openstack/nova
Code Issues Proposed changes
nova/nova/cmd/manage.py

3683 lines
162 KiB
Python
Raw Blame History

# Copyright (c) 2011 X.commerce, a business unit of eBay Inc.
# Copyright 2010 United States Government as represented by the
# Administrator of the National Aeronautics and Space Administration.
# All Rights Reserved.
# Copyright 2013 Red Hat, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""
CLI interface for nova management.
"""
import collections
from contextlib import contextmanager
import functools
import os
import re
import sys
import time
import traceback
import typing as ty
from urllib import parse as urlparse
from dateutil import parser as dateutil_parser
from keystoneauth1 import exceptions as ks_exc
from neutronclient.common import exceptions as neutron_client_exc
from os_brick.initiator import connector
import os_resource_classes as orc
from oslo_config import cfg
from oslo_db import exception as db_exc
from oslo_log import log as logging
import oslo_messaging as messaging
from oslo_serialization import jsonutils
from oslo_utils import encodeutils
from oslo_utils import uuidutils
import prettytable
from sqlalchemy.engine import url as sqla_url
from nova.cmd import common as cmd_common
from nova.compute import api
from nova.compute import instance_actions
from nova.compute import rpcapi
import nova.conf
from nova import config
from nova import context
from nova.db import constants as db_const
from nova.db.main import api as db
from nova.db import migration
from nova import exception
from nova.i18n import _
from nova.limit import local as local_limit
from nova.limit import placement as placement_limit
from nova.network import constants
from nova.network import neutron as neutron_api
from nova import objects
from nova.objects import block_device as block_device_obj
from nova.objects import compute_node as compute_node_obj
from nova.objects import fields as obj_fields
from nova.objects import host_mapping as host_mapping_obj
from nova.objects import instance as instance_obj
from nova.objects import instance_mapping as instance_mapping_obj
from nova.objects import pci_device as pci_device_obj
from nova.objects import quotas as quotas_obj
from nova.objects import virtual_interface as virtual_interface_obj
import nova.quota
from nova import rpc
from nova.scheduler.client import report
from nova.scheduler import utils as scheduler_utils
from nova import utils
from nova import version
from nova.virt.libvirt import machine_type_utils
from nova.volume import cinder
CONF = nova.conf.CONF
LOG = logging.getLogger(__name__)
# Keep this list sorted and one entry per line for readability.
_EXTRA_DEFAULT_LOG_LEVELS = [
'nova=ERROR',
'oslo_concurrency=INFO',
'oslo_db=INFO',
'oslo_policy=INFO',
'oslo.privsep=ERROR',
'os_brick=ERROR',
]
# Consts indicating whether allocations need to be healed by creating them or
# by updating existing allocations.
_CREATE = 'create'
_UPDATE = 'update'
# Decorators for actions
args = cmd_common.args
action_description = cmd_common.action_description
def mask_passwd_in_url(url):
parsed = urlparse.urlparse(url)
safe_netloc = re.sub(':.*@', ':****@', parsed.netloc)
new_parsed = urlparse.ParseResult(
parsed.scheme, safe_netloc,
parsed.path, parsed.params,
parsed.query, parsed.fragment)
return urlparse.urlunparse(new_parsed)
def format_dict(dct, dict_property="Property", dict_value='Value',
sort_key=None):
"""Print a `dict` as a table of two columns.
:param dct: `dict` to print
:param dict_property: name of the first column
:param dict_value: header label for the value (second) column
:param sort_key: key used for sorting the dict
"""
pt = prettytable.PrettyTable([dict_property, dict_value])
pt.align = 'l'
# starting in PrettyTable 3.4.0 we need to also set the header
# as align now only applies to the data.
if hasattr(pt, 'header_align'):
pt.header_align = 'l'
for k, v in sorted(dct.items(), key=sort_key):
# convert dict to str to check length
if isinstance(v, dict):
v = str(v)
# if value has a newline, add in multiple rows
# e.g. fault with stacktrace
if v and isinstance(v, str) and r'\n' in v:
lines = v.strip().split(r'\n')
col1 = k
for line in lines:
pt.add_row([col1, line])
col1 = ''
else:
pt.add_row([k, v])
return encodeutils.safe_encode(pt.get_string()).decode()
@contextmanager
def locked_instance(cell_mapping, instance, reason):
"""Context manager to lock and unlock instance,
lock state will be restored regardless of the success or failure
of target functionality.
:param cell_mapping: instance-cell-mapping
:param instance: instance to be lock and unlock
:param reason: reason, why lock is required
"""
compute_api = api.API()
initial_state = 'locked' if instance.locked else 'unlocked'
if not instance.locked:
with context.target_cell(
context.get_admin_context(), cell_mapping) as cctxt:
compute_api.lock(cctxt, instance, reason=reason)
try:
yield
finally:
if initial_state == 'unlocked':
with context.target_cell(
context.get_admin_context(), cell_mapping) as cctxt:
compute_api.unlock(cctxt, instance)
class DbCommands(object):
"""Class for managing the main database."""
# NOTE(danms): These functions are called with a DB context and a
# count, which is the maximum batch size requested by the
# user. They must be idempotent. At most $count records should be
# migrated. The function must return a tuple of (found, done). The
# found value indicates how many unmigrated/candidate records existed in
# the database prior to the migration (either total, or up to the
# $count limit provided), and a nonzero found value may tell the user
# that there is still work to do. The done value indicates whether
# or not any records were actually migrated by the function. Thus
# if both (found, done) are nonzero, work was done and some work
# remains. If found is nonzero and done is zero, some records are
# not migratable (or don't need migrating), but all migrations that can
# complete have finished.
# NOTE(stephenfin): These names must be unique
online_migrations = (
# Added in Pike
quotas_obj.migrate_quota_limits_to_api_db,
# Added in Pike
quotas_obj.migrate_quota_classes_to_api_db,
# Added in Queens
db.migration_migrate_to_uuid,
# Added in Queens
block_device_obj.BlockDeviceMapping.populate_uuids,
# Added in Rocky
# NOTE(tssurya): This online migration is going to be backported to
# Queens and Pike since instance.avz of instances before Pike
# need to be populated if it was not specified during boot time.
instance_obj.populate_missing_availability_zones,
# Added in Rocky
instance_mapping_obj.populate_queued_for_delete,
# Added in Stein
compute_node_obj.migrate_empty_ratio,
# Added in Stein
virtual_interface_obj.fill_virtual_interface_list,
# Added in Stein
instance_mapping_obj.populate_user_id,
# Added in Victoria
pci_device_obj.PciDevice.populate_dev_uuids,
# Added in 2023.2
instance_obj.populate_instance_compute_id,
)
@args('--local_cell', action='store_true',
help='Only sync db in the local cell: do not attempt to fan-out '
'to all cells')
@args('version', metavar='VERSION', nargs='?', help='Database version')
def sync(self, version=None, local_cell=False):
"""Sync the database up to the most recent version."""
if not local_cell:
ctxt = context.RequestContext()
# NOTE(mdoff): Multiple cells not yet implemented. Currently
# fanout only looks for cell0.
try:
cell_mapping = objects.CellMapping.get_by_uuid(
ctxt, objects.CellMapping.CELL0_UUID,
)
with context.target_cell(ctxt, cell_mapping) as cctxt:
migration.db_sync(version, context=cctxt)
except exception.CellMappingNotFound:
msg = _(
'WARNING: cell0 mapping not found - not syncing cell0.'
)
print(msg)
except Exception as e:
msg = _(
'ERROR: Could not access cell0.\n'
'Has the nova_api database been created?\n'
'Has the nova_cell0 database been created?\n'
'Has "nova-manage api_db sync" been run?\n'
'Has "nova-manage cell_v2 map_cell0" been run?\n'
'Is [api_database]/connection set in nova.conf?\n'
'Is the cell0 database connection URL correct?\n'
'Error: %s'
)
print(msg % str(e))
return 1
return migration.db_sync(version)
def version(self):
"""Print the current database version."""
print(migration.db_version())
@args('--max_rows', type=int, metavar='<number>', dest='max_rows',
help='Maximum number of deleted rows to archive per table. Defaults '
'to 1000. Note that this number is a soft limit and does not '
'include the corresponding rows, if any, that are removed '
'from the API database for deleted instances.')
@args('--before', metavar='<date>',
help=('Archive rows that have been deleted before this date. '
'Accepts date strings in the default format output by the '
'``date`` command, as well as ``YYYY-MM-DD [HH:mm:ss]``.'))
@args('--verbose', action='store_true', dest='verbose', default=False,
help='Print how many rows were archived per table.')
@args('--until-complete', action='store_true', dest='until_complete',
default=False,
help=('Run continuously until all deleted rows are archived. Use '
'max_rows as a batch size for each iteration.'))
@args('--purge', action='store_true', dest='purge', default=False,
help='Purge all data from shadow tables after archive completes')
@args('--all-cells', action='store_true', dest='all_cells',
default=False, help='Run command across all cells.')
@args('--task-log', action='store_true', dest='task_log', default=False,
help=('Also archive ``task_log`` table records. Note that '
'``task_log`` records are never deleted, so archiving them '
'will move all of the ``task_log`` records up to now into the '
'shadow tables. It is recommended to also specify the '
'``--before`` option to avoid races for those consuming '
'``task_log`` record data via the '
'``/os-instance_usage_audit_log`` API (example: Telemetry).'))
@args('--sleep', type=int, metavar='<seconds>', dest='sleep',
help='The amount of time in seconds to sleep between batches when '
'``--until-complete`` is used. Defaults to 0.')
def archive_deleted_rows(
self, max_rows=1000, verbose=False,
until_complete=False, purge=False,
before=None, all_cells=False, task_log=False, sleep=0,
):
"""Move deleted rows from production tables to shadow tables.
Returns 0 if nothing was archived, 1 if some number of rows were
archived, 2 if max_rows is invalid, 3 if no connection could be
established to the API DB, 4 if before date is invalid. If automating,
this should be run continuously while the result
is 1, stopping at 0.
