Replacing Cache Pool Devices using btrfs Replace [Solved]

[Gico]

Hi,

 

Can anyone help with the procedure to replace the 2 cache pool devices (hard disks) with 2 SSDs (sdab, sdak) using CLI?

I prefer using CLI (btrfs replace commands) instead of moving the data to the array and then back because I have Plex data,

so it will probably take ages.

I understand that using the GUI to do it is broken on v6.9.2.

 

Thanks.

juno-diagnostics-20211128-0938.zip

Edited December 1, 2021 by Gico

[JorgeB]

https://forums.unraid.net/topic/115767-solved-unraid-v692-upgrading-cache-pool-drive/?do=findComment&comment=1052576

 

[Gico]

Thank you. Ran the replace commands but both have same error because SSDs are somewhat smaller than the current cache disks (all 4 disks are "4TB"):

ERROR: target device smaller than source device (required 4000786980352 bytes)

 

Any reasonable option to overcome this issue? btrfs resize maybe?

Or  alternatively add the 2 SSDs to the pool (through GUI?) to have 4 devices on the cache pool, then do "btrfs device delete" one by one?

 

Edited November 28, 2021 by Gico

[JorgeB]

14 hours ago, Gico said:

because SSDs are somewhat smaller than the current cache disks

You can't direct replace to smaller devices.

 

14 hours ago, Gico said:

Or  alternatively add the 2 SSDs to the pool (through GUI?) to have 4 devices on the cache pool, then do "btrfs device delete" one by one?

You can do that, when done the cache reset procedure is the same as the one posted in the link above.

 

P.S.: metadata profile is DUP, you should change to RAID1.

[Gico]

Metadata profile changed to RAID1.

Stopped the balancing because anyway the device delete should balance the data to the new disks.

Currently running "btrfs device delete /dev/sdm1 /mnt/cache"

but counter my expectations, there are no meaningful writes to the new devices.

Is that normal?

Running "btrfs device usage" I see the data on the two old disks decreasing 3.18Tib --> 3.17Tib --> 3.11Tib...

That's freaking me out a bit, although used space on the cache pool seems to be steady.

 

juno-diagnostics-20211129-1636.zip

[JorgeB]

2 minutes ago, Gico said:

Is that normal?

Yes, first it will balance the partiality filled chunks, after that's done it should start moving the data.

[Gico]

All done, Thanks a lot. 

1st disk took about 26 hours, the 2nd one 6 hours.