I understand from another post that re-scanning of user shares will be going away soon. I'll go ahead and report this though... If I'm in the middle of transferring a lot of files to the server, and hit re-scan in my browser, it spins until the transfer is complete. (At least I assume so... I'm still waiting to finish transferring 300 GB.)

 

While I'm here, I have a further feature request that I suspect is already planned... Full merging of folders when constructing the user share. Instead of using the conflict resolution scheme described here: http://www.lime-technology.com/wordpress/?page_id=32, merge the conflicting folders at the subfolder level.

 

This way, if my folder structure looks like

 

//tower/disk1/Media/Movies/Action

//tower/disk1/Backups

//tower/disk2/Media/Movies/Suspense

 

my movies will all get merged. The scheme described in the Wiki page above says that only one "Movies" folder will be visible. NOTE: I haven't tried this in 3.1b2 yet... Perhaps it is already fixed?

 

You don't want to try to re-scan while you are writing data.

 

Regarding folder merge - in example you posed, you would get two shares:

 

Media

Backup

 

The Media share would have this directory structure:

 

Media

  Movies

    Action

    Suspense

You don't want to try to re-scan while you are writing data.

 

Could unRAID periodically do the rescan itself (when it's not writing, obviously)? This would automate the process and prevent the user from hosing his/her system.

 

Regarding folder merge - in example you posed, you would get two shares

 

Yes, I found it does the right thing. I did find a problem, however... I accidentally moved some directories from the deep merged level up one level. Windows explorer seemed to do the right thing, at which point I said "Doh!", not knowing what unRAID would do on the disks. When I checked them, they were unchanged. (?)

 

By the way, have you fully worked out the semantics for all disk operations on the merged virtual drive? I could imagine that this might be tricky. If it would help, I could create a formal model in Alloy and figure out the right semantics. Basically I would bi-simulate a non-virtual drive along with the virtual drive, and ensure that all operations on the latter look like they would on the former.

 

This would help resolve tricky edge conditions. For example, suppose your virtual drive reports 2GB of total space, but the space is divided evenly across two physical drives. If you were to write a 1.8GB file, you should get an out of space message since unRAID can't split the file across the drives. (This is one case where the virtual drive doesn't behave exactly like an equivalent physical drive.)