nova glance image service needs better retries and logging

[Chris Behrens]

The glance image service in nova does not log the host:port when connections fail.
Since retries select random hosts, we could potentially retry the same hosts.

And finally, there's a bug in the test_download_with_retries test in nova/tests/image/test_glance.py

The 2nd part of the test should be testing retries.. but forgets to reset tries = [0]... so there's no exception generated that would force it to retry.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/10339

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/10339
Committed: http://github.com/openstack/nova/commit/0c59003b1cc2ce17e3215e6c69b0dc2407f1b38b
Submitter: Jenkins
Branch: master

commit 0c59003b1cc2ce17e3215e6c69b0dc2407f1b38b
Author: Chris Behrens <email address hidden>
Date: Thu Jul 26 08:59:24 2012 +0000

    Refactor glance image service code

    Fixes bug 1029397

    This adds better retry support, making sure that if we need to retry
    requests to glance, we cycle through all hosts before potentially
    attemping the hosts that failed. The old random selection could cause
    immediate retrying of the same host that failed.

    This also adds logging of the host:port that failed and fixes a bug in
    the retry test, which didn't actually successfully test retrying.

    Tests for new code added.

    Change-Id: I400616081e1e547b9ca2e0be622889d3a399a5bf