"""
max_rows = int(max_rows)
if max_rows < 0:
print(_("Must supply a positive value for max_rows"))
return 2
if max_rows > db_const.MAX_INT:
print(_('max rows must be <= %(max_value)d') %
{'max_value': db_const.MAX_INT})
return 2
ctxt = context.get_admin_context()
try:
# NOTE(tssurya): This check has been added to validate if the API
# DB is reachable or not as this is essential for purging the
# related API database records of the deleted instances.
cell_mappings = objects.CellMappingList.get_all(ctxt)
except db_exc.CantStartEngineError:
print(_('Failed to connect to API DB so aborting this archival '
'attempt. Please check your config file to make sure that '
'[api_database]/connection is set and run this '
'command again.'))
return 3
if before:
try:
before_date = dateutil_parser.parse(before, fuzzy=True)
except ValueError as e:
print(_('Invalid value for --before: %s') % e)
return 4
else:
before_date = None
table_to_rows_archived = {}
if until_complete and verbose:
sys.stdout.write(_('Archiving') + '..') # noqa
interrupt = False
if all_cells:
# Sort first by cell name, then by table:
# +--------------------------------+-------------------------+
# | Table | Number of Rows Archived |
# +--------------------------------+-------------------------+
# | cell0.block_device_mapping | 1 |
# | cell1.block_device_mapping | 1 |
# | cell1.instance_actions | 2 |
# | cell1.instance_actions_events | 2 |
# | cell2.block_device_mapping | 1 |
# | cell2.instance_actions | 2 |
# | cell2.instance_actions_events | 2 |
# ...
def sort_func(item):
cell_name, table = item[0].split('.')
return cell_name, table
print_sort_func = sort_func
else:
cell_mappings = [None]
print_sort_func = None
total_rows_archived = 0
for cell_mapping in cell_mappings:
# NOTE(Kevin_Zheng): No need to calculate limit for each
# cell if until_complete=True.
# We need not adjust max rows to avoid exceeding a specified total
# limit because with until_complete=True, we have no total limit.
if until_complete:
max_rows_to_archive = max_rows
elif max_rows > total_rows_archived:
# We reduce the max rows to archive based on what we've
# archived so far to avoid potentially exceeding the specified
# total limit.
max_rows_to_archive = max_rows - total_rows_archived
else:
break
# If all_cells=False, cell_mapping is None
with context.target_cell(ctxt, cell_mapping) as cctxt:
cell_name = cell_mapping.name if cell_mapping else None
try:
rows_archived = self._do_archive(
table_to_rows_archived,
cctxt,
max_rows_to_archive,
until_complete,
verbose,
before_date,
cell_name,
task_log,
sleep)
except KeyboardInterrupt:
interrupt = True
break
# TODO(melwitt): Handle skip/warn for unreachable cells. Note
# that cell_mappings = [None] if not --all-cells
total_rows_archived += rows_archived
if until_complete and verbose:
if interrupt:
print('.' + _('stopped')) # noqa
else:
print('.' + _('complete')) # noqa
if verbose:
if table_to_rows_archived:
print(format_dict(
table_to_rows_archived,
dict_property=_('Table'),
dict_value=_('Number of Rows Archived'),
sort_key=print_sort_func,
))
else:
print(_('Nothing was archived.'))
if table_to_rows_archived and purge:
if verbose:
print(_('Rows were archived, running purge...'))
self.purge(purge_all=True, verbose=verbose, all_cells=all_cells)
# NOTE(danms): Return nonzero if we archived something
return int(bool(table_to_rows_archived))
def _do_archive(
self, table_to_rows_archived, cctxt, max_rows,
until_complete, verbose, before_date, cell_name, task_log, sleep,
):
"""Helper function for archiving deleted rows for a cell.
This will archive deleted rows for a cell database and remove the
associated API database records for deleted instances.
:param table_to_rows_archived: Dict tracking the number of rows
archived by <cell_name>.<table name>. Example:
{'cell0.instances': 2,
'cell1.instances': 5}
:param cctxt: Cell-targeted nova.context.RequestContext if archiving
across all cells
:param max_rows: Maximum number of deleted rows to archive per table.
Note that this number is a soft limit and does not include the
corresponding rows, if any, that are removed from the API database
for deleted instances.
:param until_complete: Whether to run continuously until all deleted
rows are archived
:param verbose: Whether to print how many rows were archived per table
:param before_date: Archive rows that were deleted before this date
:param cell_name: Name of the cell or None if not archiving across all
cells
:param task_log: Whether to archive task_log table rows
:param sleep: The amount of time in seconds to sleep between batches
when ``until_complete`` is True.
"""
ctxt = context.get_admin_context()
while True:
# table_to_rows = {table_name: number_of_rows_archived}
# deleted_instance_uuids = ['uuid1', 'uuid2', ...]
table_to_rows, deleted_instance_uuids, total_rows_archived = \
db.archive_deleted_rows(
cctxt, max_rows, before=before_date, task_log=task_log)
for table_name, rows_archived in table_to_rows.items():
if cell_name:
table_name = cell_name + '.' + table_name
table_to_rows_archived.setdefault(table_name, 0)
table_to_rows_archived[table_name] += rows_archived
# deleted_instance_uuids does not necessarily mean that any
# instances rows were archived because it is obtained by a query
# separate from the archive queries. For example, if a
# DBReferenceError was raised while processing the instances table,
# we would have skipped the table and had 0 rows archived even
# though deleted instances rows were found.
instances_archived = table_to_rows.get('instances', 0)
if deleted_instance_uuids and instances_archived:
table_to_rows_archived.setdefault(
'API_DB.instance_mappings', 0)
table_to_rows_archived.setdefault(
'API_DB.request_specs', 0)
table_to_rows_archived.setdefault(
'API_DB.instance_group_member', 0)
deleted_mappings = objects.InstanceMappingList.destroy_bulk(
ctxt, deleted_instance_uuids)
table_to_rows_archived[
'API_DB.instance_mappings'] += deleted_mappings
deleted_specs = objects.RequestSpec.destroy_bulk(
ctxt, deleted_instance_uuids)
table_to_rows_archived[
'API_DB.request_specs'] += deleted_specs
deleted_group_members = (
objects.InstanceGroup.destroy_members_bulk(
ctxt, deleted_instance_uuids))
table_to_rows_archived[
'API_DB.instance_group_member'] += deleted_group_members
# If we're not archiving until there is nothing more to archive, we
# have reached max_rows in this cell DB or there was nothing to
# archive. We check the values() in case we get something like
# table_to_rows = {'instances': 0} back somehow.
if not until_complete or not any(table_to_rows.values()):
break
if verbose:
sys.stdout.write('.')
# Optionally sleep between batches to throttle the archiving.
time.sleep(sleep)
return total_rows_archived
@args('--before', metavar='<before>', dest='before',
help='If specified, purge rows from shadow tables that are older '
'than this. Accepts date strings in the default format output '
'by the ``date`` command, as well as ``YYYY-MM-DD '
'[HH:mm:ss]``.')
@args('--all', dest='purge_all', action='store_true',
help='Purge all rows in the shadow tables')
@args('--verbose', dest='verbose', action='store_true', default=False,
help='Print information about purged records')
@args('--all-cells', dest='all_cells', action='store_true', default=False,
help='Run against all cell databases')
def purge(self, before=None, purge_all=False, verbose=False,
all_cells=False):
if before is None and purge_all is False:
print(_('Either --before or --all is required'))
return 1
if before:
try:
before_date = dateutil_parser.parse(before, fuzzy=True)
except ValueError as e:
print(_('Invalid value for --before: %s') % e)
return 2
else:
before_date = None
def status(msg):
if verbose:
print('%s: %s' % (identity, msg))
deleted = 0
admin_ctxt = context.get_admin_context()
if all_cells:
try:
cells = objects.CellMappingList.get_all(admin_ctxt)
except db_exc.DBError:
print(_('Unable to get cell list from API DB. '
'Is it configured?'))
return 4
for cell in cells:
identity = _('Cell %s') % cell.identity
with context.target_cell(admin_ctxt, cell) as cctxt:
deleted += db.purge_shadow_tables(
cctxt, before_date, status_fn=status)
else:
identity = _('DB')
deleted = db.purge_shadow_tables(
admin_ctxt, before_date, status_fn=status)
if deleted:
return 0
else:
return 3
def _run_migration(self, ctxt, max_count):
ran = 0
exceptions = False
migrations = {}
for migration_meth in self.online_migrations:
count = max_count - ran
try:
found, done = migration_meth(ctxt, count)
except Exception:
msg = (_("Error attempting to run %(method)s") % dict(
method=migration_meth))
print(msg)
LOG.exception(msg)
exceptions = True
found = done = 0
name = migration_meth.__name__
if found:
print(_('%(total)i rows matched query %(meth)s, %(done)i '
'migrated') % {'total': found,
'meth': name,
'done': done})
# This is the per-migration method result for this batch, and
# _run_migration will either continue on to the next migration,
# or stop if up to this point we've processed max_count of
# records across all migration methods.
migrations[name] = found, done
if max_count is not None:
ran += done
if ran >= max_count:
break
return migrations, exceptions
@args('--max-count', metavar='<number>', dest='max_count',
help='Maximum number of objects to consider')
def online_data_migrations(self, max_count=None):
ctxt = context.get_admin_context()
if max_count is not None:
try:
max_count = int(max_count)
except ValueError:
max_count = -1
unlimited = False
if max_count < 1:
print(_('Must supply a positive value for max_number'))
return 127
else:
unlimited = True
max_count = 50
print(_('Running batches of %i until complete') % max_count)
ran = None
migration_info = {}
exceptions = False
while ran is None or ran != 0:
migrations, exceptions = self._run_migration(ctxt, max_count)
ran = 0
# For each batch of migration method results, build the cumulative
# set of results.
for name in migrations:
migration_info.setdefault(name, (0, 0))
migration_info[name] = (
migration_info[name][0] + migrations[name][0],
migration_info[name][1] + migrations[name][1],
)
ran += migrations[name][1]
if not unlimited:
break
t = prettytable.PrettyTable([_('Migration'),
_('Total Needed'), # Really: Total Found
_('Completed')])
for name in sorted(migration_info.keys()):
info = migration_info[name]
t.add_row([name, info[0], info[1]])
print(t)
# NOTE(imacdonn): In the "unlimited" case, the loop above will only
# terminate when all possible migrations have been effected. If we're
# still getting exceptions, there's a problem that requires
# intervention. In the max-count case, exceptions are only considered
# fatal if no work was done by any other migrations ("not ran"),
# because otherwise work may still remain to be done, and that work
# may resolve dependencies for the failing migrations.
if exceptions and (unlimited or not ran):
print(_("Some migrations failed unexpectedly. Check log for "
"details."))
return 2
# TODO(mriedem): Potentially add another return code for
# "there are more migrations, but not completable right now"
return ran and 1 or 0
@args('--ironic-node-uuid', metavar='<uuid>', dest='compute_node_uuid',
help='UUID of Ironic node to be moved between services')
@args('--destination-host', metavar='<host>',
dest='destination_service_host',
help='Destination ironic nova-compute service CONF.host')
def ironic_compute_node_move(self, compute_node_uuid,
destination_service_host):
ctxt = context.get_admin_context()
destination_service = objects.Service.get_by_compute_host(
ctxt, destination_service_host)
if destination_service.forced_down:
raise exception.NovaException(
"Destination compute is forced down!")
target_compute_node = objects.ComputeNode.get_by_uuid(
ctxt, compute_node_uuid)
source_service = objects.Service.get_by_id(
ctxt, target_compute_node.service_id)
if not source_service.forced_down:
raise exception.NovaException(
"Source service is not yet forced down!")
instances = objects.InstanceList.get_by_host_and_node(
ctxt, target_compute_node.host,
target_compute_node.hypervisor_hostname)
if len(instances) > 1:
raise exception.NovaException(
"Found an ironic host with more than one instance! "
"Please delete all Nova instances that do not match "
"the instance uuid recorded on the Ironic node.")
target_compute_node.service_id = destination_service.id
target_compute_node.host = destination_service.host
target_compute_node.save()
for instance in instances:
# this is a bit like evacuate, except no need to rebuild
instance.host = destination_service.host
instance.save()
class ApiDbCommands(object):
"""Class for managing the api database."""
