OpenStack Orchestration (Heat)
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 

2638 lines
104 KiB

#
# 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.
import collections
import contextlib
import datetime as dt
import itertools
import pydoc
import tenacity
import weakref
from oslo_config import cfg
from oslo_log import log as logging
from oslo_utils import excutils
from oslo_utils import reflection
import six
from heat.common import exception
from heat.common.i18n import _
from heat.common import identifier
from heat.common import short_id
from heat.common import timeutils
from heat.engine import attributes
from heat.engine.cfn import template as cfn_tmpl
from heat.engine import clients
from heat.engine.clients import default_client_plugin
from heat.engine import environment
from heat.engine import event
from heat.engine import function
from heat.engine.hot import template as hot_tmpl
from heat.engine import node_data
from heat.engine import properties
from heat.engine import resources
from heat.engine import rsrc_defn
from heat.engine import scheduler
from heat.engine import status
from heat.engine import support
from heat.engine import sync_point
from heat.engine import template
from heat.objects import resource as resource_objects
from heat.objects import resource_data as resource_data_objects
from heat.objects import resource_properties_data as rpd_objects
from heat.rpc import client as rpc_client
cfg.CONF.import_opt('action_retry_limit', 'heat.common.config')
cfg.CONF.import_opt('observe_on_update', 'heat.common.config')
cfg.CONF.import_opt('error_wait_time', 'heat.common.config')
LOG = logging.getLogger(__name__)
datetime = dt.datetime
def _register_class(resource_type, resource_class):
resources.global_env().register_class(resource_type, resource_class)
# Attention developers about to move/delete this: STOP IT!!!
UpdateReplace = exception.UpdateReplace
# Attention developers about to move this: STOP IT!!!
class NoActionRequired(Exception):
"""Exception raised when a signal is ignored.
Resource subclasses should raise this exception from handle_signal() to
suppress recording of an event corresponding to the signal.
"""
def __init__(self, res_name='Unknown', reason=''):
msg = (_("The resource %(res)s could not perform "
"scaling action: %(reason)s") %
{'res': res_name, 'reason': reason})
super(Exception, self).__init__(six.text_type(msg))
class PollDelay(Exception):
"""Exception to delay polling of the resource.
This exception may be raised by a Resource subclass's check_*_complete()
methods to indicate that it need not be polled again immediately. If this
exception is raised, the check_*_complete() method will not be called
again until the nth time that the resource becomes eligible for polling.
A PollDelay period of 1 is equivalent to returning False.
"""
def __init__(self, period):
assert period >= 1
self.period = period
@six.python_2_unicode_compatible
class Resource(status.ResourceStatus):
BASE_ATTRIBUTES = (SHOW, ) = (attributes.SHOW_ATTR, )
LOCK_ACTIONS = (
LOCK_NONE, LOCK_ACQUIRE, LOCK_RELEASE, LOCK_RESPECT,
) = (
None, 1, -1, 0,
)
# If True, this resource must be created before it can be referenced.
strict_dependency = True
# Resource implementation set this to the subset of resource properties
# supported for handle_update, used by update_template_diff_properties
update_allowed_properties = ()
# Resource implementations set this to the name: description dictionary
# that describes the appropriate resource attributes
attributes_schema = {}
# Resource implementations set this to update policies
update_policy_schema = {}
# Default entity of resource, which is used for during resolving
# show attribute
entity = None
# Description dictionary, that describes the common attributes for all
# resources
base_attributes_schema = {
SHOW: attributes.Schema(
_("Detailed information about resource."),
cache_mode=attributes.Schema.CACHE_NONE,
type=attributes.Schema.MAP
)
}
# If True, this resource may perform authenticated API requests
# throughout its lifecycle
requires_deferred_auth = False
# Limit to apply to physical_resource_name() size reduction algorithm.
# If set to None no limit will be applied.
physical_resource_name_limit = 255
support_status = support.SupportStatus()
# Default name to use for calls to self.client()
default_client_name = None
# Required service extension for this resource
required_service_extension = None
# no signal actions
no_signal_actions = (status.ResourceStatus.SUSPEND,
status.ResourceStatus.DELETE)
# Whether all other resources need a metadata_update() after
# a signal to this resource
signal_needs_metadata_updates = True
# Whether the resource is always replaced when CHECK_FAILED
always_replace_on_check_failed = True
def __new__(cls, name, definition, stack):
"""Create a new Resource of the appropriate class for its type."""
