File: utils.py

package info (click to toggle)
nova 2:14.0.0-4+deb9u1
  • links: PTS, VCS
  • area: main
  • in suites: stretch
  • size: 33,804 kB
  • sloc: python: 315,557; sh: 1,317; xml: 1,184; pascal: 1,168; makefile: 126; sql: 43
file content (397 lines) | stat: -rw-r--r-- 14,975 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
# All Rights Reserved.
#
#    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.

"""Utility methods for scheduling."""

import collections
import functools
import sys

from oslo_log import log as logging
import oslo_messaging as messaging
from oslo_serialization import jsonutils

from nova.compute import flavors
from nova.compute import utils as compute_utils
import nova.conf
from nova import exception
from nova.i18n import _, _LE, _LW
from nova import objects
from nova.objects import base as obj_base
from nova.objects import instance as obj_instance
from nova import rpc


LOG = logging.getLogger(__name__)

CONF = nova.conf.CONF

GroupDetails = collections.namedtuple('GroupDetails', ['hosts', 'policies',
                                                       'members'])


def build_request_spec(ctxt, image, instances, instance_type=None):
    """Build a request_spec for the scheduler.

    The request_spec assumes that all instances to be scheduled are the same
    type.
    """
    instance = instances[0]
    if instance_type is None:
        if isinstance(instance, obj_instance.Instance):
            instance_type = instance.get_flavor()
        else:
            instance_type = flavors.extract_flavor(instance)

    if isinstance(instance, obj_instance.Instance):
        instance = obj_base.obj_to_primitive(instance)
        # obj_to_primitive doesn't copy this enough, so be sure
        # to detach our metadata blob because we modify it below.
        instance['system_metadata'] = dict(instance.get('system_metadata', {}))

    if isinstance(instance_type, objects.Flavor):
        instance_type = obj_base.obj_to_primitive(instance_type)
        # NOTE(danms): Replicate this old behavior because the
        # scheduler RPC interface technically expects it to be
        # there. Remove this when we bump the scheduler RPC API to
        # v5.0
        try:
            flavors.save_flavor_info(instance.get('system_metadata', {}),
                                     instance_type)
        except KeyError:
            # If the flavor isn't complete (which is legit with a
            # flavor object, just don't put it in the request spec
            pass

    request_spec = {
            'image': image or {},
            'instance_properties': instance,
            'instance_type': instance_type,
            'num_instances': len(instances)}
    return jsonutils.to_primitive(request_spec)


def set_vm_state_and_notify(context, instance_uuid, service, method, updates,
                            ex, request_spec):
    """changes VM state and notifies."""
    LOG.warning(_LW("Failed to %(service)s_%(method)s: %(ex)s"),
                {'service': service, 'method': method, 'ex': ex})

    vm_state = updates['vm_state']
    properties = request_spec.get('instance_properties', {})
    # NOTE(vish): We shouldn't get here unless we have a catastrophic
    #             failure, so just set the instance to its internal state
    notifier = rpc.get_notifier(service)
    state = vm_state.upper()
    LOG.warning(_LW('Setting instance to %s state.'), state,
                instance_uuid=instance_uuid)

    instance = objects.Instance(context=context, uuid=instance_uuid,
                                **updates)
    instance.obj_reset_changes(['uuid'])
    instance.save()
    compute_utils.add_instance_fault_from_exc(context,
            instance, ex, sys.exc_info())

    payload = dict(request_spec=request_spec,
                    instance_properties=properties,
                    instance_id=instance_uuid,
                    state=vm_state,
                    method=method,
                    reason=ex)

    event_type = '%s.%s' % (service, method)
    notifier.error(context, event_type, payload)


def build_filter_properties(scheduler_hints, forced_host,
        forced_node, instance_type):
    """Build the filter_properties dict from data in the boot request."""
    filter_properties = dict(scheduler_hints=scheduler_hints)
    filter_properties['instance_type'] = instance_type
    # TODO(alaski): It doesn't seem necessary that these are conditionally
    # added.  Let's just add empty lists if not forced_host/node.
    if forced_host:
        filter_properties['force_hosts'] = [forced_host]
    if forced_node:
        filter_properties['force_nodes'] = [forced_node]
    return filter_properties


def populate_filter_properties(filter_properties, host_state):
    """Add additional information to the filter properties after a node has
    been selected by the scheduling process.
    """
    if isinstance(host_state, dict):
        host = host_state['host']
        nodename = host_state['nodename']
        limits = host_state['limits']
    else:
        host = host_state.host
        nodename = host_state.nodename
        limits = host_state.limits