def __init__(self):
pass
@args('version', metavar='VERSION', nargs='?', help='Database version')
def sync(self, version=None):
"""Sync the database up to the most recent version."""
return migration.db_sync(version, database='api')
def version(self):
"""Print the current database version."""
print(migration.db_version(database='api'))
class CellV2Commands(object):
"""Commands for managing cells v2."""
def _validate_transport_url(self, transport_url, warn_about_none=True):
if not transport_url:
if not CONF.transport_url:
if warn_about_none:
print(_(
'Must specify --transport-url if '
'[DEFAULT]/transport_url is not set in the '
'configuration file.'))
return None
print(_('--transport-url not provided in the command line, '
'using the value [DEFAULT]/transport_url from the '
'configuration file'))
transport_url = CONF.transport_url
try:
messaging.TransportURL.parse(conf=CONF,
url=objects.CellMapping.format_mq_url(
transport_url))
except (messaging.InvalidTransportURL, ValueError) as e:
print(_('Invalid transport URL: %s') % str(e))
return None
return transport_url
def _validate_database_connection(
self, database_connection, warn_about_none=True):
if not database_connection:
if not CONF.database.connection:
if warn_about_none:
print(_(
'Must specify --database_connection if '
'[database]/connection is not set in the '
'configuration file.'))
return None
print(_('--database_connection not provided in the command line, '
'using the value [database]/connection from the '
'configuration file'))
return CONF.database.connection
return database_connection
def _non_unique_transport_url_database_connection_checker(self, ctxt,
cell_mapping, transport_url, database_connection):
for cell in objects.CellMappingList.get_all(ctxt):
if cell_mapping and cell.uuid == cell_mapping.uuid:
# If we're looking for a specific cell, then don't check
# that one for same-ness to allow idempotent updates
continue
if (cell.database_connection == database_connection or
cell.transport_url == transport_url):
print(_('The specified transport_url and/or '
'database_connection combination already exists '
'for another cell with uuid %s.') % cell.uuid)
return True
return False
@args('--transport-url', metavar='<transport_url>', dest='transport_url',
help='The transport url for the cell message queue')
def simple_cell_setup(self, transport_url=None):
"""Simple cellsv2 setup.
This simplified command is for use by existing non-cells users to
configure the default environment. Returns 0 if setup is completed (or
has already been done) and 1 if no hosts are reporting (and this cannot
be mapped).
"""
transport_url = self._validate_transport_url(transport_url)
if not transport_url:
return 1
ctxt = context.RequestContext()
try:
cell0_mapping = self._map_cell0()
except db_exc.DBDuplicateEntry:
print(_('Cell0 is already setup'))
cell0_mapping = objects.CellMapping.get_by_uuid(
ctxt, objects.CellMapping.CELL0_UUID)
# Run migrations so cell0 is usable
with context.target_cell(ctxt, cell0_mapping) as cctxt:
try:
migration.db_sync(None, context=cctxt)
except db_exc.DBError as ex:
print(_('Unable to sync cell0 schema: %s') % ex)
cell_uuid = self._map_cell_and_hosts(transport_url)
if cell_uuid is None:
# There are no compute hosts which means no cell_mapping was
# created. This should also mean that there are no instances.
return 1
self.map_instances(cell_uuid)
return 0
@args('--database_connection',
metavar='<database_connection>',
help='The database connection url for cell0. '
'This is optional. If not provided, a standard database '
'connection will be used based on the main database connection '
'from the Nova configuration.'
)
def map_cell0(self, database_connection=None):
"""Create a cell mapping for cell0.
cell0 is used for instances that have not been scheduled to any cell.
This generally applies to instances that have encountered an error
before they have been scheduled.
This command creates a cell mapping for this special cell which
requires a database to store the instance data.
Returns 0 if cell0 created successfully or already setup.
"""
try:
self._map_cell0(database_connection=database_connection)
except db_exc.DBDuplicateEntry:
print(_('Cell0 is already setup'))
return 0
def _map_cell0(self, database_connection=None):
"""Facilitate creation of a cell mapping for cell0.
See map_cell0 for more.
"""
def cell0_default_connection():
# If no database connection is provided one is generated
# based on the database connection url.
# The cell0 database will use the same database scheme and
# netloc as the main database, with a related path.
# NOTE(sbauza): The URL has to be RFC1738 compliant in order to
# be usable by sqlalchemy.
connection = CONF.database.connection
# sqlalchemy has a nice utility for parsing database connection
# URLs so we use that here to get the db name so we don't have to
# worry about parsing and splitting a URL which could have special
# characters in the password, which makes parsing a nightmare.
url = sqla_url.make_url(connection)
url = url.set(database=url.database + '_cell0')
return urlparse.unquote(url.render_as_string(hide_password=False))
dbc = database_connection or cell0_default_connection()
ctxt = context.RequestContext()
# A transport url of 'none://' is provided for cell0. RPC should not
# be used to access cell0 objects. Cells transport switching will
# ignore any 'none' transport type.
cell_mapping = objects.CellMapping(
ctxt, uuid=objects.CellMapping.CELL0_UUID, name="cell0",
transport_url="none:///",
database_connection=dbc)
cell_mapping.create()
return cell_mapping
def _get_and_map_instances(self, ctxt, cell_mapping, limit, marker):
filters = {}
with context.target_cell(ctxt, cell_mapping) as cctxt:
instances = objects.InstanceList.get_by_filters(
cctxt.elevated(read_deleted='yes'), filters,
sort_key='created_at', sort_dir='asc', limit=limit,
marker=marker)
for instance in instances:
try:
mapping = objects.InstanceMapping(ctxt)
mapping.instance_uuid = instance.uuid
mapping.cell_mapping = cell_mapping
mapping.project_id = instance.project_id
mapping.user_id = instance.user_id
mapping.create()
except db_exc.DBDuplicateEntry:
continue
if len(instances) == 0 or len(instances) < limit:
# We've hit the end of the instances table
marker = None
else:
marker = instances[-1].uuid
return marker
@args('--cell_uuid', metavar='<cell_uuid>', dest='cell_uuid',
required=True,
help='Unmigrated instances will be mapped to the cell with the '
'uuid provided.')
@args('--max-count', metavar='<max_count>', dest='max_count',
help='Maximum number of instances to map. If not set, all instances '
'in the cell will be mapped in batches of 50. If you have a '
'large number of instances, consider specifying a custom value '
'and run the command until it exits with 0.')
@args('--reset', action='store_true', dest='reset_marker',
help='The command will start from the beginning as opposed to the '
'default behavior of starting from where the last run '
'finished')
def map_instances(self, cell_uuid, max_count=None, reset_marker=None):
"""Map instances into the provided cell.
Instances in the nova database of the provided cell (nova database
info is obtained from the nova-api database) will be queried from
oldest to newest and if unmapped, will be mapped to the provided cell.
A max-count can be set on the number of instance to map in a single
run. Repeated runs of the command will start from where the last run
finished so it is not necessary to increase max-count to finish. A
reset option can be passed which will reset the marker, thus making the
command start from the beginning as opposed to the default behavior of
starting from where the last run finished. An exit code of 0 indicates
that all instances have been mapped.
"""
# NOTE(stephenfin): The support for batching in this command relies on
# a bit of a hack. We initially process N instance-cell mappings, where
# N is the value of '--max-count' if provided else 50. To ensure we
# can continue from N on the next iteration, we store a instance-cell
# mapping object with a special name and the UUID of the last
# instance-cell mapping processed (N - 1) in munged form. On the next
# iteration, we search for the special name and unmunge the UUID to
# pick up where we left off. This is done until all mappings are
# processed. The munging is necessary as there's a unique constraint on
# the UUID field and we need something reversible. For more
# information, see commit 9038738d0.
if max_count is not None:
try:
max_count = int(max_count)
except ValueError:
max_count = -1
map_all = False
if max_count < 1:
print(_('Must supply a positive value for max-count'))
return 127
else:
map_all = True
max_count = 50
ctxt = context.RequestContext()
marker_project_id = 'INSTANCE_MIGRATION_MARKER'
# Validate the cell exists, this will raise if not
cell_mapping = objects.CellMapping.get_by_uuid(ctxt, cell_uuid)
# Check for a marker from a previous run
marker_mapping = objects.InstanceMappingList.get_by_project_id(ctxt,
marker_project_id)
if len(marker_mapping) == 0:
marker = None
else:
# There should be only one here
marker = marker_mapping[0].instance_uuid.replace(' ', '-')
if reset_marker:
marker = None
marker_mapping[0].destroy()
next_marker = True
while next_marker is not None:
next_marker = self._get_and_map_instances(ctxt, cell_mapping,
max_count, marker)
marker = next_marker
if not map_all:
break
if next_marker:
# Don't judge me. There's already an InstanceMapping with this UUID
# so the marker needs to be non destructively modified.
next_marker = next_marker.replace('-', ' ')
# This is just the marker record, so set user_id to the special
# marker name as well.
objects.InstanceMapping(ctxt, instance_uuid=next_marker,
project_id=marker_project_id,
user_id=marker_project_id).create()
return 1
return 0
def _map_cell_and_hosts(self, transport_url, name=None, verbose=False):
ctxt = context.RequestContext()
cell_mapping_uuid = cell_mapping = None
# First, try to detect if a CellMapping has already been created
compute_nodes = objects.ComputeNodeList.get_all(ctxt)
if not compute_nodes:
print(_('No hosts found to map to cell, exiting.'))
return None
missing_nodes = set()
for compute_node in compute_nodes:
try:
host_mapping = objects.HostMapping.get_by_host(
ctxt, compute_node.host)
except exception.HostMappingNotFound:
missing_nodes.add(compute_node.host)
else:
if verbose:
print(_(
'Host %(host)s is already mapped to cell %(uuid)s'
) % {'host': host_mapping.host,
'uuid': host_mapping.cell_mapping.uuid})
# Re-using the existing UUID in case there is already a mapping
# NOTE(sbauza): There could be possibly multiple CellMappings
# if the operator provides another configuration file and moves
# the hosts to another cell v2, but that's not really something
# we should support.
cell_mapping_uuid = host_mapping.cell_mapping.uuid
if not missing_nodes:
print(_('All hosts are already mapped to cell(s).'))
return cell_mapping_uuid
# Create the cell mapping in the API database
if cell_mapping_uuid is not None:
cell_mapping = objects.CellMapping.get_by_uuid(
ctxt, cell_mapping_uuid)
if cell_mapping is None:
cell_mapping_uuid = uuidutils.generate_uuid()
cell_mapping = objects.CellMapping(
ctxt, uuid=cell_mapping_uuid, name=name,
transport_url=transport_url,
database_connection=CONF.database.connection)
cell_mapping.create()
# Pull the hosts from the cell database and create the host mappings
for compute_host in missing_nodes:
host_mapping = objects.HostMapping(
ctxt, host=compute_host, cell_mapping=cell_mapping)
host_mapping.create()
if verbose:
print(cell_mapping_uuid)
return cell_mapping_uuid
@args('--transport-url', metavar='<transport_url>', dest='transport_url',
help='The transport url for the cell message queue')
@args('--name', metavar='<cell_name>', help='The name of the cell')
@args('--verbose', action='store_true',
help='Output the cell mapping uuid for any newly mapped hosts.')
def map_cell_and_hosts(self, transport_url=None, name=None, verbose=False):
"""EXPERIMENTAL. Create a cell mapping and host mappings for a cell.
