Slow transfer speed from one share to another

[SphericalRedundancy]

So I've got 8 data drives, a parity drive, and a cache drive.

 

So lets say I have two shares, share A is set as yes to use the cache drive while share B is set to not use it at all.

 

I will sometimes need to transfer large (100 GB or more) folders from share B to the share A. Whenever I do this the transfer speed is very slow ranging from 5MB/s to 10MB/s and as far as I can tell this is because share B is not only writing to share A but also the parity.

 

Why doesn't share B transfer to the cache drive first since share A is set to use it?

[JonathanM]

This is more complex than it seems on the surface. If both shares are allowed to occupy the same physical disk, then that 100GB move will happen almost instantaneously, without involving the cache drive, because moving from share A to share B is a simple single folder rename operation.

 

The simple answer to your question is, if you allow all shares to occupy all disks, then any such operation as you describe will be pretty much instant and not need the cache. If you insist on forcing specific files to specific disks, it has to involve both data disks and the parity drive, so it will be slow. I suppose limetech could implement the logic needed to force the write to cache so the mover will put it back on the array later, but they haven't. If they did, then the people who let unraid use all the disks for all the shares would complain that a move that used to take half a second to complete is suddenly taking a very long time.

[SphericalRedundancy]

I guess they could also implement logic to only move to cache if moving to a different drive and skip cache if moving to the same disk just a different folder.

 

For now I guess I'll have to move share B to an unassigned device since I don't actually care about it being protected.

 

I assume since that is outside of the array that I'd then be able to transfer to share A and it'd use the cache drive?

[pwm]

3 hours ago, SphericalRedundancy said:

I guess they could also implement logic to only move to cache if moving to a different drive and skip cache if moving to the same disk just a different folder.

 

But unless you use a redundant disk pool for cache, files on the cache isn't protected from media failures. So a move through cache introduces interesting problems that an user isn't likely to realize.