I am running unRAID 7.1.2 and have a BTRFS cache pool. I started with 2 SSDs, but wanted to free up slots for additional HDD data drives, so I put a PCIe card with two NVMe drives on it and added those to the pool.

What I am trying to do now is remove the 2 SSDs from the cache pool, but have been unable to do so. I've tried following instructions here:

<https://forums.unraid.net/topic/133522-cache-pool-remove-a-disk/>

But have been unsuccessful, and I'm hoping someone can help me through this.

I use the cache for my appdata share and docker.img and have incoming files go there to other shares (set for Cache->Array).

I've tried stopping the array, and unassigning the 2 SSDs from the slots, the checking the "I Want To Do This Box" but the array will not restart giving an error about too many missing devices.

I'm not sure what I'm missing and how to fix this, so any help is appreciated.

Thank you.

Edited July 7, 20251 yr by kolepard
problem solved

Please post the diagnostics and the output from btrfs fi show

btrfs_vi_show_cache.txt tower-diagnostics-20250629-1101.zip

And I thank you very much for your help.

Kevin

Edited June 29, 20251 yr by kolepard
text not included with files

Syslog is full of what looks like a power/connection issue with this device:

Device Model: Samsung SSD 860 EVO 1TB

Serial Number: S4CRNE0M305956E

Replace both cables and post new diags after array start.

New diagnostics. Thank you again!

tower-diagnostics-20250701-1426.zip

Pool mounted degraded, what device do you want to be uyed for the pool in the end?

What I ultimately want is 2 drives in the cache pool that are mirrored, so that if there is a problem with one, I can just pull it out until I can replace it. I thought that was how the cache pool worked with multiple devices, but this experience makes me think that is not how it works. Is there a better way to set up the cache so it is mirrored (either actually or effectively) so that if there is a problem with one I can easily switch to another until I can replace the faulty drive?

In any case I want to remove the Samsung and Ironwolf SSDs and use the two NVMe drives in a redundant pool.

Is this an unstable enough situation that I should move everything off the cache pool while I can?

As an aside, I think the problem with the Samsung drive itself, as it dropped off again this morning even with new cables.

Thank you again.

Kevin

The pool is currently only using one device:

Data Metadata System

Id Path single single single Unallocated Total Slack

-- --------- --------- --------- -------- ----------- --------- -----

1 /dev/sdj1 150.00GiB 2.00GiB 32.00MiB 779.48GiB 931.51GiB -

4 missing - - - 931.51GiB 931.51GiB -

-- --------- --------- --------- -------- ----------- --------- -----

Total 150.00GiB 2.00GiB 32.00MiB 1.67TiB 1.82TiB 0.00B

Used 132.38GiB 931.47MiB 48.00KiB

You can do this:

With the array still running, type btrfs dev remove missing /mnt/cache

Then reimport the pool as a single device pool:

on main, click on the first device for that pool and then "remove pool"

back on main, create a new pool with the same name and 1 slot

assign the pool device (sdj), leave the filesystem set to auto

start the array to import the pool

Now stop the array, change slots to 2 and add one of the NVMe devices, this will create a mirror.

stop the array, and replace sdj with the other NVMe device, start array

That should do it.

I ran

btrfs dev remove missing /mnt/cache

Then clicked on the first device in the pool which took a rather long time to load but ultimately showed this, with the "Remove Pool" option greyed out.

The device is currently listed as in Standby mode, and is not spinning up when I click on the grey dot to make it spin up.

I appreciate your ongoing efforts.

Kevin

Looks like the device dropped offline, post new diags.

New diagnostics.tower-diagnostics-20250702-0919.zip

Yep, full of errors again, if you have already replaced the cables, try connecting that SSD to another controller, like the onboard SATA controller.

tower-diagnostics-20250705-1318.zip

Most of my last message disappeared, an only the diagnostics appeared; I'm not sure why.

That Seagate drive is will appear "spun up" for a while after a power cycle, no matter how I attach it, which I think is the major problem. The last set of diagnostics was obtained while it was "spun up" (i.e., green dot). What I did was run

btrfs dev remove missing /mnt/cache

It reported no device was missing, which was true, so I did this:

1. Ran a balance operation and let it finish.

2. Stopped array

3. Set the problematic drive, sdj, to unassigned

4. Checked "I Want to Do This"

5. Restarted the array

6. Let the BTRFS Operation complete, which showed a cache pool of 3 devices (Huzzah! Progress!)

7. Stopped the array again

8. Removed the second SSD

9. Again checked "I Want to Do this"

10. Restarted array

11. Let the BTRFS operation complete again

This has left me with a pool 2 devices of the 2 NVMe drives on the PCIe riser card. Fortunately I was able to complete the first 5 steps before the troublesome SSD dropped off and reported "spun down". I'm not sure what I would have needed to do if the troublesome drive "spun down" before I could complete those steps.

I presume these remaining 2 drives are mirrored, correct?

Is there a better way for me to set this up so I avoid this problem in the future? Or was this just bad luck?

Thank you again for your help.

Kevin

Edited July 5, 20251 yr by kolepard
Add detail/clarity

The diags still show all 4 devices as part of the pool, if it finished the removal post new ones.

Those were before the steps I outlined in the next post, so you're right.

These are after.

Thank you for checking them for me.

Kevin

tower-diagnostics-20250706-0914.zip

Looks good.

Thanks for verifying.

I presume these remaining 2 drives are mirrored, correct?

Is there a better way for me to set this up so I avoid this problem in the future? Or was this just bad luck?

Kevin

5 minutes ago, kolepard said:

I presume these remaining 2 drives are mirrored, correct?

They are.

5 minutes ago, kolepard said:

Is there a better way for me to set this up so I avoid this problem in the future?

That the best way for redundancy, but keep in mind that it's not a backup, e.g. a filesystem corruption issue will affect both devices.

Thanks. I keep backups of everything on the cache drive.

Thank you again.

Kevin