Users not dividing their cloud into multiple cells will be a single
cell v2 deployment and should specify:
nova-manage cell_v2 map_cell_and_hosts --config-file <nova.conf>
Users running multiple cells can add a cell v2 by specifying:
nova-manage cell_v2 map_cell_and_hosts --config-file <cell nova.conf>
"""
transport_url = self._validate_transport_url(transport_url)
if not transport_url:
return 1
self._map_cell_and_hosts(transport_url, name, verbose)
# online_data_migrations established a pattern of 0 meaning everything
# is done, 1 means run again to do more work. This command doesn't do
# partial work so 0 is appropriate.
return 0
@args('--uuid', metavar='<instance_uuid>', dest='uuid', required=True,
help=_('The instance UUID to verify'))
@args('--quiet', action='store_true', dest='quiet',
help=_('Do not print anything'))
def verify_instance(self, uuid, quiet=False):
"""Verify instance mapping to a cell.
This command is useful to determine if the cellsv2 environment is
properly setup, specifically in terms of the cell, host, and instance
mapping records required.
This prints one of three strings (and exits with a code) indicating
whether the instance is successfully mapped to a cell (0), is unmapped
due to an incomplete upgrade (1), unmapped due to normally transient
state (2), it is a deleted instance which has instance mapping (3),
or it is an archived instance which still has an instance mapping (4).
"""
def say(string):
if not quiet:
print(string)
ctxt = context.get_admin_context()
try:
mapping = objects.InstanceMapping.get_by_instance_uuid(
ctxt, uuid)
except exception.InstanceMappingNotFound:
say('Instance %s is not mapped to a cell '
'(upgrade is incomplete) or instance '
'does not exist' % uuid)
return 1
if mapping.cell_mapping is None:
say('Instance %s is not mapped to a cell' % uuid)
return 2
else:
with context.target_cell(ctxt, mapping.cell_mapping) as cctxt:
try:
instance = objects.Instance.get_by_uuid(cctxt, uuid)
except exception.InstanceNotFound:
try:
el_ctx = cctxt.elevated(read_deleted='yes')
instance = objects.Instance.get_by_uuid(el_ctx, uuid)
# instance is deleted
if instance:
say('The instance with uuid %s has been deleted.'
% uuid)
say('Execute '
'`nova-manage db archive_deleted_rows` '
'command to archive this deleted '
'instance and remove its instance_mapping.')
return 3
except exception.InstanceNotFound:
# instance is archived
say('The instance with uuid %s has been archived.'
% uuid)
say('However its instance_mapping remains.')
return 4
# instance is alive and mapped to a cell
say('Instance %s is in cell: %s (%s)' % (
uuid,
mapping.cell_mapping.name,
mapping.cell_mapping.uuid))
return 0
@args('--cell_uuid', metavar='<cell_uuid>', dest='cell_uuid',
help='If provided only this cell will be searched for new hosts to '
'map.')
@args('--verbose', action='store_true',
help=_('Provide detailed output when discovering hosts.'))
@args('--strict', action='store_true',
help=_('Considered successful (exit code 0) only when an unmapped '
'host is discovered. Any other outcome will be considered a '
'failure (non-zero exit code).'))
@args('--by-service', action='store_true', default=False,
dest='by_service',
help=_('Discover hosts by service instead of compute node'))
def discover_hosts(self, cell_uuid=None, verbose=False, strict=False,
by_service=False):
"""Searches cells, or a single cell, and maps found hosts.
When a new host is added to a deployment it will add a service entry
to the db it's configured to use. This command will check the db for
each cell, or a single one if passed in, and map any hosts which are
not currently mapped. If a host is already mapped nothing will be done.
This command should be run once after all compute hosts have been
deployed and should not be run in parallel. When run in parallel,
the commands will collide with each other trying to map the same hosts
in the database at the same time.
"""
def status_fn(msg):
if verbose:
print(msg)
ctxt = context.RequestContext()
try:
hosts = host_mapping_obj.discover_hosts(ctxt, cell_uuid, status_fn,
by_service)
except exception.HostMappingExists as exp:
print(_('ERROR: Duplicate host mapping was encountered. This '
'command should be run once after all compute hosts have '
'been deployed and should not be run in parallel. When '
'run in parallel, the commands will collide with each '
'other trying to map the same hosts in the database at '
'the same time. Error: %s') % exp)
return 2
# discover_hosts will return an empty list if no hosts are discovered
if strict:
return int(not hosts)
@action_description(
_("Add a new cell to nova API database. "
"DB and MQ urls can be provided directly "
"or can be taken from config. The result is cell uuid."))
@args('--name', metavar='<cell_name>', help=_('The name of the cell'))
@args('--database_connection', metavar='<database_connection>',
dest='database_connection',
help=_('The database url for the cell database'))
@args('--transport-url', metavar='<transport_url>', dest='transport_url',
help=_('The transport url for the cell message queue'))
@args('--verbose', action='store_true',
help=_('Output the uuid of the created cell'))
@args('--disabled', action='store_true',
help=_('To create a pre-disabled cell.'))
def create_cell(self, name=None, database_connection=None,
transport_url=None, verbose=False, disabled=False):
ctxt = context.get_context()
transport_url = self._validate_transport_url(transport_url)
if not transport_url:
return 1
database_connection = self._validate_database_connection(
database_connection)
if not database_connection:
return 1
if (self._non_unique_transport_url_database_connection_checker(ctxt,
None, transport_url, database_connection)):
return 2
cell_mapping_uuid = uuidutils.generate_uuid()
cell_mapping = objects.CellMapping(
ctxt,
uuid=cell_mapping_uuid, name=name,
transport_url=transport_url,
database_connection=database_connection,
disabled=disabled)
cell_mapping.create()
if verbose:
print(cell_mapping_uuid)
return 0
@args('--verbose', action='store_true',
help=_('Show sensitive details, such as passwords'))
def list_cells(self, verbose=False):
"""Lists the v2 cells in the deployment.
By default the cell name, uuid, disabled state, masked transport
URL and database connection details are shown. Use the --verbose
option to see transport URL and database connection with their
sensitive details.
"""
cell_mappings = objects.CellMappingList.get_all(
context.get_admin_context())
field_names = [_('Name'), _('UUID'), _('Transport URL'),
_('Database Connection'), _('Disabled')]
t = prettytable.PrettyTable(field_names)
for cell in sorted(cell_mappings,
# CellMapping.name is optional
key=lambda _cell: _cell.name or ''):
fields = [cell.name or '', cell.uuid]
if verbose:
fields.extend([cell.transport_url, cell.database_connection])
else:
fields.extend([
mask_passwd_in_url(cell.transport_url),
mask_passwd_in_url(cell.database_connection)])
fields.extend([cell.disabled])
t.add_row(fields)
print(t)
return 0
@args('--force', action='store_true', default=False,
help=_('Delete hosts and instance_mappings that belong '
'to the cell as well.'))
@args('--cell_uuid', metavar='<cell_uuid>', dest='cell_uuid',
required=True, help=_('The uuid of the cell to delete.'))
def delete_cell(self, cell_uuid, force=False):
"""Delete an empty cell by the given uuid.
This command will return a non-zero exit code in the following cases.
* The cell is not found by uuid.
* It has hosts and force is False.
* It has instance mappings and force is False.
If force is True and the cell has hosts and/or instance_mappings, they
are deleted as well (as long as there are no living instances).
Returns 0 in the following cases.
* The empty cell is found and deleted successfully.
* The cell has hosts and force is True then the cell, hosts and
instance_mappings are deleted successfully; if there are no
living instances.
"""
ctxt = context.get_admin_context()
# Find the CellMapping given the uuid.
try:
cell_mapping = objects.CellMapping.get_by_uuid(ctxt, cell_uuid)
except exception.CellMappingNotFound:
print(_('Cell with uuid %s was not found.') % cell_uuid)
return 1
# Check to see if there are any HostMappings for this cell.
host_mappings = objects.HostMappingList.get_by_cell_id(
ctxt, cell_mapping.id)
nodes = []
if host_mappings:
if not force:
print(_('There are existing hosts mapped to cell with uuid '
'%s.') % cell_uuid)
return 2
# We query for the compute nodes in the cell,
# so that they can be unmapped.
with context.target_cell(ctxt, cell_mapping) as cctxt:
nodes = objects.ComputeNodeList.get_all(cctxt)
# Check to see if there are any InstanceMappings for this cell.
instance_mappings = objects.InstanceMappingList.get_by_cell_id(
ctxt, cell_mapping.id)
if instance_mappings:
with context.target_cell(ctxt, cell_mapping) as cctxt:
instances = objects.InstanceList.get_all(cctxt)
if instances:
# There are instances in the cell.
print(_('There are existing instances mapped to cell with '
'uuid %s.') % cell_uuid)
return 3
else:
if not force:
# There are no instances in the cell but the records remain
# in the 'instance_mappings' table.
print(_("There are instance mappings to cell with uuid "
"%s, but all instances have been deleted "
"in the cell.") % cell_uuid)
print(_("So execute 'nova-manage db archive_deleted_rows' "
"to delete the instance mappings."))
return 4
# Delete instance_mappings of the deleted instances
for instance_mapping in instance_mappings:
instance_mapping.destroy()
# Unmap the compute nodes so that they can be discovered
# again in future, if needed.
for node in nodes:
node.mapped = 0
node.save()
# Delete hosts mapped to the cell.
for host_mapping in host_mappings:
host_mapping.destroy()
# There are no hosts or instances mapped to the cell so delete it.
cell_mapping.destroy()
return 0
@args('--cell_uuid', metavar='<cell_uuid>', dest='cell_uuid',
required=True, help=_('The uuid of the cell to update.'))
@args('--name', metavar='<cell_name>', dest='name',
help=_('Set the cell name.'))
@args('--transport-url', metavar='<transport_url>', dest='transport_url',
help=_('Set the cell transport_url. NOTE that running nodes '
'will not see the change until restart!'))
@args('--database_connection', metavar='<database_connection>',
dest='db_connection',
help=_('Set the cell database_connection. NOTE that running nodes '
'will not see the change until restart!'))
@args('--disable', action='store_true', dest='disable',
help=_('Disables the cell. Note that the scheduling will be blocked '
'to this cell until its enabled and followed by a SIGHUP of '
'nova-scheduler service.'))
@args('--enable', action='store_true', dest='enable',
help=_('Enables the cell. Note that this makes a disabled cell '
'available for scheduling after a SIGHUP of the '
'nova-scheduler service'))
def update_cell(self, cell_uuid, name=None, transport_url=None,
db_connection=None, disable=False, enable=False):
"""Updates the properties of a cell by the given uuid.
If the cell is not found by uuid, this command will return an exit
code of 1. If the provided transport_url or/and database_connection
is/are same as another cell, this command will return an exit code
of 3. If the properties cannot be set, this will return 2. If an
attempt is made to disable and enable a cell at the same time, this
command will exit with a return code of 4. If an attempt is made to
disable or enable cell0 this command will exit with a return code of 5.
