cache pool corruption and strange behaviour [6.10.3]

[icedfusion]

Yesterday evening I created an additional cache pool that I wanted to use only for plex app data. I setup another share called plexAppdata that was set to use only this new cache pool (from the drop down list I could select my nvme cache pool drive, or the ssd cache pool drive)

I noticed that this ssd cache pool was mirroring the contents of my nvme cache pool - I assumed because it was a separate cache pool this would not happen - is my understanding not correct?

 

I then went into the settings and 'erased' the ssd cache pool - but this also erased everything on the nvme cache pool. They really were separeate cache pools, both with a slot size of 1.

 

After being extremely annoyed (but i had backups of everything) - I have now had to reformat both drives, added them both back as separate cache pools and now they are not mirroring (the only change i did was to change the ssd cache pool to xfs instead of btrfs).

 

Is my understanding incorrect or did I do something wrong?

 

Cheers

 

ice.

 

[JonathanM]

If the two drives were EVER assigned to the same pool, even briefly, they "permanently" remember that status because BTRFS keeps a record of pool members totally separate of Unraid's pool definition. To remove them from the BTRFS pool can require extra steps, @JorgeB is the resident expert on what is required. Theoretically if the drive is removed from Unraid's pool correctly, the status is updated and things work as intended. If you physically remove the drive, I believe it may still have the BTRFS headers on it that make it a member of the pool, I think you have to leave the drive installed and remove it from Unraid's pool definition, then the pool is rebalanced to remove it.

 

I'm a little fuzzy on the exact procedure, but what you experienced is a symptom of both drives being assigned to a single pool at some point in the past, and not being properly BTRFS balanced to separate drives.

[icedfusion]

Thank you for this, yes, the drive was originally part of the other pool, so this explains a lot. A bit annoying as it caused some issues that as a user I would never know.

Good to know my understanding of function was correct but now I know what went wrong and not necessarily user error.

 

Cheers

ice.

[trurl]

Probably your docker/VM related files are on the array now since they would have been recreated there when cache was missing.

 

https://wiki.unraid.net/Manual/Shares#Default_Shares

[JorgeB]

This won't happen if you properly remove the device from the pool first, or wipe them manually, if you don't and separate both members to different pools it will happen, but we don't want Unraid to wipe new members of a pool, or it would not be possible to import old pools.