assert isinstance(definition, rsrc_defn.ResourceDefinition)
if cls != Resource:
# Call is already for a subclass, so pass it through
ResourceClass = cls
else:
registry = stack.env.registry
ResourceClass = registry.get_class_to_instantiate(
definition.resource_type,
resource_name=name)
assert issubclass(ResourceClass, Resource)
return super(Resource, cls).__new__(ResourceClass)
@classmethod
def _validate_service_availability(cls, context, resource_type):
try:
(svc_available, reason) = cls.is_service_available(context)
except Exception as exc:
LOG.exception("Resource type %s unavailable",
resource_type)
ex = exception.ResourceTypeUnavailable(
resource_type=resource_type,
service_name=cls.default_client_name,
reason=six.text_type(exc))
raise ex
else:
if not svc_available:
ex = exception.ResourceTypeUnavailable(
resource_type=resource_type,
service_name=cls.default_client_name,
reason=reason)
LOG.info(six.text_type(ex))
raise ex
def __init__(self, name, definition, stack):
def _validate_name(res_name):
if '/' in res_name:
message = _('Resource name may not contain "/"')
raise exception.StackValidationFailed(message=message)
_validate_name(name)
self.stack = stack
self.context = stack.context
self.name = name
self.t = definition
self.reparse(client_resolve=False)
self.update_policy = self.t.update_policy(self.update_policy_schema,
self.context)
self._update_allowed_properties = self.calc_update_allowed(
self.properties)
self.attributes_schema.update(self.base_attributes_schema)
self.attributes = attributes.Attributes(self.name,
self.attributes_schema,
self._make_resolver(
weakref.ref(self)))
self.abandon_in_progress = False
self.resource_id = None
# if the stack is being deleted, assume we've already been deleted.
# or if the resource has not been created yet, and the stack was
# rollback, we set the resource to rollback
if stack.action == stack.DELETE or stack.action == stack.ROLLBACK:
self.action = stack.action
else:
self.action = self.INIT
self.status = self.COMPLETE
self.status_reason = ''
self.id = None
self.uuid = None
self._data = None
self._attr_data_id = None
self._rsrc_metadata = None
self._rsrc_prop_data_id = None
self._stored_properties_data = None
self.created_time = stack.created_time
self.updated_time = stack.updated_time
self._rpc_client = None
self.requires = set()
self.replaces = None
self.replaced_by = None
self.current_template_id = None
self.old_template_id = None
self.root_stack_id = None
self._calling_engine_id = None
self._atomic_key = None
self.converge = False
if not self.stack.in_convergence_check:
resource = stack.db_resource_get(name)
if resource:
self._load_data(resource)
else:
proxy = self.stack.defn[self.name]
node_data = proxy._resource_data
if node_data is not None:
self.action, self.status = proxy.state
self.id = node_data.primary_key
self.uuid = node_data.uuid
def rpc_client(self):
"""Return a client for making engine RPC calls."""
if not self._rpc_client:
self._rpc_client = rpc_client.EngineClient()
return self._rpc_client
def _load_data(self, resource):
"""Load the resource state from its DB representation."""
self.resource_id = resource.physical_resource_id
self.action = resource.action
self.status = resource.status
self.status_reason = resource.status_reason
self.id = resource.id
self.uuid = resource.uuid
try:
self._data = resource_data_objects.ResourceData.get_all(
self, resource.data)
except exception.NotFound:
self._data = {}
self.attributes.cached_attrs = resource.attr_data or None
self._attr_data_id = resource.attr_data_id
self._rsrc_metadata = resource.rsrc_metadata
self._stored_properties_data = resource.properties_data
self._rsrc_prop_data_id = resource.rsrc_prop_data_id
self.created_time = resource.created_at
self.updated_time = resource.updated_at
self.requires = set(resource.requires)
self.replaces = resource.replaces
self.replaced_by = resource.replaced_by
self.current_template_id = resource.current_template_id
self.root_stack_id = resource.root_stack_id
self._atomic_key = resource.atomic_key
@property
def external_id(self):
return self.t.external_id()
@classmethod
def getdoc(cls):
if cls.__doc__ is None:
return _('No description available')
return pydoc.getdoc(cls)
@property
def stack(self):
stack = self._stackref()
assert stack is not None, "Need a reference to the Stack object"
return stack
@stack.setter
def stack(self, stack):
self._stackref = weakref.ref(stack)
@classmethod
def load(cls, context, resource_id, current_traversal, is_update, data):
"""Load a specified resource from the database to check.
