If a coverage-report directory conflicts, the error message is
incomplete like the following:
$ nova coverage-start --combine
Coverage collection started
$
$ nova coverage-report --html test01
Report path: /tmp/nova-coverage_Krl__k/test01
$
$ nova coverage-start --combine
Coverage collection started
$ nova coverage-report --html test01
ERROR: Directory conflict: %s already exists (HTTP 400)
$
This patch fixes the message.
Fixes bug #1217315
Change-Id: Ie21b0e87262a3caef36d93711ee96529ec699cec
Currently an open file like object is passed into the image service
and thus into the download modules. This is a limiting for certain
optimization (system calls to copy, hard links, external services,
out of order protocols like bittorrent, etc). This patch passes the
file name into the image service when used by the nova.virt.images.
The previous means of passing in a data stream is still supported
but cannot be used with download modules.
Change-Id: Id17d426faab8d63dcffca22b300ebac3f15c104a
blueprint: image-multiple-location
1) Child cell scheduling fails. The change to use Instance.create() in
child cells broke things. The instance properties was being overwritten
due to a typo. I also found a number of other things that need to be
popped out before passing to create_db_entry_for_new_instance().
2) Notification to the API cell was failing. We cannot pass an object to
instance_update_at_top().
3) When cells is enabled, deleting an instance hits the '_local_delete'
part of code in the API because we don't see the compute node. There's
a trap in cells_api.py to forward the delete on to child cells, but we
need to make sure not to try to deallocate network in the API cell.
Fixes bug 1216113
Change-Id: Ie989c22cb3b6a1359a0540e60e06fa4153e93476
CONF.default_flavor is only used in the EC2 API, so update the help
message to reflect that.
Partial-Bug: #1207553
Change-Id: I3f263c564b619a32a69f5170ab07eddfba77f60b
This reverts commit a4ad62ac9f.
Nova baremetal architecture scheduling was broken by it. While
workarounds exist it is backwards incompatible, and baremetal is
supported since Grizzly.
Fix bug: 1213967
Change-Id: I319b8a17f3ae7a3b527d388c6ff2954c0bcc0108
Resource tracker checking is added so that we have reasonable assurance
of being able to build the instance before it starts. This allows for a
quick reschedule if the compute node does not have available resources.
This is one part of a series of patches to build up the
build_and_run_instance method.
Part of bp query-scheduler
Change-Id: Ife6e5aef4fda46f74e2db7be1a09b210be397152
Ken'ichi Ohmichi