    # Adds a retry entry for the selected compute host and node:
    _add_retry_host(filter_properties, host, nodename)

    # Adds oversubscription policy
    if not filter_properties.get('force_hosts'):
        filter_properties['limits'] = limits


def populate_retry(filter_properties, instance_uuid):
    max_attempts = CONF.scheduler_max_attempts
    force_hosts = filter_properties.get('force_hosts', [])
    force_nodes = filter_properties.get('force_nodes', [])

    # In the case of multiple force hosts/nodes, scheduler should not
    # disable retry filter but traverse all force hosts/nodes one by
    # one till scheduler gets a valid target host.
    if (max_attempts == 1 or len(force_hosts) == 1
                           or len(force_nodes) == 1):
        # re-scheduling is disabled.
        return

    # retry is enabled, update attempt count:
    retry = filter_properties.setdefault(
        'retry', {
            'num_attempts': 0,
            'hosts': []  # list of compute hosts tried
    })
    retry['num_attempts'] += 1

    _log_compute_error(instance_uuid, retry)
    exc_reason = retry.pop('exc_reason', None)

    if retry['num_attempts'] > max_attempts:
        msg = (_('Exceeded max scheduling attempts %(max_attempts)d '
                 'for instance %(instance_uuid)s. '
                 'Last exception: %(exc_reason)s')
               % {'max_attempts': max_attempts,
                  'instance_uuid': instance_uuid,
                  'exc_reason': exc_reason})
        raise exception.MaxRetriesExceeded(reason=msg)


def _log_compute_error(instance_uuid, retry):
    """If the request contained an exception from a previous compute
    build/resize operation, log it to aid debugging
    """
    exc = retry.get('exc')  # string-ified exception from compute
    if not exc:
        return  # no exception info from a previous attempt, skip

    hosts = retry.get('hosts', None)
    if not hosts:
        return  # no previously attempted hosts, skip

    last_host, last_node = hosts[-1]
    LOG.error(_LE('Error from last host: %(last_host)s (node %(last_node)s):'
                  ' %(exc)s'),
              {'last_host': last_host,
               'last_node': last_node,
               'exc': exc},
              instance_uuid=instance_uuid)


def _add_retry_host(filter_properties, host, node):
    """Add a retry entry for the selected compute node. In the event that
    the request gets re-scheduled, this entry will signal that the given
    node has already been tried.
    """
    retry = filter_properties.get('retry', None)
    if not retry:
        return
    hosts = retry['hosts']
    hosts.append([host, node])


def parse_options(opts, sep='=', converter=str, name=""):
    """Parse a list of options, each in the format of <key><sep><value>. Also
    use the converter to convert the value into desired type.

    :params opts: list of options, e.g. from oslo_config.cfg.ListOpt
    :params sep: the separator
    :params converter: callable object to convert the value, should raise
                       ValueError for conversion failure
    :params name: name of the option

    :returns: a lists of tuple of values (key, converted_value)
    """
    good = []
    bad = []
    for opt in opts:
        try:
            key, seen_sep, value = opt.partition(sep)
            value = converter(value)
        except ValueError:
            key = None
            value = None
        if key and seen_sep and value is not None:
            good.append((key, value))
        else:
            bad.append(opt)
    if bad:
        LOG.warning(_LW("Ignoring the invalid elements of the option "
                        "%(name)s: %(options)s"),
                    {'name': name,
                     'options': ", ".join(bad)})
    return good


def validate_filter(filter):
    """Validates that the filter is configured in the default filters."""
    return filter in CONF.scheduler_default_filters


def validate_weigher(weigher):
    """Validates that the weigher is configured in the default weighers."""
    if 'nova.scheduler.weights.all_weighers' in CONF.scheduler_weight_classes:
        return True
    return weigher in CONF.scheduler_weight_classes


_SUPPORTS_AFFINITY = None
_SUPPORTS_ANTI_AFFINITY = None
_SUPPORTS_SOFT_AFFINITY = None
_SUPPORTS_SOFT_ANTI_AFFINITY = None


def _get_group_details(context, instance_uuid, user_group_hosts=None):
    """Provide group_hosts and group_policies sets related to instances if
    those instances are belonging to a group and if corresponding filters are
    enabled.