Otherwise, the exit code will be 0.
NOTE: Updating the transport_url or database_connection fields on
a running system will NOT result in all nodes immediately using the
new values. Use caution when changing these values.
NOTE (tssurya): The scheduler will not notice that a cell has been
enabled/disabled until it is restarted or sent the SIGHUP signal.
"""
ctxt = context.get_admin_context()
try:
cell_mapping = objects.CellMapping.get_by_uuid(ctxt, cell_uuid)
except exception.CellMappingNotFound:
print(_('Cell with uuid %s was not found.') % cell_uuid)
return 1
if name:
cell_mapping.name = name
# Having empty transport_url and db_connection means leaving the
# existing values
transport_url = self._validate_transport_url(
transport_url, warn_about_none=False)
db_connection = self._validate_database_connection(
db_connection, warn_about_none=False)
if (self._non_unique_transport_url_database_connection_checker(ctxt,
cell_mapping, transport_url, db_connection)):
# We use the return code 3 before 2 to avoid changing the
# semantic meanings of return codes.
return 3
if transport_url:
cell_mapping.transport_url = transport_url
if db_connection:
cell_mapping.database_connection = db_connection
if disable and enable:
print(_('Cell cannot be disabled and enabled at the same time.'))
return 4
if disable or enable:
if cell_mapping.is_cell0():
print(_('Cell0 cannot be disabled.'))
return 5
elif disable and not cell_mapping.disabled:
cell_mapping.disabled = True
elif enable and cell_mapping.disabled:
cell_mapping.disabled = False
elif disable and cell_mapping.disabled:
print(_('Cell %s is already disabled') % cell_uuid)
elif enable and not cell_mapping.disabled:
print(_('Cell %s is already enabled') % cell_uuid)
try:
cell_mapping.save()
except Exception as e:
print(_('Unable to update CellMapping: %s') % e)
return 2
return 0
@args('--cell_uuid', metavar='<cell_uuid>', dest='cell_uuid',
help=_('The uuid of the cell.'))
def list_hosts(self, cell_uuid=None):
"""Lists the hosts in one or all v2 cells."""
ctxt = context.get_admin_context()
if cell_uuid:
# Find the CellMapping given the uuid.
try:
cell_mapping = objects.CellMapping.get_by_uuid(ctxt, cell_uuid)
except exception.CellMappingNotFound:
print(_('Cell with uuid %s was not found.') % cell_uuid)
return 1
host_mappings = objects.HostMappingList.get_by_cell_id(
ctxt, cell_mapping.id)
else:
host_mappings = objects.HostMappingList.get_all(ctxt)
field_names = [_('Cell Name'), _('Cell UUID'), _('Hostname')]
t = prettytable.PrettyTable(field_names)
for host in sorted(host_mappings, key=lambda _host: _host.host):
fields = [host.cell_mapping.name, host.cell_mapping.uuid,
host.host]
t.add_row(fields)
print(t)
return 0
@args('--cell_uuid', metavar='<cell_uuid>', dest='cell_uuid',
required=True, help=_('The uuid of the cell.'))
@args('--host', metavar='<host>', dest='host',
required=True, help=_('The host to delete.'))
def delete_host(self, cell_uuid, host):
"""Delete a host in a cell (host mappings) by the given host name
This command will return a non-zero exit code in the following cases.
* The cell is not found by uuid.
* The host is not found by host name.
* The host is not in the cell.
* The host has instances.
Returns 0 if the host is deleted successfully.
NOTE: The scheduler caches host-to-cell mapping information so when
deleting a host the scheduler may need to be restarted or sent the
SIGHUP signal.
"""
ctxt = context.get_admin_context()
# Find the CellMapping given the uuid.
try:
cell_mapping = objects.CellMapping.get_by_uuid(ctxt, cell_uuid)
except exception.CellMappingNotFound:
print(_('Cell with uuid %s was not found.') % cell_uuid)
return 1
try:
host_mapping = objects.HostMapping.get_by_host(ctxt, host)
except exception.HostMappingNotFound:
print(_('The host %s was not found.') % host)
return 2
if host_mapping.cell_mapping.uuid != cell_mapping.uuid:
print(_('The host %(host)s was not found '
'in the cell %(cell_uuid)s.') % {'host': host,
'cell_uuid': cell_uuid})
return 3
with context.target_cell(ctxt, cell_mapping) as cctxt:
instances = objects.InstanceList.get_by_host(cctxt, host)
try:
nodes = objects.ComputeNodeList.get_all_by_host(cctxt, host)
except exception.ComputeHostNotFound:
nodes = []
if instances:
print(_('There are instances on the host %s.') % host)
return 4
for node in nodes:
node.mapped = 0
node.save()
host_mapping.destroy()
return 0
class PlacementCommands(object):
"""Commands for managing placement resources."""
@staticmethod
def _get_compute_node_uuid(ctxt, instance, node_cache):
"""Find the ComputeNode.uuid for the given Instance
:param ctxt: cell-targeted nova.context.RequestContext
:param instance: the instance to lookup a compute node
:param node_cache: dict of Instance.node keys to ComputeNode.uuid
values; this cache is updated if a new node is processed.
:returns: ComputeNode.uuid for the given instance
:raises: nova.exception.ComputeHostNotFound
"""
if instance.node in node_cache:
return node_cache[instance.node]
compute_node = objects.ComputeNode.get_by_host_and_nodename(
ctxt, instance.host, instance.node)
node_uuid = compute_node.uuid
node_cache[instance.node] = node_uuid
return node_uuid
@staticmethod
def _get_ports(ctxt, instance, neutron):
"""Return the ports that are bound to the instance
:param ctxt: nova.context.RequestContext
:param instance: the instance to return the ports for
:param neutron: nova.network.neutron.ClientWrapper to
communicate with Neutron
:return: a list of neutron port dict objects
:raise UnableToQueryPorts: If the neutron list ports query fails.
"""
try:
return neutron.list_ports(
ctxt, device_id=instance.uuid,
fields=['id', constants.RESOURCE_REQUEST,
constants.BINDING_PROFILE]
)['ports']
except neutron_client_exc.NeutronClientException as e:
raise exception.UnableToQueryPorts(
instance_uuid=instance.uuid, error=str(e))
@staticmethod
def _has_request_but_no_allocation(port, neutron):
has_res_req = neutron_api.API()._has_resource_request(
context.get_admin_context(), port, neutron)
binding_profile = neutron_api.get_binding_profile(port)
allocation = binding_profile.get(constants.ALLOCATION)
return has_res_req and not allocation
@staticmethod
def _merge_allocations(alloc1, alloc2):
"""Return a new allocation dict that contains the sum of alloc1 and
alloc2.
:param alloc1: a dict in the form of
{
<rp_uuid>: {'resources': {<resource class>: amount,
<resource class>: amount},
<rp_uuid>: {'resources': {<resource class>: amount},
}
:param alloc2: a dict in the same form as alloc1
:return: the merged allocation of alloc1 and alloc2 in the same format
"""
allocations = collections.defaultdict(
lambda: {'resources': collections.defaultdict(int)})
for alloc in [alloc1, alloc2]:
for rp_uuid in alloc:
for rc, amount in alloc[rp_uuid]['resources'].items():
allocations[rp_uuid]['resources'][rc] += amount
return allocations
@staticmethod
def _get_resource_request_from_ports(
ctxt: context.RequestContext,
ports: ty.List[ty.Dict[str, ty.Any]]
) -> ty.Tuple[
ty.Dict[str, ty.List['objects.RequestGroup']],
'objects.RequestLevelParams']:
"""Collect RequestGroups and RequestLevelParams for all ports
:param ctxt: the request context
:param ports: a list of port dicts
:returns: A two tuple where the first item is a dict mapping port
uuids to a list of request groups coming from that port, the
second item is a combined RequestLevelParams object from all ports.
"""
groups = {}
request_level_params = objects.RequestLevelParams()
extended_res_req = (
neutron_api.API().has_extended_resource_request_extension(
ctxt)
)
for port in ports:
resource_request = port.get(constants.RESOURCE_REQUEST)
if extended_res_req:
groups[port['id']] = (
objects.RequestGroup.from_extended_port_request(
ctxt, resource_request
)
)
request_level_params.extend_with(
objects.RequestLevelParams.from_port_request(
resource_request
)
)
else:
# This is the legacy format, only one group per port and no
# request level param support
# TODO(gibi): remove this path once the extended resource
# request extension is mandatory in neutron
groups[port['id']] = [
objects.RequestGroup.from_port_request(
ctxt, port['id'], resource_request
)
]
return groups, request_level_params
@staticmethod
def _get_port_binding_profile_allocation(
ctxt: context.RequestContext,
neutron: neutron_api.ClientWrapper,
port: ty.Dict[str, ty.Any],
request_groups: ty.List['objects.RequestGroup'],
resource_provider_mapping: ty.Dict[str, ty.List[str]],
) -> ty.Dict[str, str]:
"""Generate the value of the allocation key of the port binding profile
based on the provider mapping returned from placement
:param ctxt: the request context
:param neutron: the neutron client
:param port: the port dict from neutron
:param request_groups: the list of RequestGroups object generated from
the port resource request
:param resource_provider_mapping: The dict of request group to resource
provider mapping returned by the Placement allocation candidate
query
:returns: a dict mapping request group ids to resource provider uuids
in the form as Neutron expects in the port binding profile.
"""
if neutron_api.API().has_extended_resource_request_extension(
ctxt, neutron
):
# The extended resource request format also means that a
# port has more than a one request groups.
# Each request group id from the port needs to be mapped to
# a single provider id from the provider mappings. Each
# group from the port is mapped to a numbered request group
# in placement so we can assume that they are mapped to
# a single provider and therefore the provider mapping list
# has a single provider id.
allocation = {
group.requester_id: resource_provider_mapping[
group.requester_id][0]
for group in request_groups
}
else:
# This is the legacy resource request format where a port
# is mapped to a single request group
# NOTE(gibi): In the resource provider mapping there can be
# more than one RP fulfilling a request group. But resource
# requests of a Neutron port is always mapped to a
# numbered request group that is always fulfilled by one
# resource provider. So we only pass that single RP UUID
# here.
allocation = resource_provider_mapping[
port['id']][0]
return allocation
def _get_port_allocations_to_heal(
self, ctxt, instance, node_cache, placement, neutron, output):
"""Return the needed extra allocation for the ports of the instance.
:param ctxt: nova.context.RequestContext
:param instance: instance to get the port allocations for
:param node_cache: dict of Instance.node keys to ComputeNode.uuid
values; this cache is updated if a new node is processed.
:param placement: nova.scheduler.client.report.SchedulerReportClient
to communicate with the Placement service API.
:param neutron: nova.network.neutron.ClientWrapper to
communicate with Neutron
:param output: function that takes a single message for verbose output
:raise UnableToQueryPorts: If the neutron list ports query fails.
:raise nova.exception.ComputeHostNotFound: if compute node of the
instance not found in the db.
