XenAPI resize down is painfully slow

[Rick Harris]

There are a few reasons for this:

1. Copy VDI is needed so that we get a base-copy that has r/w permissions (we maybe able to optimize this away later by doing things differently

2. _copy_partition is using dd to copy the data into the blank VHD which is causing the VHD to balloon up to its max size (instead of retaining its sparseness)

3. rsync is having to send those long runs of zeros across the network

We can alleviate the problems of 2) and 3) by using a sparse_dd routing instead of plain dd. This will keep the sparseness of the VHD eliminating the need to send so much data across the network.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/4435

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/4435
Committed: http://github.com/openstack/nova/commit/82d78b979558b0e26b5159ea0f6cb205610b0260
Submitter: Jenkins
Branch: master

commit 82d78b979558b0e26b5159ea0f6cb205610b0260
Author: Rick Harris <email address hidden>
Date: Wed Feb 22 23:17:06 2012 +0000

    Speeding up resize down with sparse_copy

    Fixes bug 939084

    Change-Id: I78f7476adda427d729b67e7c32efc2a4dd2d7234