Related to blueprint xenapi-config-drive
Move the configdrive related functions to the nova.virt.configdrive
module, so that it could be used by other drivers as well. Also move the
configdrive related configuration options to the same module.
Change-Id: Ib97ab907ac820e616017f770c2f3a8366af57a0f
This removes the database instance list operation from the libvirt
imagecache module and pushes it up into compute/manager.py.
This is a step in the proposed strategy to move db usage out of the virt
drivers up into compute/manager.py in support of bp:no-db-compute.
Change-Id: I2ea39f14304ed4b44c7ac9650703a5f780411c18
API indicates the status parameter is a string, hence status=BALONEY is a valid
filter that should return zero records since no records match that status. The
return should be an empty list and 200 OK.
Fixes bug 1061712
Change-Id: If796110a18dd75ab5c5d4ba41e9c3b58180341f8
Fixes Bug 1065092
Adds a single QUOTA.rollback for all exceptions rather than performing
the rollback individually. Ensures that no stale "reservations" are left
in place after a failed instance creation.
Change-Id: I354726b651ae8feb6153d5865e0b14b54c73314b
Related to blueprint xenapi-config-drive
find_guest_agent is related to agent, thus moved to that volume. With
all the agent calls in one module, making the agent optional will be as
easy as implementing a null agent.
Change-Id: I5c8cf2fd3d08518e1794a2018429162e4bd767d7
In multi_host mode, floating ip(s) addr and NAT rules are still on
source node after resize/migrate instance. This patch fixes it up by
adding new methods in network.api to moving them to the destination
node.
Also adds the new methods to network/quantumv2/api.py. They do nothing
but pass for now.
This patch updates network RPC API to version 1.1
Fixes bug 1053344
Change-Id: If9f30050d37831f108ac4a1c8a018d820818f3b6
Handle ImportErrors's when importing the quantumv2 API
in the Nova API servers controller. This makes it possible
to run nova-api when quantumclient is not present.
This was recently broken in commit 83a8418 which changed how we
import the quantum v2 API class.
Fixes LP Bug #1070509.
Change-Id: I3c452c3bd88cb5b431cb23e525ff78e3d677e1a8
Fix the list in the qemu output and break out of the loop
when we encounter that instead of failing to parse.
Fixes bug 1070088.
Change-Id: I5f501c599b94e0ac13892c97d60ec740ac7e641d
Going forward with the early stages of no-db-compute, we will
consistently hit cases where we need to eject some database usage
from one virt driver up a few layers. Since those are not commonly
used by all, some way of determining (and the higher layer) the
properties of the virt driver being used will be useful.
We have discussed using something like a mixin interface class
and multiple inheritance. We have discussed tricks like detecting
if the manage_image_cache() method has been overridden. We have
discussed trying it once and catching NotImplementedError and then
never trying again. Aside from the first, none of these are
unsneaky enough to bite us later.
This approach will provide us a way to declare such properties
succinctly in the compute driver to help the higher layers know
what we want them to do on our behalf.
Change-Id: I74dea9322a5b4688319ebf5d9afe416e93401c58
Related to blueprint xenapi-config-drive
Move mkfs function to nova.utils, so configdrive module does not depend
on the libvirt module. This way, the configdrive module could be used
by other virt modules.
Change-Id: Ibda2aacfa0b66bc087e397df1474b5dfbe86d923
Fixes bug 1058273.
The path of the disk for snapshotting LXC instance is debugged.
Also, virt_dom.managedSave(0) and self._create_domain(domain=virt_dom)
are not called for LXC instance, because they are supported by libvirt.
Change-Id: Idf8275a0222b2eb289e16dd245f7721e17fce21d
There is an error in the logic for a unit test in the Scheduler DiskFilter.
We need to test the failure of the host_passes check with the
"test_disk_filter_fails" test.
Fixes bug 1070264.
Change-Id: Ib07f47b74a6634c573dca8fb4efa2d95b2e1ed87
Related to blueprint xenapi-volume-drivers
parse_volume_info method did too many things. It was checking the
connection_data for iscsi information, and also checked the mountpoint.
This change extracts the mountpoint conversion, and performs that check
earlier in the volume attach process.
Change-Id: I7c26e813319f6d9de287782ad188b22804d032a9
Related to blueprint xenapi-volume-drivers
The connection_info parameter, that is coming from volume contains only
two keys, 'driver_volume_type', and 'data'.
See:
https://github.com/openstack/nova/blob/master/nova/volume/manager.py#L375
parse_volume_info was checking the connection_info for 'auth_method',
and authentication parameters. These informations live inside
connection_data. This patch is fixing this, and covering with a test.
Also rename the generic data variable to connection_data, to make it
easier to read the code.
parse_volume_info thus no longer depends on the whole connection_info,
only on the connection_data.
Also clean-up some copy-paste test setup code.
Change-Id: I5289120d46176c8528c3317a650269c4f384f930
Changes since last sync:
- Added is_newer_than function
- Normalize_time() always return naive object
- Use pep8 v1.3.3
Change-Id: I47c74671907a2f64a3fdd2caf857999c33033bdd
Andrew Laski