:raise PlacementAPIConnectFailure: if placement API cannot be reached
:raise AllocationUpdateFailed: if there is either no allocation
candidate returned from placement for the missing port allocations
or there are more than one candidates making the healing
ambiguous.
:return: A two tuple where the first item is a dict of resources keyed
by RP uuid to be included in the instance allocation dict. The
second item is a list of port dicts to be updated in Neutron.
"""
# We need to heal port allocations for ports that have resource_request
# but do not have an RP uuid in the binding:profile.allocation field.
# We cannot use the instance info_cache to check the binding profile
# as this code needs to be able to handle ports that were attached
# before nova in stein started updating the allocation key in the
# binding:profile.
# In theory a port can be assigned to an instance without it being
# bound to any host (e.g. in case of shelve offload) but
# _heal_allocations_for_instance() already filters out instances that
# are not on any host.
ports_to_heal = [
port for port in self._get_ports(ctxt, instance, neutron)
if self._has_request_but_no_allocation(port, neutron)]
if not ports_to_heal:
# nothing to do, return early
return {}, []
node_uuid = self._get_compute_node_uuid(
ctxt, instance, node_cache)
# NOTE(gibi): We need to handle both legacy and extended resource
# request. So we need to handle ports with multiple request groups
# allocating from multiple providers.
# The logic what we follow here is pretty similar to the logic
# implemented in ComputeManager._allocate_port_resource_for_instance
# for the interface attach case. We just apply it to more then one
# ports here.
request_groups_per_port, req_lvl_params = (
self._get_resource_request_from_ports(ctxt, ports_to_heal)
)
# flatten the list of list of groups
request_groups = [
group
for groups in request_groups_per_port.values()
for group in groups
]
# we can have multiple request groups, it would be enough to restrict
# only one of them to the compute tree but for symmetry we restrict
# all of them
for request_group in request_groups:
request_group.in_tree = node_uuid
# If there are multiple groups then the group_policy is mandatory in
# the allocation candidate query. We can assume that if this instance
# booted successfully then we have the policy in the flavor. If there
# is only one group and therefore no policy then the value of the
# policy in the allocation candidate query is ignored, so we simply
# default it here.
group_policy = instance.flavor.extra_specs.get("group_policy", "none")
rr = scheduler_utils.ResourceRequest.from_request_groups(
request_groups, req_lvl_params, group_policy)
res = placement.get_allocation_candidates(ctxt, rr)
# NOTE(gibi): the get_allocation_candidates method has the
# @safe_connect decorator applied. Such decorator will return None
# if the connection to Placement is failed. So we raise an exception
# here. The case when Placement successfully return a response, even
# if it is a negative or empty response, the method will return a three
# tuple. That case is handled couple of lines below.
if not res:
raise exception.PlacementAPIConnectFailure()
alloc_reqs, __, __ = res
if not alloc_reqs:
port_ids = [port['id'] for port in ports_to_heal]
raise exception.AllocationUpdateFailed(
consumer_uuid=instance.uuid,
error=f'Placement returned no allocation candidate to fulfill '
f'the resource request of the port(s) {port_ids}'
)
if len(alloc_reqs) > 1:
# If there is more than one candidates then it is an ambiguous
# situation that we cannot handle here because selecting the right
# one might need extra information from the compute node. For
# example which PCI PF the VF is allocated from and which RP
# represents that PCI PF in placement.
# TODO(gibi): One way to get that missing information to resolve
# ambiguity would be to load up the InstancePciRequest objects and
# try to use the parent_if_name in their spec to find the proper
# candidate that allocates for the same port from the PF RP that
# has the same name.
port_ids = [port['id'] for port in ports_to_heal]
raise exception.AllocationUpdateFailed(
consumer_uuid=instance.uuid,
error=f'Placement returned more than one possible allocation '
f'candidates to fulfill the resource request of the '
f'port(s) {port_ids}. This script does not have enough '
f'information to select the proper candidate to heal the'
f'missing allocations. A possible way to heal the'
f'allocation of this instance is to migrate it to '
f'another compute as the migration process re-creates '
f'the full allocation on the target host.'
)
# so we have one candidate, lets use that to get the needed allocations
# and the provider mapping for the ports' binding profile
alloc_req = alloc_reqs[0]
allocations = alloc_req["allocations"]
provider_mappings = alloc_req["mappings"]
for port in ports_to_heal:
# We also need to record the RPs we are allocated from in the
# port. This will be sent back to Neutron before the allocation
# is updated in placement
profile_allocation = self._get_port_binding_profile_allocation(
ctxt, neutron, port, request_groups_per_port[port['id']],
provider_mappings
)
binding_profile = neutron_api.get_binding_profile(port)
binding_profile[constants.ALLOCATION] = profile_allocation
port[constants.BINDING_PROFILE] = binding_profile
output(_(
"Found a request group : resource provider mapping "
"%(mapping)s for the port %(port_uuid)s with resource request "
"%(request)s attached to the instance %(instance_uuid)s") %
{"mapping": profile_allocation, "port_uuid": port['id'],
"request": port.get(constants.RESOURCE_REQUEST),
"instance_uuid": instance.uuid}
)
return allocations, ports_to_heal
def _update_ports(self, neutron, ports_to_update, output):
succeeded = []
try:
for port in ports_to_update:
profile = neutron_api.get_binding_profile(port)
body = {
'port': {
constants.BINDING_PROFILE: profile
}
}
output(
_('Updating port %(port_uuid)s with attributes '
'%(attributes)s') %
{'port_uuid': port['id'], 'attributes': body['port']})
neutron.update_port(port['id'], body=body)
succeeded.append(port)
except neutron_client_exc.NeutronClientException as e:
output(
_('Updating port %(port_uuid)s failed: %(error)s') %
{'port_uuid': port['id'], 'error': str(e)})
# one of the port updates failed. We need to roll back the updates
# that succeeded before
self._rollback_port_updates(neutron, succeeded, output)
# we failed to heal so we need to stop but we successfully rolled
# back the partial updates so the admin can retry the healing.
raise exception.UnableToUpdatePorts(error=str(e))
@staticmethod
def _rollback_port_updates(neutron, ports_to_rollback, output):
# _update_ports() added the allocation key to these ports, so we need
# to remove them during the rollback.
manual_rollback_needed = []
last_exc = None
for port in ports_to_rollback:
profile = neutron_api.get_binding_profile(port)
profile.pop(constants.ALLOCATION)
body = {
'port': {
constants.BINDING_PROFILE: profile
}
}
try:
output(_('Rolling back port update for %(port_uuid)s') %
{'port_uuid': port['id']})
neutron.update_port(port['id'], body=body)
except neutron_client_exc.NeutronClientException as e:
output(
_('Rolling back update for port %(port_uuid)s failed: '
'%(error)s') % {'port_uuid': port['id'],
'error': str(e)})
# TODO(gibi): We could implement a retry mechanism with
# back off.
manual_rollback_needed.append(port['id'])
last_exc = e
if manual_rollback_needed:
# At least one of the port operation failed so we failed to roll
# back. There are partial updates in neutron. Human intervention
# needed.
raise exception.UnableToRollbackPortUpdates(
error=str(last_exc),
port_uuids=manual_rollback_needed)
def _heal_missing_alloc(self, ctxt, instance, node_cache):
node_uuid = self._get_compute_node_uuid(
ctxt, instance, node_cache)
# Now get the resource allocations for the instance based
# on its embedded flavor.
resources = scheduler_utils.resources_from_flavor(
instance, instance.flavor)
payload = {
'allocations': {
node_uuid: {'resources': resources},
},
'project_id': instance.project_id,
'user_id': instance.user_id,
'consumer_generation': None
}
return payload
def _heal_missing_project_and_user_id(self, allocations, instance):
allocations['project_id'] = instance.project_id
allocations['user_id'] = instance.user_id
return allocations
@staticmethod
def ensure_instance_has_no_vgpu_request(instance):
if instance.flavor.extra_specs.get("resources:VGPU"):
raise exception.HealvGPUAllocationNotSupported(
instance_uuid=instance.uuid)
@staticmethod
def ensure_instance_has_no_cyborg_device_profile_request(instance):
if instance.flavor.extra_specs.get("accel:device_profile"):
raise exception.HealDeviceProfileAllocationNotSupported(
instance_uuid=instance.uuid)
def _heal_allocations_for_instance(self, ctxt, instance, node_cache,
output, placement, dry_run,
heal_port_allocations, neutron,
force):
"""Checks the given instance to see if it needs allocation healing
:param ctxt: cell-targeted nova.context.RequestContext
:param instance: the instance to check for allocation healing
:param node_cache: dict of Instance.node keys to ComputeNode.uuid
values; this cache is updated if a new node is processed.
:param output: function that takes a single message for verbose output
:param placement: nova.scheduler.client.report.SchedulerReportClient
to communicate with the Placement service API.
:param dry_run: Process instances and print output but do not commit
any changes.
:param heal_port_allocations: True if healing port allocation is
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
given instance cannot be found
:raises: AllocationCreateFailed if unable to create allocations for
a given instance against a given compute node resource provider
:raises: AllocationUpdateFailed if unable to update allocations for
a given instance with consumer project/user information
:raise UnableToQueryPorts: If the neutron list ports query fails.
:raise PlacementAPIConnectFailure: if placement API cannot be reached
:raise UnableToUpdatePorts: if a port update failed in neutron but any
partial update was rolled back successfully.
:raise UnableToRollbackPortUpdates: if a port update failed in neutron
and the rollback of the partial updates also failed.
"""
if instance.task_state is not None:
output(_('Instance %(instance)s is undergoing a task '
'state transition: %(task_state)s') %
{'instance': instance.uuid,
'task_state': instance.task_state})
return
if instance.node is None:
output(_('Instance %s is not on a host.') % instance.uuid)
return
self.ensure_instance_has_no_vgpu_request(instance)
self.ensure_instance_has_no_cyborg_device_profile_request(instance)
try:
allocations = placement.get_allocs_for_consumer(
ctxt, instance.uuid)
except (ks_exc.ClientException,
exception.ConsumerAllocationRetrievalFailed) as e:
raise exception.AllocationUpdateFailed(
consumer_uuid=instance.uuid,
error=_("Allocation retrieval failed: %s") % e)
need_healing = False
# Placement response can have an empty {'allocations': {}} in it if
# there are no allocations for the instance
if not allocations.get('allocations'):
# This instance doesn't have allocations
need_healing = _CREATE
allocations = self._heal_missing_alloc(ctxt, instance, node_cache)
if (allocations.get('project_id') != instance.project_id or
allocations.get('user_id') != instance.user_id):
