Disk resource management broken due to unit confusion

[iain MacDonnell]

Disk resource management is broken because disk space is mostly measured in Gb, but a couple of places in the NC code measure it in Mb, and the two don't mix well.

The two places are:

1) When the instance filesystem is queried for available space, the result (in bytes) is divided by BYTES_PER_DISK_UNIT, which is erroneously set to 1048576 - should be 1073741824 - so the node controller reports 1024x as much disk as it really has.

2) in handlers_default, doDescribeResource() attempts to account for swap space, based on the #define'd SWAP_SIZE, which is 512 (Mb), but the code treats it as if it was measured in Gb, so each instance consumes 512Gb+ of disk resource!

Since the definition of SWAP_SIZE is a fixed 512Mb, and everything else is measured in Gb, I've further hacked my version to count 1Gb for swap when SWAP_SIZE is less than 1024Mb (always, for now...)

[Andy Grimm (agrimm)]

This issue is now being tracked upstream at http://eucalyptus.atlassian.net/browse/EUCA-2750

Please watch that issue for further updates.