Returns a tuple of the Resource, the StackDefinition corresponding to
the resource's ResourceDefinition (i.e. the one the resource was last
updated to if it has already been created, or the one it will be
created with if it hasn't been already), and the Stack containing the
latest StackDefinition (i.e. the one that the latest traversal is
updating to.
The latter two must remain in-scope, because the Resource holds weak
references to them.
"""
from heat.engine import stack as stack_mod
db_res = resource_objects.Resource.get_obj(context, resource_id)
curr_stack = stack_mod.Stack.load(context, stack_id=db_res.stack_id,
cache_data=data)
initial_stk_defn = latest_stk_defn = curr_stack.defn
current_template_id = db_res.current_template_id
using_new_template = (current_template_id != curr_stack.t.id and
current_template_id is not None)
will_create = (db_res.action == cls.INIT and
is_update and
current_traversal == curr_stack.current_traversal)
if using_new_template and not will_create:
# load the definition associated with the resource's template
current_template = template.Template.load(context,
current_template_id)
initial_stk_defn = curr_stack.defn.clone_with_new_template(
current_template,
curr_stack.identifier())
curr_stack.defn = initial_stk_defn
res_defn = initial_stk_defn.resource_definition(db_res.name)
res_type = initial_stk_defn.env.registry.get_class_to_instantiate(
res_defn.resource_type, resource_name=db_res.name)
# If the resource type has changed and the new one is a valid
# substitution, use that as the class to instantiate.
if is_update and (latest_stk_defn is not initial_stk_defn):
try:
new_res_defn = latest_stk_defn.resource_definition(db_res.name)
except KeyError:
pass
else:
new_registry = latest_stk_defn.env.registry
new_res_type = new_registry.get_class_to_instantiate(
new_res_defn.resource_type, resource_name=db_res.name)
if res_type.check_is_substituted(new_res_type):
res_type = new_res_type
# Load only the resource in question; don't load all resources
# by invoking stack.resources. Maintain light-weight stack.
resource = res_type(db_res.name, res_defn, curr_stack)
resource._load_data(db_res)
curr_stack.defn = latest_stk_defn
return resource, initial_stk_defn, curr_stack
def make_replacement(self, new_tmpl_id, requires):
"""Create a replacement resource in the database.
Returns the DB ID of the new resource, or None if the new resource
cannot be created (generally because the template ID does not exist).
Raises UpdateInProgress if another traversal has already locked the
current resource.
"""
# 1. create the replacement with "replaces" = self.id
# Don't set physical_resource_id so that a create is triggered.
rs = {'stack_id': self.stack.id,
'name': self.name,
'rsrc_prop_data_id': None,
'needed_by': [],
'requires': sorted(requires, reverse=True),
'replaces': self.id,
'action': self.INIT,
'status': self.COMPLETE,
'current_template_id': new_tmpl_id,
'stack_name': self.stack.name,
'root_stack_id': self.root_stack_id}
update_data = {'status': self.COMPLETE}
# Retry in case a signal has updated the atomic_key
attempts = max(cfg.CONF.client_retry_limit, 0) + 1
def prepare_attempt(fn, attempt):
if attempt > 1:
res_obj = resource_objects.Resource.get_obj(
self.context, self.id)
if (res_obj.engine_id is not None or
res_obj.updated_at != self.updated_time):
raise exception.UpdateInProgress(resource_name=self.name)
self._atomic_key = res_obj.atomic_key
@tenacity.retry(
stop=tenacity.stop_after_attempt(attempts),
retry=tenacity.retry_if_exception_type(
exception.UpdateInProgress),
before=prepare_attempt,
wait=tenacity.wait_random(max=2),
reraise=True)
def create_replacement():
return resource_objects.Resource.replacement(self.context,
self.id,
update_data,
rs,
self._atomic_key)
new_rs = create_replacement()
if new_rs is None:
return None
self._incr_atomic_key(self._atomic_key)
self.replaced_by = new_rs.id
return new_rs.id
def reparse(self, client_resolve=True):
"""Reparse the resource properties.