    :param instance_uuid: UUID of the instance to check
    :param user_group_hosts: Hosts from the group or empty set

    :returns: None or namedtuple GroupDetails
    """
    global _SUPPORTS_AFFINITY
    if _SUPPORTS_AFFINITY is None:
        _SUPPORTS_AFFINITY = validate_filter(
            'ServerGroupAffinityFilter')
    global _SUPPORTS_ANTI_AFFINITY
    if _SUPPORTS_ANTI_AFFINITY is None:
        _SUPPORTS_ANTI_AFFINITY = validate_filter(
            'ServerGroupAntiAffinityFilter')
    global _SUPPORTS_SOFT_AFFINITY
    if _SUPPORTS_SOFT_AFFINITY is None:
        _SUPPORTS_SOFT_AFFINITY = validate_weigher(
            'nova.scheduler.weights.affinity.ServerGroupSoftAffinityWeigher')
    global _SUPPORTS_SOFT_ANTI_AFFINITY
    if _SUPPORTS_SOFT_ANTI_AFFINITY is None:
        _SUPPORTS_SOFT_ANTI_AFFINITY = validate_weigher(
            'nova.scheduler.weights.affinity.'
            'ServerGroupSoftAntiAffinityWeigher')

    if not instance_uuid:
        return

    try:
        group = objects.InstanceGroup.get_by_instance_uuid(context,
                                                           instance_uuid)
    except exception.InstanceGroupNotFound:
        return

    policies = set(('anti-affinity', 'affinity', 'soft-affinity',
                    'soft-anti-affinity'))
    if any((policy in policies) for policy in group.policies):
        if not _SUPPORTS_AFFINITY and 'affinity' in group.policies:
            msg = _("ServerGroupAffinityFilter not configured")
            LOG.error(msg)
            raise exception.UnsupportedPolicyException(reason=msg)
        if not _SUPPORTS_ANTI_AFFINITY and 'anti-affinity' in group.policies:
            msg = _("ServerGroupAntiAffinityFilter not configured")
            LOG.error(msg)
            raise exception.UnsupportedPolicyException(reason=msg)
        if (not _SUPPORTS_SOFT_AFFINITY
                and 'soft-affinity' in group.policies):
            msg = _("ServerGroupSoftAffinityWeigher not configured")
            LOG.error(msg)
            raise exception.UnsupportedPolicyException(reason=msg)
        if (not _SUPPORTS_SOFT_ANTI_AFFINITY
                and 'soft-anti-affinity' in group.policies):
            msg = _("ServerGroupSoftAntiAffinityWeigher not configured")
            LOG.error(msg)
            raise exception.UnsupportedPolicyException(reason=msg)
        group_hosts = set(group.get_hosts())
        user_hosts = set(user_group_hosts) if user_group_hosts else set()
        return GroupDetails(hosts=user_hosts | group_hosts,
                            policies=group.policies, members=group.members)


def setup_instance_group(context, request_spec, filter_properties):
    """Add group_hosts and group_policies fields to filter_properties dict
    based on instance uuids provided in request_spec, if those instances are
    belonging to a group.

    :param request_spec: Request spec
    :param filter_properties: Filter properties
    """
    group_hosts = filter_properties.get('group_hosts')
    # NOTE(sbauza) If there are multiple instance UUIDs, it's a boot
    # request and they will all be in the same group, so it's safe to
    # only check the first one.
    instance_uuid = request_spec.get('instance_properties', {}).get('uuid')
    group_info = _get_group_details(context, instance_uuid, group_hosts)
    if group_info is not None:
        filter_properties['group_updated'] = True
        filter_properties['group_hosts'] = group_info.hosts
        filter_properties['group_policies'] = group_info.policies
        filter_properties['group_members'] = group_info.members


def retry_on_timeout(retries=1):
    """Retry the call in case a MessagingTimeout is raised.

    A decorator for retrying calls when a service dies mid-request.

    :param retries: Number of retries
    :returns: Decorator
    """
    def outer(func):
        @functools.wraps(func)
        def wrapped(*args, **kwargs):
            attempt = 0
            while True:
                try:
                    return func(*args, **kwargs)
                except messaging.MessagingTimeout:
                    attempt += 1
                    if attempt <= retries:
                        LOG.warning(_LW(
                            "Retrying %(name)s after a MessagingTimeout, "
                            "attempt %(attempt)s of %(retries)s."),
                                 {'attempt': attempt, 'retries': retries,
                                  'name': func.__name__})
                    else:
                        raise
        return wrapped
    return outer

retry_select_destinations = retry_on_timeout(CONF.scheduler_max_attempts - 1)


def request_is_rebuild(spec_obj):
    """Returns True if request is for a rebuild.

    :param spec_obj: An objects.RequestSpec to examine (or None).
    """
    if not spec_obj:
        return False
    if 'scheduler_hints' not in spec_obj:
        return False
    check_type = spec_obj.scheduler_hints.get('_nova_check_type')
    return check_type == ['rebuild']