Tim Mackey created CLOUDSTACK-9223:
--------------------------------------
Summary: UNMAP XenServer Primary Storage prior to snapshot
Key: CLOUDSTACK-9223
URL: https://issues.apache.org/jira/browse/CLOUDSTACK-9223
Project: CloudStack
Issue Type: Improvement
Security Level: Public (Anyone can view this level - this is the default.)
Components: Storage Controller
Reporter: Tim Mackey
For Syed Mushtaq:
So I was looking around the Swift/S3 integration of SolidFire and did a bunch
of tests. When I backup a snpashot to Swift, SolidFire uploads the whole
volume's data instead of just the diff. Also since the SolidFire is unaware of
any changes that are happening on he FS inside the LUN, it always uploads the
total number of blocks allocated (instead of actual used space). For example:
if you have a 100 GB volume, you create a 100 GB file on it. This makes
SolidFire allocate blocks whose size totals 100GB. Now delete that file so the
usage in the filesystem is back to 0. However, SolidFire didn't know about that
deletion so it still keeps 100GB worth of blocks allocated to the volume. Now
if I take a snapshot and invoke the upload to Swift. SolidFire copies 100GB
(despite the fact that there is no file currently in the FS). Infact, all
future snapshots will also copy 100GB worth of data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