Optional translate flag for property translation and
client_resolve flag for resolving properties by doing
client lookup.
"""
self.properties = self.t.properties(self.properties_schema,
self.context)
self.translate_properties(self.properties, client_resolve)
def calc_update_allowed(self, props):
update_allowed_set = set(self.update_allowed_properties)
for (psk, psv) in six.iteritems(props.props):
if psv.update_allowed():
update_allowed_set.add(psk)
return update_allowed_set
def __eq__(self, other):
"""Allow == comparison of two resources."""
# For the purposes of comparison, we declare two resource objects
# equal if their names and resolved templates are the same
if isinstance(other, Resource):
return ((self.name == other.name) and
(self.t.freeze() == other.t.freeze()))
return NotImplemented
def __ne__(self, other):
"""Allow != comparison of two resources."""
result = self.__eq__(other)
if result is NotImplemented:
return result
return not result
def __hash__(self):
return id(self)
def metadata_get(self, refresh=False):
if refresh:
self._rsrc_metadata = None
if self.id is None or self.action == self.INIT:
return self.t.metadata()
if self._rsrc_metadata is not None:
return self._rsrc_metadata
rs = resource_objects.Resource.get_obj(self.stack.context, self.id,
refresh=True,
fields=('rsrc_metadata', ))
self._rsrc_metadata = rs.rsrc_metadata
return rs.rsrc_metadata
@resource_objects.retry_on_conflict
def metadata_set(self, metadata, merge_metadata=None):
"""Write new metadata to the database.
The caller may optionally provide a merge_metadata() function, which
takes two arguments - the metadata passed to metadata_set() and the
current metadata of the resource - and returns the merged metadata to
write. If merge_metadata is not provided, the metadata passed to
metadata_set() is written verbatim, overwriting any existing metadata.
If a race condition is detected, the write will be retried with the new
result of merge_metadata() (if it is supplied) or the verbatim data (if
it is not).
"""
if self.id is None or self.action == self.INIT:
raise exception.ResourceNotAvailable(resource_name=self.name)
refresh = merge_metadata is not None
db_res = resource_objects.Resource.get_obj(
self.stack.context, self.id, refresh=refresh,
fields=('name', 'rsrc_metadata', 'atomic_key', 'engine_id',
'action', 'status'))
if db_res.action == self.DELETE:
self._db_res_is_deleted = True
LOG.debug("resource %(name)s, id: %(id)s is DELETE_%(st)s, "
"not setting metadata",
{'name': self.name, 'id': self.id, 'st': db_res.status})
raise exception.ResourceNotAvailable(resource_name=self.name)
LOG.debug('Setting metadata for %s', six.text_type(self))
if refresh:
metadata = merge_metadata(metadata, db_res.rsrc_metadata)
if db_res.update_metadata(metadata):
self._incr_atomic_key(db_res.atomic_key)
self._rsrc_metadata = metadata
def handle_metadata_reset(self):
"""Default implementation; should be overridden by resources.
Now we override this method to reset the metadata for scale-policy
and scale-group resources, because their metadata might hang in a
wrong state ('scaling_in_progress' is always True) if engine restarts
while scaling.
"""
pass
def _break_if_required(self, action, hook):
"""Block the resource until the hook is cleared if there is one."""
if self.stack.env.registry.matches_hook(self.name, hook):
self.trigger_hook(hook)
self._add_event(self.action, self.status,
"%(a)s paused until Hook %(h)s is cleared"
% {'a': action, 'h': hook})
LOG.info('Reached hook on %s', self)
while self.has_hook(hook):
try:
yield
except BaseException as exc:
self.clear_hook(hook)
self._add_event(
self.action, self.status,
"Failure occurred while waiting.")
if (isinstance(exc, AssertionError) or
not isinstance(exc, Exception)):
raise
def has_nested(self):
"""Return True if the resource has an existing nested stack.
