openstack
/
nova
Code Issues Proposed changes

Merge "Support for --force flag for nova-manage placement heal_allocations command"

This commit is contained in:
Zuul 2020-05-11 18:15:34 +00:00 committed by Gerrit Code Review
parent 7dfe999180 87936baaac
commit 5b76ae4e66
4 changed files with 152 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
doc/source/cli/nova-manage.rst
View File

@ -546,7 +546,7 @@ Placement
.. _heal_allocations_cli:
``nova-manage placement heal_allocations [--max-count <max_count>] [--verbose] [--skip-port-allocations] [--dry-run] [--instance <instance_uuid>] [--cell <cell_uuid]``
``nova-manage placement heal_allocations [--max-count <max_count>] [--verbose] [--skip-port-allocations] [--dry-run] [--instance <instance_uuid>] [--cell <cell_uuid] [--force]``
Iterates over non-cell0 cells looking for instances which do not have
allocations in the Placement service and which are not undergoing a task
state transition. For each instance found, allocations are created against
@ -606,6 +606,9 @@ Placement
Specify ``--cell`` to process heal allocations within a specific cell.
This is mutually exclusive with the ``--instance`` option.
Specify ``--force`` to forcefully heal single instance allocation. This
option needs to be passed with ``--instance``.
This command requires that the
:oslo.config:option:`api_database.connection` and
:oslo.config:group:`placement` configuration options are set. Placement API

37
nova/cmd/manage.py
View File

@ -1825,7 +1825,8 @@ class PlacementCommands(object):
def _heal_allocations_for_instance(self, ctxt, instance, node_cache,
output, placement, dry_run,
heal_port_allocations, neutron):
heal_port_allocations, neutron,
force):
"""Checks the given instance to see if it needs allocation healing
:param ctxt: cell-targeted nova.context.RequestContext
@ -1841,6 +1842,8 @@ class PlacementCommands(object):
requested, False otherwise.
:param neutron: nova.network.neutron.ClientWrapper to
communicate with Neutron
:param force: True if force healing is requested for particular
instance, False otherwise.
:return: True if allocations were created or updated for the instance,
None if nothing needed to be done
:raises: nova.exception.ComputeHostNotFound if a compute node for a
@ -1905,6 +1908,16 @@ class PlacementCommands(object):
allocations = self._heal_missing_project_and_user_id(
allocations, instance)
if force:
output(_('Force flag passed for instance %s') % instance.uuid)
need_healing = _UPDATE
# get default allocations
alloc = self._heal_missing_alloc(ctxt, instance, node_cache)
# set consumer generation of existing allocations
alloc["consumer_generation"] = allocations["consumer_generation"]
# set allocations
allocations = alloc
if heal_port_allocations:
to_heal = self._get_port_allocations_to_heal(
ctxt, instance, node_cache, placement, neutron, output)
@ -1974,7 +1987,8 @@ class PlacementCommands(object):
def _heal_instances_in_cell(self, ctxt, max_count, unlimited, output,
placement, dry_run, instance_uuid,
heal_port_allocations, neutron):
heal_port_allocations, neutron,
force):
"""Checks for instances to heal in a given cell.
:param ctxt: cell-targeted nova.context.RequestContext
@ -1991,6 +2005,8 @@ class PlacementCommands(object):
requested, False otherwise.
:param neutron: nova.network.neutron.ClientWrapper to
communicate with Neutron
:param force: True if force healing is requested for particular
instance, False otherwise.
:return: Number of instances that had allocations created.
:raises: nova.exception.ComputeHostNotFound if a compute node for a
given instance cannot be found
@ -2044,7 +2060,7 @@ class PlacementCommands(object):
for instance in instances:
if self._heal_allocations_for_instance(
ctxt, instance, node_cache, output, placement,
dry_run, heal_port_allocations, neutron):
dry_run, heal_port_allocations, neutron, force):
num_processed += 1
# Make sure we don't go over the max count. Note that we
@ -2101,9 +2117,11 @@ class PlacementCommands(object):
@args('--cell', metavar='<cell_uuid>', dest='cell_uuid',
help='Heal allocations within a specific cell. '
'The --cell and --instance options are mutually exclusive.')
@args('--force', action='store_true', dest='force', default=False,
help='Force heal allocations. Requires the --instance argument.')
def heal_allocations(self, max_count=None, verbose=False, dry_run=False,
instance_uuid=None, skip_port_allocations=False,
cell_uuid=None):
cell_uuid=None, force=False):
"""Heals instance allocations in the Placement service
Return codes:
@ -2126,9 +2144,6 @@ class PlacementCommands(object):
# for example, this could cleanup ironic instances that have
# allocations on VCPU/MEMORY_MB/DISK_GB but are now using a custom
# resource class
# - add an option to overwrite allocations for instances which already
# have allocations (but the operator thinks might be wrong?); this
# would probably only be safe with a specific instance.
# - deal with nested resource providers?
heal_port_allocations = not skip_port_allocations
@ -2144,6 +2159,11 @@ class PlacementCommands(object):
'are mutually exclusive.'))
return 127
if force and not instance_uuid:
print(_('The --instance flag is required'
'when using --force flag.'))
return 127
# TODO(mriedem): Rather than --max-count being both a total and batch
# count, should we have separate options to be specific, i.e. --total
# and --batch-size? Then --batch-size defaults to 50 and --total
@ -2220,7 +2240,8 @@ class PlacementCommands(object):
try:
num_processed += self._heal_instances_in_cell(
cctxt, limit_per_cell, unlimited, output, placement,
dry_run, instance_uuid, heal_port_allocations, neutron)
dry_run, instance_uuid, heal_port_allocations, neutron,
force)
except exception.ComputeHostNotFound as e:
print(e.format_message())
return 2

