parity1 doesn't care what order the disks are in, as long as they are all there.

Parity2 uses disk slot number as well, so it's invalid when you move slots around.

 

There is no easy way to tell unraid to only rebuild parity2, so you have to set a new config with all the drives where you want them, WITHOUT assigning parity2, and select that parity is already valid, start the array, cancel the automatic parity check, stop the array, add parity2 back, allow it to build.

 

All this assumes the drive being removed is properly zeroed, that can take a very long time.

 

Most people just take the risk of being unprotected while parity is being rebuilt when removing a drive slot. That assumes all your drives are healthy, but if you don't think all your drives are healthy that needs to be addressed BEFORE going through this drastic of a procedure.

[jbuszkie]

Ok.. So something like this:

 

Zero the disk (I'm like 25% done with that.)

Take a screen shot the array config

stop the array.

unassign disk10

assign disk10's drive to disk 9

unassign parity2

New configuration (keep existing assignments)

start array with trust parity

Start a parity check and let it go a little to make sure  it starts and doesn't report tons of errors. (sanity check)

Stop parity check

Stop array

re-assign parity2

Start array

Parity 2 rebuild

done?

 

Jim

 

Edit:  added/changed the steps I did.

 

 

 

 

 

Edited September 21, 2024 by jbuszkie
update the steps I did

[JonathanM]

Have to set the new config before you change the assignments.

[jbuszkie]

17 minutes ago, JonathanM said:

Have to set the new config before you change the assignments.

It's always scary doing that! lol

 

[itimpi]

2 hours ago, jbuszkie said:

It's always scary doing that! lol

 

I would recommend you always start by using the option to retain all current assignments, and then return to the Main tab and make the minimal. Hangers required.   That at least reduces the chance of messing up disk assignments.

[Veah]

It does this

[jbuszkie]

21 hours ago, jbuszkie said:

Ok.. So something like this:

 

Zero the disk (I'm like 25% done with that.)

Take a screen shot the array config

stop the array.

unassign disk10

assign disk10's drive to disk 9

unassign parity2

New configuration (keep existing assignments)

start array with trust parity

Start a parity check and let it go a little to make sure  it starts and doesn't report tons of errors. (sanity check)

Stop parity check

Stop array

re-assign parity2

Start array

Parity 2 rebuild

done?

 

Jim

 

Edit:  added/changed the steps I did.

 

 

 

 

 

 

Ok  I did the above steps and I seen to be working ok.  (parity 2 is in the process of rebuilding)

 

caveats:

This documentation needs to be updated:  It's pre 6.12 and the mdX device numbers have changed to mdXp1 for 6.12

The Clear Drive script as I had loaded in User Scripts plugins is not working for my unraid.  The script used pre 6.12 nomenclature.

The procedure calls for unmounting the drive before doing the zeroing.. 

umount /mnt/diskX
dd bs=1M if=/dev/zero of=/dev/mdX status=progress

Again the mdX needs to be mdXp1

And this worked fine until I tried to shutdown the array.

The shutdown script didn't like the fact that the drive was already unmounted and looped trying to unmount that drive.  I believe this is a bug

and the shutdown script should check to see if the drive is already unmounted...

 

Now one can just do a new config make all the changes you want and just have it reconstruct the parity..  but even though my system is healthy and I'm sure it would have been fine...  I'm paranoid and wanted to be protected at (almost) all times!

 

Thanks for the input, guys!