Shares missing after adding 2nd parity drive

[Singularity42]

Last night I powered down my unraid, added an additional disk to be used for dual parity, and now Ive got issues in unraid.

- No shares appear in unraid - but the shares are still accessible and mounted

- Cache drive shows up as a cache device and unassigned device - shows it has data use but there are no files/folders:

 

Diag attached.  Also, Im supposed to copy and paste this message from "Apps"?

 

Hoping someone can help me here.  

 

 

singularity-diagnostics-20200520-0855.zip

[Singularity42]

This has something to do with my cache drive showing up as two devices.  If I spin up the array, I see all my shares, everything works.  Then the logs of the disk go into IO Errors, then I lose everything.   

When this happens the cache disk shows up as cache (sdi) and unassigned (sdg).  Im assuming unraid doesnt know which is which and thats why it halts everything and stops the cache disk.  

Now, how to fix?

[Singularity42]

I was able to resolve this by: 

- stop array

- un-assign cache device

- remove missing device (same drive, different id) 

- assign new cache

- start array again.

[JorgeB]

Cache device dropped offline and reconnected with a different identifier:

May 20 08:14:00 Singularity kernel: ata7.00: disabled
May 20 08:14:00 Singularity kernel: sd 8:0:0:0: [sdg] tag#6 UNKNOWN(0x2003) Result: hostbyte=0x04 driverbyte=0x00
May 20 08:14:00 Singularity kernel: sd 8:0:0:0: [sdg] tag#6 CDB: opcode=0x28 28 00 0a 07 6e c8 00 01 00 00
May 20 08:14:00 Singularity kernel: print_req_error: I/O error, dev sdg, sector 168259272
May 20 08:14:00 Singularity kernel: print_req_error: I/O error, dev sdg, sector 168259528
May 20 08:14:00 Singularity kernel: print_req_error: I/O error, dev sdg, sector 168259272
May 20 08:14:00 Singularity kernel: print_req_error: I/O error, dev loop2, sector 6207736

Check cables.

[Singularity42]

Ya, I thought about this, but I wasnt close to  these cables at all and I was able to reproduce the issue in a very strange way.  

 

Stop/Start array, cache disk comes in as sdi ( i dont know why ).
Go mess with mounts (change directory location for mount) on a virtual machine that pulls scripts off original (sdg) cache disk.

Disk now changes its identifier (verified by watching lsscsi) to sdg

Disk IO errors, and IO is halted, due to id changing.

 

I was able to do that 5 times with same result each time.

 

I fixed it with the above steps, and its no longer an issue.  🤷‍♂️
Now just issues with IOMMU and booting that vm , yay.

[JorgeB]

Then you might have something trying to use that SSD/port, like a VM.

[Singularity42]

Yep I think thats exactly what was happening.  Once I cleared the mounts off the vm to the cache drive, and went through the stop/remove missing/reassign cache/start, then recreated the mounts to the same disk.  Working like a charm now!