xenapi: must cleanup tar process on glance download errors

[John Garbutt]

We recently ensured there is a socket timeout on glance download errors.

However this now leaves the tar processes behind. We need to kill the tar process when these kinds of errors occur.

The code is also inconsistent, we should really do something similar for the upload code path, to ensure both behave in a similar way to any network errors.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/76169

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/76169
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=6b578665d4573b248e01ae5e0a77f200833c3b5c
Submitter: Jenkins
Branch: master

commit 6b578665d4573b248e01ae5e0a77f200833c3b5c
Author: John Garbutt <email address hidden>
Date: Tue Feb 25 11:46:39 2014 +0000

    xenapi: Cleanup tar process on glance error

    Currently, when the network connection to glance is interrupted, there
    is different behaviour on upload and download.

    This change ensures the behaviour between the two code paths is more
    consistent.

    Uploads generally need to be given more time before they timeout, so
    to keep a single timeout between upload and download, the timeout is
    increased to 90 seconds.

    At the same time, it ensures the tar process gets killed when any
    issues occur with the communication between the hypervisor and
    glance.

    Change-Id: Id5396e5d3c1052dc2979476a886412da65e08670
    Closes-Bug: #1284596