115
nova/tests/functional/test_nova_manage.py
View File

@ -783,6 +783,121 @@ class TestNovaManagePlacementHealAllocations(
self.assertIn('Found 1 candidate instances', output)
self.assertIn('Processed 0 instances.', output)
def test_heal_allocations_force_allocation(self):
"""Tests the case that a specific instance allocations are
forcefully changed.
1. create server without allocations
2. heal allocations without forcing them.
Assert the allocations match the flavor
3. update the allocations to change MEMORY_MB to not match the flavor
4. run heal allocations without --force.
Assert the allocations still have the bogus
MEMORY_MB value since they were not forcefully updated.
5. run heal allocations with --force.
Assert the allocations match the flavor again
6. run heal allocations again.
You should get rc=4 back since nothing changed.
"""
# 1. Create server that we will forcefully heal specifically.
server, rp_uuid = self._boot_and_assert_no_allocations(
self.flavor, 'cell1', volume_backed=True)
# 2. heal allocations without forcing them
result = self.cli.heal_allocations(
verbose=True, instance_uuid=server['id']
)
self.assertEqual(0, result, self.output.getvalue())
# assert the allocations match the flavor
allocs = self._get_allocations_by_server_uuid(
server['id'])[rp_uuid]['resources']
self.assertEqual(self.flavor['vcpus'], allocs['VCPU'])
self.assertEqual(self.flavor['ram'], allocs['MEMORY_MB'])
# 3. update the allocations to change MEMORY_MB
# to not match the flavor
alloc_body = {
"allocations": [
{
"resource_provider": {
"uuid": rp_uuid
},
"resources": {
"MEMORY_MB": 1024,
"VCPU": self.flavor['vcpus'],
"DISK_GB": self.flavor['disk']
}
}
]
}
self.placement_api.put('/allocations/%s' % server['id'], alloc_body)
# Check allocation to see if memory has changed
allocs = self._get_allocations_by_server_uuid(
server['id'])[rp_uuid]['resources']
self.assertEqual(self.flavor['vcpus'], allocs['VCPU'])
self.assertEqual(1024, allocs['MEMORY_MB'])
# 4. run heal allocations without --force
result = self.cli.heal_allocations(
verbose=True, instance_uuid=server['id']
)
self.assertEqual(0, result, self.output.getvalue())
self.assertIn(
'Successfully updated allocations for',
self.output.getvalue())
# assert the allocations still have the bogus memory
allocs = self._get_allocations_by_server_uuid(
server['id'])[rp_uuid]['resources']
self.assertEqual(1024, allocs['MEMORY_MB'])
# call heal without force flag
# rc should be 4 since force flag was not used.
result = self.cli.heal_allocations(
verbose=True, instance_uuid=server['id']
)
self.assertEqual(4, result, self.output.getvalue())
# call heal with force flag and dry run
result = self.cli.heal_allocations(
dry_run=True, verbose=True,
instance_uuid=server['id'],
force=True
)
self.assertEqual(4, result, self.output.getvalue())
self.assertIn(
'[dry-run] Update allocations for instance',
self.output.getvalue())
# assert nothing has changed after dry run
allocs = self._get_allocations_by_server_uuid(
server['id'])[rp_uuid]['resources']
self.assertEqual(1024, allocs['MEMORY_MB'])
# 5. run heal allocations with --force
result = self.cli.heal_allocations(
verbose=True, instance_uuid=server['id'],
force=True
)
self.assertEqual(0, result, self.output.getvalue())
self.assertIn('Force flag passed for instance',
self.output.getvalue())
self.assertIn('Successfully updated allocations',
self.output.getvalue())
# assert the allocations match the flavor again
allocs = self._get_allocations_by_server_uuid(
server['id'])[rp_uuid]['resources']
self.assertEqual(self.flavor['ram'], allocs['MEMORY_MB'])
# 6. run heal allocations again and you should get rc=4
# back since nothing changed
result = self.cli.heal_allocations(
verbose=True, instance_uuid=server['id']
)
self.assertEqual(4, result, self.output.getvalue())
class TestNovaManagePlacementHealPortAllocations(
test_servers.PortResourceRequestBasedSchedulingTestBase):

4
releasenotes/notes/force-heal-allocations-7834f3156be90c94.yaml Normal file
View File

@ -0,0 +1,4 @@
---
features:
- Add ``--force`` option to the ``nova-manage placement heal_allocations``
command to forcefully heal allocations for a specific instance.