We are rolling out a new OES Linux server build for our customers and I think we have hit an unfortunate design snag...

Our smallest server is 32TB and we often build servers with much larger capacity. We are no longer using NSS because of the 8TB volume limitation and are now using XFS (its also MUCH faster). However, we still have to limit the maximum volume size to less than 16 TB due to limitations (we reported last year) in the NCP architecture. As such, our server architecture requires at least two drive letters to mount the associated NCP volumes with a single 32 TB RAID array. When installing a server with multiple RAID arrays the required drive letter assignments for mapping can start to get out of hand. We are installing a customer server today with 4 x 32 TB RAID arrays and they require 8 drive letters just for the RAID storage...

So... We decided to use a single consolidated NCP Share Point one directory above the actual RAID Volume Mount points. Our goal was to only require a single drive letter to give our customers access to all their data arrays.

For example: We create a subdirectory called "raids" under "/media". We then create a new directory under "/media/raids" as a mount point for each individual RAID to be attached (raid1, raid2, raid3, etc). We set up /media/raids as an NCP share point (RAIDS:) and then the customer can see all of their individual RAID arrays as unique folder at the root of the "RAID:" volume. Reiterating:

NCP "RAIDS:" Share: /media/raids

Individual file system mount points: /media/raids/raid1, /media/raids/raid2, etc

Windows XP seems to have no problem with this. However, the Win7 client will report that insufficient space exists when trying to copy large amounts of data up to any of the RAID arrays (folders). Apparently, the free space calculation is being performed at the root of the NCP share point (the relatively small "/" file system) rather than the free space which actually exists in the target folder (on the larger RAID arrays).

Is this type of architecture simply asking too much of the NCP server? Am I missing something? I still have some testing to do to quantify the issue further, but I believe the core complication is that I have multiple volume mount points underneath a single NCP share.

This may be a bug in the Win7 client as XP does not seem to have this issue. It may also simply be that XP doesn't bother to try to pre-calculate the amount of space required before it begins a file copy.

Any additional insight is greatly appreciated!