Nova is working on support for Python 3.4. via
https://blueprints.launchpad.net/nova/+spec/nova-python3
This changeset moves from py33 to py34 in the tox.ini file to be
inline with Nova. It also incorporates changes produced from the
'sixer' tool run with 'all'.
sixer can be found at:
https://pypi.python.org/pypi/sixer
At the present time, the py34 environment does not run
successfully because it pulls in portions of Nova that have not
yet been fully converted to Python 3.
Change-Id: If84df3017a722d0ad3546e81f97e67f1a07a763f
Create new module nova_powervm.virt.powervm.mgmt containing utilities
around the PowerVM management partition.
Moved vm.get_mgmt_partition into the new module.
Change-Id: I67d88148c78a5c966e16ddb4b490a398250b2120
Live testing revealed a bug in
nova_powervm.virt.powervm.disk.driver.DiskAdapter.instance_disk_iter
whereby Adapter.read's xag parameter was a single xag rather than a
list. Fixed the bug and added UTs.
Change-Id: I9f0d78129fca97dacbea3bcbbdda25763850804e
In the disk drivers, provide the ability to map and unmap an instance's
boot disk to/from the management partition.
Change-Id: I992100da5618bb804da547e62308b83114819b0d
PowerVM systems use a technology called Resource Monitoring and Control
(RMC). This allows the hypervisor to perform dynamic operations to the
virtual machine (resize, migration, etc...).
This technology requires an Ethernet connection between the VM and the
PowerVM API. This change set enables this via two means.
- Refactoring the existing 'mgmt' nic connection into tasks. This
enables further tasks to determine if a mgmt VIF was created.
- Send an additional network info element into the config drive to
bring up the VIF on IPv6 link local.
Change-Id: I11881203d753061d7a2127fd2a939462f69d39e7
Add a method to the disk.driver API which can find the backing (usually
boot) disk associated with an instance.
This is in anticipation of the snapshot functionality, which will need
to find an instance's disk in order to map it to the management
partition.
Change-Id: I7b200fb20481c69c3cb2d5eb96a6ab1596b8ef2b
Previously, the SSP disk driver was gleaning the list of VIOS UUIDs from
the list of Nodes in the Cluster, and optionally restricting that list
to VIOSes from a specified host.
It will almost never be useful to know about VIOSes residing on a
different host. Any attempt to talk to those VIOSes through the PowerVM
API will fail, as we don't have a pypowervm connection to that host.
Therefore this change set simplifies the retrieval of VIOS UUIDs,
eliminating the host_uuid option and restricting the result to VIOSes on
the current host.
Change-Id: I2fe1e93b7039383889e56b809d5348ee2d5222a9
Implement a helper method to return the LPAR wrapper representing the
host's management partition.
This is needed for such operations as snapshot, which will need to map
a (normal) instance's storage to the management partition.
Change-Id: I0fb3f8e9191289b4778d4d7171547475dc6bcca5
There is a 15-character limit on virtual disk names, imposed by VIOS
(but not by the PowerVM REST API). There is no such limit for SSP LUs.
The recently-changed disk name generation algorithm allowed vdisk names
to exceed 15 characters, causing errors like:
Failed to perform operation on Volume Group on VIOS with ID 2 in CEC
8247-22L*1234A0A - RMC command for Virtual Disk failed with error -
Backing device names can not exceed "15" characters in length.
This change set adds an optional 'short' parameter to the private
_get_disk_name helper which causes it to limit the generated name to 15
characters. The localdisk driver is changed to use this new parameter.
Unit tests are improved such that they no longer mock _get_disk_name,
and such that they verify both of its variants.
Change-Id: I9efbb9d904e52f2a74bba7d8e631cd45862aa504
The PowerVM API introduced a new parameter to determine the management
partition. This should be excluded from the operations for VMs. This
change set updates the nova code such that it excludes that partition
from VM feeds.
Change-Id: I800e0df774403ebc62ad8a5c5df0e77bb0017b25
Previously, we were using the instance's name verbatim to generate the
name of its config ISO. Since the format of the instance name is very
permissive, this can lead to rejection of the name by the PowerVM REST
API:
Value 'Linux Test_0ce757b1_thorst_config.iso' is not facet-valid with
respect to pattern '([A-Za-z0-9_\.]{1,79})' for type 'FileName.Pattern'.
This change set exploits a new pypowervm utility to sanitize the
instance name and combine it with the desired prefix and suffix to
generate a valid config ISO name.
The same is being done for an instance's boot/rescue disks and image LUs
(for SSP).
Change-Id: Id8a77e20ede3ec3339ef5e5910e05943683a2b6c
This change updates the vscsi.py to not send the adapter down when
requesting that the scsi adapter be connected. This was due to a recent
change in pypowervm, and the refactor was missed in the nova-powervm
code.
An additional unit test was added to cover this.
Change-Id: Ic8555f49f213e3e7e46a59e1518984ef5e0f431b
The nova-powervm driver will operate on a native management partition
which manages only the host on which it resides. There is therefore no
need to specify that host's MTMS in the configuration; and it is an
error if the PowerVM REST API does not return exactly one host.
Change-Id: I0f77add7c09760a54c8d7662bfe3c6810095620b
This change set adds support for the PowerVM VNC console.
Changes the remove vterm to always be invoked. This shouldn't have a
performance impact if the console is not running. Ensures proper clean
up on the system running the PowerVM API.
Change-Id: Ia6a5dd445410fcd5351c3528003a83ba659805bb
When retrieving the vios uuids of the cluster we have to throw
out any vios that is on another node, since we won't have
their uuid or url.
