I recently had issues with my iGPU and this recreated my entire USB. The only thing I kept is the data on my three data drives. Beforehand, I had copied all data off my cache SSD and reformated it with the new install.

 

Now, after copying all Docker data back over, I now encounter issues with reinstalling the old containers. Via SSH, I get the message that my cache is read-only. The SSD sometimes appears twice (see image); once in the cache pool and a second time as an unassigned device.

tower-diagnostics-20240426-1508.zip

Pool went read only, with btrfs suggest you backup the data and re-format.

On 4/26/2024 at 4:03 PM, JorgeB said:

Pool went read only, with btrfs suggest you backup the data and re-format.

 

In hoping that I might have done something wrong earlier that you can point out, here's how I already did that this morning:

 

1. Stop the array.

2. Remove the SSD from the pool.

3. Delete the partition via Unassigned Devices.

4. Add the SDD to the pool.

5. Start the array and reformat the pool drive.

6. Restarted the server for good measure.

 

Edit: After reformatting again, I still get this issue. Docker, for example, says: `docker: Error response from daemon: error creating temporary lease: file resize error: truncate /var/lib/docker/containerd/daemon/io.containerd.metadata.v1.bolt/meta.db: read-only file system: unknown.`

Edit 2: For now, I've removed the cache drive entirely and running everything off the array HDDs. I'm still unsure on how to proceed due to the cache drive's behavior (especially it appearing twice).

Edited April 30, 20242 yr by DesertCookie

On 4/26/2024 at 4:03 PM, JorgeB said:

re-format.

I've tried reformatting for the third time now. I've used a brand new SSD. I've used a carrier card. I've swapped RAM. Every time, the cache turns read-only within minutes of starting the array.

tower-diagnostics-20240505-1843.zip

The NVMe device is being passed through to the Spiele VM, correct that and it should resolve the issue.

1 hour ago, JorgeB said:

The NVMe device is being passed through to the Spiele VM, correct that and it should resolve the issue.

That's... what? Thanks for pointing that out! That VM has been dormant for months now and I never started it up since removing its GPU. I never had the NVMe drive passed to it, but had a HDD passed though that I removed when upgrading to 6.12. Does it make sense that the NVMe bound itself to the VM and that this causes issues, despite the VM not being turned on? Perhaps the new USB I made reassigned the VFIO IDs and the SSD thus moved up into the place of the now removed HDD...? I assume this is not expected behavior. Do you think it's worth investigating and potentially opening a bug report?

Edited May 6, 20242 yr by DesertCookie

If you add or remove hardware the devices IDs will change, so what before was the GPU is now the NVMe device, this is not a bug.