# We have an instance with allocations but not the correct
# project_id/user_id, so we want to update the allocations
# and re-put them. We don't use put_allocations here
# because we don't want to mess up shared or nested
# provider allocations.
need_healing = _UPDATE
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)
port_allocations, ports_to_update = to_heal
else:
port_allocations, ports_to_update = {}, []
if port_allocations:
need_healing = need_healing or _UPDATE
# Merge in any missing port allocations
allocations['allocations'] = self._merge_allocations(
allocations['allocations'], port_allocations)
if need_healing:
if dry_run:
# json dump the allocation dict as it contains nested default
# dicts that is pretty hard to read in the verbose output
alloc = jsonutils.dumps(allocations)
if need_healing == _CREATE:
output(_('[dry-run] Create allocations for instance '
'%(instance)s: %(allocations)s') %
{'instance': instance.uuid,
'allocations': alloc})
elif need_healing == _UPDATE:
output(_('[dry-run] Update allocations for instance '
'%(instance)s: %(allocations)s') %
{'instance': instance.uuid,
'allocations': alloc})
else:
# First update ports in neutron. If any of those operations
# fail, then roll back the successful part of it and fail the
# healing. We do this first because rolling back the port
# updates is more straight-forward than rolling back allocation
# changes.
self._update_ports(neutron, ports_to_update, output)
# Now that neutron update succeeded we can try to update
# placement. If it fails we need to rollback every neutron port
# update done before.
resp = placement.put_allocations(ctxt, instance.uuid,
allocations)
if resp:
if need_healing == _CREATE:
output(_('Successfully created allocations for '
'instance %(instance)s.') %
{'instance': instance.uuid})
elif need_healing == _UPDATE:
output(_('Successfully updated allocations for '
'instance %(instance)s.') %
{'instance': instance.uuid})
return True
else:
# Rollback every neutron update. If we succeed to
# roll back then it is safe to stop here and let the admin
# retry. If the rollback fails then
# _rollback_port_updates() will raise another exception
# that instructs the operator how to clean up manually
# before the healing can be retried
self._rollback_port_updates(
neutron, ports_to_update, output)
raise exception.AllocationUpdateFailed(
consumer_uuid=instance.uuid, error='')
else:
output(_('The allocation of instance %s is up-to-date. '
'Nothing to be healed.') % instance.uuid)
return
def _heal_instances_in_cell(self, ctxt, max_count, unlimited, output,
placement, dry_run, instance_uuid,
heal_port_allocations, neutron,
force):
"""Checks for instances to heal in a given cell.
:param ctxt: cell-targeted nova.context.RequestContext
:param max_count: batch size (limit per instance query)
:param unlimited: True if all instances in the cell should be
processed, else False to just process $max_count instances
:param output: function that takes a single message for verbose output
:param placement: nova.scheduler.client.report.SchedulerReportClient
to communicate with the Placement service API.
:param dry_run: Process instances and print output but do not commit
any changes.
:param instance_uuid: UUID of a specific instance to process.
:param heal_port_allocations: True if healing port allocation is
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
:raises: AllocationCreateFailed if unable to create allocations for
a given instance against a given compute node resource provider
:raises: AllocationUpdateFailed if unable to update allocations for
a given instance with consumer project/user information
:raise UnableToQueryPorts: If the neutron list ports query fails.
:raise PlacementAPIConnectFailure: if placement API cannot be reached
:raise UnableToUpdatePorts: if a port update failed in neutron but any
partial update was rolled back successfully.
:raise UnableToRollbackPortUpdates: if a port update failed in neutron
and the rollback of the partial updates also failed.
"""
# Keep a cache of instance.node to compute node resource provider UUID.
# This will save some queries for non-ironic instances to the
# compute_nodes table.
node_cache = {}
# Track the total number of instances that have allocations created
# for them in this cell. We return when num_processed equals max_count
# and unlimited=True or we exhaust the number of instances to process
# in this cell.
num_processed = 0
# Get all instances from this cell which have a host and are not
# undergoing a task state transition. Go from oldest to newest.
# NOTE(mriedem): Unfortunately we don't have a marker to use
# between runs where the user is specifying --max-count.
# TODO(mriedem): Store a marker in system_metadata so we can
# automatically pick up where we left off without the user having
# to pass it in (if unlimited is False).
filters = {'deleted': False}
if instance_uuid:
filters['uuid'] = instance_uuid
instances = objects.InstanceList.get_by_filters(
ctxt, filters=filters, sort_key='created_at', sort_dir='asc',
limit=max_count, expected_attrs=['flavor'])
while instances:
output(_('Found %s candidate instances.') % len(instances))
# For each instance in this list, we need to see if it has
# allocations in placement and if so, assume it's correct and
# continue.
for instance in instances:
if self._heal_allocations_for_instance(
ctxt, instance, node_cache, output, placement,
dry_run, heal_port_allocations, neutron, force):
num_processed += 1
# Make sure we don't go over the max count. Note that we
# don't include instances that already have allocations in the
# max_count number, only the number of instances that have
# successfully created allocations.
# If a specific instance was requested we return here as well.
if (not unlimited and num_processed == max_count) or instance_uuid:
return num_processed
# Use a marker to get the next page of instances in this cell.
# Note that InstanceList doesn't support slice notation.
marker = instances[len(instances) - 1].uuid
instances = objects.InstanceList.get_by_filters(
ctxt, filters=filters, sort_key='created_at', sort_dir='asc',
limit=max_count, marker=marker, expected_attrs=['flavor'])
return num_processed
@action_description(
_("Iterates over non-cell0 cells looking for instances which do "
"not have allocations in the Placement service, or have incomplete "
"consumer project_id/user_id values in existing allocations or "
"missing allocations for ports having resource request, and "
"which are not undergoing a task state transition. For each "
"instance found, allocations are created (or updated) against the "
"compute node resource provider for that instance based on the "
"flavor associated with the instance. This command requires that "
"the [api_database]/connection and [placement] configuration "
"options are set."))
@args('--max-count', metavar='<max_count>', dest='max_count',
help='Maximum number of instances to process. If not specified, all '
'instances in each cell will be mapped in batches of 50. '
'If you have a large number of instances, consider specifying '
'a custom value and run the command until it exits with '
'0 or 4.')
@args('--verbose', action='store_true', dest='verbose', default=False,
help='Provide verbose output during execution.')
@args('--dry-run', action='store_true', dest='dry_run', default=False,
help='Runs the command and prints output but does not commit any '
'changes. The return code should be 4.')
@args('--instance', metavar='<instance_uuid>', dest='instance_uuid',
help='UUID of a specific instance to process. If specified '
'--max-count has no effect. '
'The --cell and --instance options are mutually exclusive.')
@args('--skip-port-allocations', action='store_true',
dest='skip_port_allocations', default=False,
help='Skip the healing of the resource allocations of bound ports. '
'E.g. healing bandwidth resource allocation for ports having '
'minimum QoS policy rules attached. If your deployment does '
'not use such a feature then the performance impact of '
'querying neutron ports for each instance can be avoided with '
'this flag.')
@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, force=False):
"""Heals instance allocations in the Placement service
Return codes:
* 0: Command completed successfully and allocations were created.
* 1: --max-count was reached and there are more instances to process.
* 2: Unable to find a compute node record for a given instance.
* 3: Unable to create (or update) allocations for an instance against
its compute node resource provider.
* 4: Command completed successfully but no allocations were created.
* 5: Unable to query ports from neutron
* 6: Unable to update ports in neutron
* 7: Cannot roll back neutron port updates. Manual steps needed.
* 8: Cannot heal instance with vGPU or Cyborg resource request
* 127: Invalid input.
"""
# NOTE(mriedem): Thoughts on ways to expand this:
# - allow filtering on enabled/disabled cells
# - add a force option to force allocations for instances which have
# task_state is not None (would get complicated during a migration);
# for example, this could cleanup ironic instances that have
# allocations on VCPU/MEMORY_MB/DISK_GB but are now using a custom
# resource class
# - deal with nested resource providers?
heal_port_allocations = not skip_port_allocations
output = lambda msg: None
if verbose:
output = lambda msg: print(msg)
# If user has provided both cell and instance
# Throw an error
if instance_uuid and cell_uuid:
print(_('The --cell and --instance options '
'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
# defaults to None to mean unlimited.
if instance_uuid:
max_count = 1
unlimited = False
elif max_count is not None:
try:
max_count = int(max_count)
except ValueError:
max_count = -1
unlimited = False
if max_count < 1:
print(_('Must supply a positive integer for --max-count.'))
return 127
else:
max_count = 50
unlimited = True
output(_('Running batches of %i until complete') % max_count)
ctxt = context.get_admin_context()
# If we are going to process a specific instance, just get the cell
# it is in up front.
if instance_uuid:
try:
im = objects.InstanceMapping.get_by_instance_uuid(
ctxt, instance_uuid)
cells = objects.CellMappingList(objects=[im.cell_mapping])
except exception.InstanceMappingNotFound:
print('Unable to find cell for instance %s, is it mapped? Try '
'running "nova-manage cell_v2 verify_instance" or '
'"nova-manage cell_v2 map_instances".' %
instance_uuid)
return 127
elif cell_uuid:
try:
# validate cell_uuid
cell = objects.CellMapping.get_by_uuid(ctxt, cell_uuid)
# create CellMappingList
cells = objects.CellMappingList(objects=[cell])
except exception.CellMappingNotFound:
print(_('Cell with uuid %s was not found.') % cell_uuid)
return 127
else:
cells = objects.CellMappingList.get_all(ctxt)
if not cells:
output(_('No cells to process.'))
return 4
placement = report.report_client_singleton()
neutron = None
if heal_port_allocations:
neutron = neutron_api.get_client(ctxt, admin=True)
num_processed = 0
# TODO(mriedem): Use context.scatter_gather_skip_cell0.
for cell in cells:
# Skip cell0 since that is where instances go that do not get
# scheduled and hence would not have allocations against a host.
if cell.uuid == objects.CellMapping.CELL0_UUID:
continue
output(_('Looking for instances in cell: %s') % cell.identity)
limit_per_cell = max_count
if not unlimited:
# Adjust the limit for the next cell. For example, if the user
# only wants to process a total of 100 instances and we did
# 75 in cell1, then we only need 25 more from cell2 and so on.
limit_per_cell = max_count - num_processed
with context.target_cell(ctxt, cell) as cctxt:
try:
num_processed += self._heal_instances_in_cell(
cctxt, limit_per_cell, unlimited, output, placement,
dry_run, instance_uuid, heal_port_allocations, neutron,
force)
except exception.ComputeHostNotFound as e:
print(e.format_message())
return 2
except (
exception.AllocationCreateFailed,
exception.AllocationUpdateFailed,
exception.PlacementAPIConnectFailure
) as e:
print(e.format_message())
return 3
except exception.UnableToQueryPorts as e:
print(e.format_message())
return 5
except exception.UnableToUpdatePorts as e:
print(e.format_message())
return 6
except exception.UnableToRollbackPortUpdates as e:
print(e.format_message())
return 7
except (
exception.HealvGPUAllocationNotSupported,
exception.HealDeviceProfileAllocationNotSupported,
) as e:
print(e.format_message())
return 8
# Make sure we don't go over the max count. Note that we
# don't include instances that already have allocations in the
# max_count number, only the number of instances that have
# successfully created allocations.
# If a specific instance was provided then we'll just exit
# the loop and process it below (either return 4 or 0).
if num_processed == max_count and not instance_uuid:
output(_('Max count reached. Processed %s instances.')
% num_processed)
return 1
output(_('Processed %s instances.') % num_processed)
if not num_processed:
return 4
return 0
@staticmethod
def _get_rp_uuid_for_host(ctxt, host):
"""Finds the resource provider (compute node) UUID for the given host.