Change-Id: I1235d935b8e6c75e470fdb0495d6d9d955d11b4b
Recently pypowervm added the ability to invoke Adapter.create() directly
from an EntryWrapper. This change set takes advantage of this addition.
Change-Id: Ic4e7c1f44e4952af0874354715acce20e9651a99
This change set supports the media repository creation against the
Virtual I/O Server if it has not already been created.
This fixes a long standing TODO in the nova-powervm code for initial
standup. This was enabled due to changes made back in pypowervm to
support unordered elements.
Change-Id: I84a8f80a8fdb16ba2d48a70862d2cd0058ebaaa1
Recent changes in pypowervm removed now-redundant adapter parameters
from several methods. Rebase on that change.
Change-Id: I88b5061150216359bcaab65a3c3587b622f61e14
This change set switches nova-powervm to utilize a base Session as the
authentication model for log in to the API. This assumes that the
nova-powervm process has the right user permissions to use the API, and
that it is running on the server hosting the PowerVM API.
Change-Id: I07be4249d5ab06d95fcda85c8c6c48c67f85bcf6
The get_active_vioses operation was added to the vios.py recently.
However, all calls down to it were with xags=None. Most of the
invocations want the active VIOS with the appropriate XAGs.
This change set allows for an optional parameter to get the XAGs.
Change-Id: Ied9b264b8062aa7d3594101ee5d91c9fcff98c68
Recently pypowervm removed the traits parameter from the Response
constructor, and added an adapter parameter to the various wrapper bld
methods.
Change-Id: I80fa269112ed305225d80f2cad84f80f937f44d3
This change updates the vscsi volume driver to ensure that the hdisk
removal is always invoked. The hdisk removal significantly speeds up
subsequent discoveries. The changes to pypowervm should remove risks
from running the API.
The change is done so that an error from the hdisk job does not block
volume disconnect. The volume will have already been disconnected from
the VM by the time the command is run. If the hdisk remove fails for
some reason, a subsequent attach should clean it up (but can be slower
when it does so).
Change-Id: Ib823f793ec4b951111301286743c6bd243b863fe
The initial code was gathering all of the FC ports for certain
operations. This could lead to very long delays as the inactive ports
would run operations, and would eventually time out.
This change will only return the active ports for the VIOSes, resulting
in faster operations when dealing with storage.
Change-Id: I5c10e8e91ef6936f6c6c44644de54db3b18a41d5
This change optimizes the use of the VIOSes in the driver. It does so
by:
- Removing an excess call to capture the VIOSes in the driver start up.
- Querying only for the active VIOSes within the vscsi volume driver.
This not only makes the code run quicker, it is also more correct
functionally. It no longer tries to connect to VIOSes that may be down
for maintenance purposes.
Change-Id: I718752385837c021c8d0f429ab71f4f7e27c0dd6
Added exception logic to handle the case where the LPAR
that is the target of a destroy operation may not be on
the host.
Change-Id: Ic7ec4934a85aa91f7f433dbad359229a41093ca2
A recent commit in pypowervm now requires a 'traits' parameter
to be specified on a Response object. Updated the unit tests
to supply 'None' since traits are not used in the testing.
Change-Id: I2272bb55b4575374ba93c77222531a9ce71afb3e
This change set moves the vif plugging to the start of the spawn
operation. This is so that the corresponding neutron agent sees the VIF
right after the LPAR create (rather than at the tail end of the deploy).
This allows the neutron agent to run more efficiently.
Change-Id: Ib5e5a6294131616f374778712c051da42ab2a233
Fixed up the logic to find the HSCL151B error code in the
REST response message and updated the unit test.
Change-Id: I04505a0b88d4c44746eaf3417f099e66a9fdf08a
Implement nova_powervm.virt.powervm.driver.PowerVMDriver.reboot.
This implementation heeds the reboot_type (HARD or SOFT) parameter.
Change-Id: Ieaba2b676343a488ead9e06d3710c6c934b4e727
When nova is managing an SSP Cluster whose VIOSes span multiple
accessible hosts (e.g. nova is running against a single HMC managing
multiple hosts), it is important that operations specific to a
particular instance - such as creating and removing disk connections -
be invoked only on the VIOS(es) residing on the same host as that
instance. This change set fixes the SSP Disk Driver to ensure this is
the case.
Ancillary to this change is an expansion and refactoring of the vm
utility to retrieve quick properties for an LPAR. Invoking the new
vm.get_vm_qp method without the qprop parameter will return a dictionary
of all the quick properties available for the LPAR.
Change-Id: Ib1d02743baf041c9abc69b25201d33e516fa6cb3
Because of recent pypowervm changes to default group = None the
vios.py method get_physical_wwpns must be updated to provide
xags=STORAGE.
Change-Id: I37bdc14b57536eeedb606b5b4fe6c1051ee5a888
Recent pypowervm changes moved pypowervm.tasks.hdisk.LUA_STATUS_* and
LUA_TYPE_* into enums LUAStatus and LUAType and renamed all existing
enumerations. This change set accomodates those shifts.
Change-Id: I06507d26aba138336d2e8a7480bf03501e773ec7
The NPIV flow needs both the FC mapping data as well as the backing
storage devices. Since pypowervm switched to a default of group=None,
that exposed an issue in the query in the npiv.py file where it was not
retrieving the Fibre Ports.
This update extends the XAG request to include Storage, as well as the
original FC_Mappings.
Change-Id: Iaf08b15183383fc1df8d9b294d2a18140e7b3e31
Recently, pypowervm refactored the common parts of logical_partition and
virtual_io_server into a base_partition module. This change set
accomodates that refactor.
Change-Id: Icc67697972f633bfb007df0b2328f3f2448e8062
Kyle L. Henderson