:param ctxt: cell-targeted nova RequestContext
:param host: name of the compute host
:returns: The UUID of the resource provider (compute node) for the host
:raises: nova.exception.HostMappingNotFound if no host_mappings record
is found for the host; indicates
"nova-manage cell_v2 discover_hosts" needs to be run on the cell.
:raises: nova.exception.ComputeHostNotFound if no compute_nodes record
is found in the cell database for the host; indicates the
nova-compute service on that host might need to be restarted.
:raises: nova.exception.TooManyComputesForHost if there are more than
one compute_nodes records in the cell database for the host which
is only possible (under normal circumstances) for ironic hosts but
ironic hosts are not currently supported with host aggregates so
if more than one compute node is found for the host, it is
considered an error which the operator will need to resolve
manually.
"""
# Get the host mapping to determine which cell it's in.
hm = objects.HostMapping.get_by_host(ctxt, host)
# Now get the compute node record for the host from the cell.
with context.target_cell(ctxt, hm.cell_mapping) as cctxt:
# There should really only be one, since only ironic
# hosts can have multiple nodes, and you can't have
# ironic hosts in aggregates for that reason. If we
# find more than one, it's an error.
nodes = objects.ComputeNodeList.get_all_by_host(
cctxt, host)
if len(nodes) > 1:
# This shouldn't happen, so we need to bail since we
# won't know which node to use.
raise exception.TooManyComputesForHost(
num_computes=len(nodes), host=host)
return nodes[0].uuid
@action_description(
_("Mirrors compute host aggregates to resource provider aggregates "
"in the Placement service. Requires the [api_database] and "
"[placement] sections of the nova configuration file to be "
"populated."))
@args('--verbose', action='store_true', dest='verbose', default=False,
help='Provide verbose output during execution.')
# TODO(mriedem): Add an option for the 'remove aggregate' behavior.
# We know that we want to mirror hosts aggregate membership to
# placement, but regarding removal, what if the operator or some external
# tool added the resource provider to an aggregate but there is no matching
# host aggregate, e.g. ironic nodes or shared storage provider
# relationships?
# TODO(mriedem): Probably want an option to pass a specific host instead of
# doing all of them.
def sync_aggregates(self, verbose=False):
"""Synchronizes nova host aggregates with resource provider aggregates
Adds nodes to missing provider aggregates in Placement.
NOTE: Depending on the size of your deployment and the number of
compute hosts in aggregates, this command could cause a non-negligible
amount of traffic to the placement service and therefore is
recommended to be run during maintenance windows.
Return codes:
* 0: Successful run
* 1: A host was found with more than one matching compute node record
* 2: An unexpected error occurred while working with the placement API
* 3: Failed updating provider aggregates in placement
* 4: Host mappings not found for one or more host aggregate members
* 5: Compute node records not found for one or more hosts
* 6: Resource provider not found by uuid for a given host
"""
# Start by getting all host aggregates.
ctxt = context.get_admin_context()
aggregate_api = api.AggregateAPI()
placement = aggregate_api.placement_client
aggregates = aggregate_api.get_aggregate_list(ctxt)
# Now we're going to loop over the existing compute hosts in aggregates
# and check to see if their corresponding resource provider, found via
# the host's compute node uuid, are in the same aggregate. If not, we
# add the resource provider to the aggregate in Placement.
output = lambda msg: None
if verbose:
output = lambda msg: print(msg)
output(_('Filling in missing placement aggregates'))
# Since hosts can be in more than one aggregate, keep track of the host
# to its corresponding resource provider uuid to avoid redundant
# lookups.
host_to_rp_uuid = {}
unmapped_hosts = set() # keep track of any missing host mappings
computes_not_found = set() # keep track of missing nodes
providers_not_found = {} # map of hostname to missing provider uuid
for aggregate in aggregates:
output(_('Processing aggregate: %s') % aggregate.name)
for host in aggregate.hosts:
output(_('Processing host: %s') % host)
rp_uuid = host_to_rp_uuid.get(host)
if not rp_uuid:
try:
rp_uuid = self._get_rp_uuid_for_host(ctxt, host)
host_to_rp_uuid[host] = rp_uuid
except exception.HostMappingNotFound:
# Don't fail on this now, we can dump it at the end.
unmapped_hosts.add(host)
continue
except exception.ComputeHostNotFound:
# Don't fail on this now, we can dump it at the end.
computes_not_found.add(host)
continue
except exception.TooManyComputesForHost as e:
# TODO(mriedem): Should we treat this like the other
# errors and not fail immediately but dump at the end?
print(e.format_message())
return 1
# We've got our compute node record, so now we can ensure that
# the matching resource provider, found via compute node uuid,
# is in the same aggregate in placement, found via aggregate
# uuid.
try:
placement.aggregate_add_host(ctxt, aggregate.uuid,
rp_uuid=rp_uuid)
output(_('Successfully added host (%(host)s) and '
'provider (%(provider)s) to aggregate '
'(%(aggregate)s).') %
{'host': host, 'provider': rp_uuid,
'aggregate': aggregate.uuid})
except exception.ResourceProviderNotFound:
# The resource provider wasn't found. Store this for later.
providers_not_found[host] = rp_uuid
except exception.ResourceProviderAggregateRetrievalFailed as e:
print(e.message)
return 2
except exception.NovaException as e:
# The exception message is too generic in this case
print(_('Failed updating provider aggregates for '
'host (%(host)s), provider (%(provider)s) '
'and aggregate (%(aggregate)s). Error: '
'%(error)s') %
{'host': host, 'provider': rp_uuid,
'aggregate': aggregate.uuid,
'error': e.message})
return 3
# Now do our error handling. Note that there is no real priority on
# the error code we return. We want to dump all of the issues we hit
# so the operator can fix them before re-running the command, but
# whether we return 4 or 5 or 6 doesn't matter.
return_code = 0
if unmapped_hosts:
print(_('The following hosts were found in nova host aggregates '
'but no host mappings were found in the nova API DB. Run '
'"nova-manage cell_v2 discover_hosts" and then retry. '
'Missing: %s') % ','.join(unmapped_hosts))
return_code = 4
if computes_not_found:
print(_('Unable to find matching compute_nodes record entries in '
'the cell database for the following hosts; does the '
'nova-compute service on each host need to be restarted? '
'Missing: %s') % ','.join(computes_not_found))
return_code = 5
if providers_not_found:
print(_('Unable to find matching resource provider record in '
'placement with uuid for the following hosts: %s. Try '
'restarting the nova-compute service on each host and '
'then retry.') %
','.join('(%s=%s)' % (host, providers_not_found[host])
for host in sorted(providers_not_found.keys())))
return_code = 6
return return_code
def _get_instances_and_current_migrations(self, ctxt, cn_uuid):
if self.cn_uuid_mapping.get(cn_uuid):
cell_uuid, cn_host, cn_node = self.cn_uuid_mapping[cn_uuid]
else:
# We need to find the compute node record from all cells.
results = context.scatter_gather_skip_cell0(
ctxt, objects.ComputeNode.get_by_uuid, cn_uuid)
for result_cell_uuid, result in results.items():
if not context.is_cell_failure_sentinel(result):
cn = result
cell_uuid = result_cell_uuid
break
else:
return False
cn_host, cn_node = (cn.host, cn.hypervisor_hostname)
self.cn_uuid_mapping[cn_uuid] = (cell_uuid, cn_host, cn_node)
cell_mapping = objects.CellMapping.get_by_uuid(ctxt, cell_uuid)
# Get all the active instances from this compute node
if self.instances_mapping.get(cn_uuid):
inst_uuids = self.instances_mapping[cn_uuid]
else:
# Get the instance list record from the cell.
with context.target_cell(ctxt, cell_mapping) as cctxt:
instances = objects.InstanceList.get_by_host_and_node(
cctxt, cn_host, cn_node, expected_attrs=[])
inst_uuids = [instance.uuid for instance in instances]
self.instances_mapping[cn_uuid] = inst_uuids
# Get all *active* migrations for this compute node
# NOTE(sbauza): Since migrations are transient, it's better to not
# cache the results as they could be stale
with context.target_cell(ctxt, cell_mapping) as cctxt:
migs = objects.MigrationList.get_in_progress_by_host_and_node(
cctxt, cn_host, cn_node)
mig_uuids = [migration.uuid for migration in migs]
return (inst_uuids, mig_uuids)
def _delete_allocations_from_consumer(self, ctxt, placement, provider,
consumer_uuid, consumer_type):
"""Deletes allocations from a resource provider with consumer UUID.
:param ctxt: nova.context.RequestContext
:param placement: nova.scheduler.client.report.SchedulerReportClient
to communicate with the Placement service API.
:param provider: Resource Provider to look at.
:param consumer_uuid: the consumer UUID having allocations.
:param consumer_type: the type of consumer,
either 'instance' or 'migration'
:returns: bool whether the allocations were deleted.
"""
# We need to be careful and only remove the allocations
# against this specific RP or we would delete the
# whole instance usage and then it would require some
# healing.
# TODO(sbauza): Remove this extra check once placement
# supports querying allocation delete on both
# consumer and resource provider parameters.
allocations = placement.get_allocs_for_consumer(
ctxt, consumer_uuid)
if len(allocations['allocations']) > 1:
# This consumer has resources spread among multiple RPs (think
# nested or shared for example)
# We then need to just update the usage to remove
# the orphaned resources on the specific RP
del allocations['allocations'][provider['uuid']]
try:
placement.put_allocations(
ctxt, consumer_uuid, allocations)
except exception.AllocationUpdateFailed:
return False
else:
try:
placement.delete_allocation_for_instance(
ctxt, consumer_uuid, consumer_type, force=True)
except exception.AllocationDeleteFailed:
return False
return True
def _check_orphaned_allocations_for_provider(self, ctxt, placement,
output, provider,
delete):
"""Finds orphaned allocations for a specific resource provider.
:param ctxt: nova.context.RequestContext
:param placement: nova.scheduler.client.report.SchedulerReportClient
to communicate with the Placement service API.
:param output: function that takes a single message for verbose output
:param provider: Resource Provider to look at.
:param delete: deletes the found orphaned allocations.
:return: a tuple (<number of orphaned allocs>, <number of faults>)
"""
num_processed = 0
faults = 0
# TODO(sbauza): Are we sure we have all Nova RCs ?
# FIXME(sbauza): Possibly use consumer types once Placement API
# supports them.
# NOTE(sbauza): We check allocations having *any* below RC, not having
# *all* of them.
NOVA_RCS = [orc.VCPU, orc.MEMORY_MB, orc.DISK_GB, orc.VGPU,
orc.NET_BW_EGR_KILOBIT_PER_SEC,
orc.NET_BW_IGR_KILOBIT_PER_SEC,
orc.PCPU, orc.MEM_ENCRYPTION_CONTEXT]
# Since the RP can be a child RP, we need to get the root RP as it's
# the compute node UUID
# NOTE(sbauza): In case Placement doesn't support 1.14 microversion,
# that means we don't have nested RPs.
# Since we ask for microversion 1.14, all RPs have a root RP UUID. </