For most resource types, this will always return False. StackResource
subclasses return True when appropriate. Resource subclasses that may
return True must also provide a nested_identifier() method to return
the identifier of the nested stack, and a nested() method to return a
Stack object for the nested stack.
"""
return False
def get_nested_parameters_stack(self):
"""Return the nested stack for schema validation.
Regular resources don't have such a thing.
"""
return
def has_hook(self, hook):
# Clear the cache to make sure the data is up to date:
self._data = None
return self.data().get(hook) == "True"
def trigger_hook(self, hook):
self.data_set(hook, "True")
def clear_hook(self, hook):
self.data_delete(hook)
def type(self):
return self.t.resource_type
def has_interface(self, resource_type):
"""Check if resource is mapped to resource_type or is "resource_type".
Check to see if this resource is either mapped to resource_type
or is a "resource_type".
"""
if self.type() == resource_type:
return True
try:
ri = self.stack.env.get_resource_info(self.type(),
self.name)
except exception.EntityNotFound:
return False
else:
return ri.name == resource_type
def identifier(self):
"""Return an identifier for this resource."""
return identifier.ResourceIdentifier(resource_name=self.name,
**self.stack.identifier())
def frozen_definition(self):
"""Return a frozen ResourceDefinition with stored property values.
The returned definition will contain the property values read from the
database, and will have all intrinsic functions resolved (note that
this makes it useless for calculating dependencies).
"""
if self._stored_properties_data is not None:
args = {'properties': self._stored_properties_data}
else:
args = {}
return self.t.freeze(**args)
@contextlib.contextmanager
def frozen_properties(self):
"""Context manager to use the frozen property values from the database.
The live property values are always substituted back when the context
ends.
"""
live_props = self.properties
props = self.frozen_definition().properties(self.properties_schema,
self.context)
try:
self.properties = props
yield props
finally:
self.properties = live_props
def update_template_diff(self, after, before):
"""Returns the difference between the before and after json snippets.
If something has been removed in after which exists in before we set it
to None.
"""
return after - before
def update_template_diff_properties(self, after_props, before_props):
"""Return changed Properties between the before and after properties.
If any property having immutable as True is updated, raises
NotSupported error.
If any properties have changed which are not in
update_allowed_properties, raises UpdateReplace.
"""
update_allowed_set = self.calc_update_allowed(after_props)
immutable_set = set()
for (psk, psv) in six.iteritems(after_props.props):
if psv.immutable():
immutable_set.add(psk)
def prop_changed(key):
try:
before = before_props.get(key)
except (TypeError, ValueError) as exc:
# We shouldn't get here usually, but there is a known issue
# with template resources and new parameters in non-convergence
# stacks (see bug 1543685). The error should be harmless
# because we're on the before properties, which have presumably
# already been validated.
LOG.warning('Ignoring error in old property value '
'%(prop_name)s: %(msg)s',
{'prop_name': key, 'msg': six.text_type(exc)})
return True
return before != after_props.get(key)
# Create a set of keys which differ (or are missing/added)
changed_properties_set = set(k for k in after_props if prop_changed(k))
# Create a list of updated properties offending property immutability
update_replace_forbidden = [k for k in changed_properties_set
if k in immutable_set]
if update_replace_forbidden:
msg = _("Update to properties %(props)s of %(name)s (%(res)s)"
) % {'props': ", ".join(sorted(update_replace_forbidden)),
'res': self.type(), 'name': self.name}
raise exception.NotSupported(feature=msg)
if changed_properties_set and self.needs_replace_with_prop_diff(
changed_properties_set,
after_props,
before_props):
raise UpdateReplace(self)
if not changed_properties_set.issubset(update_allowed_set):
raise UpdateReplace(self.name)
return dict((k, after_props.get(k)) for k in changed_properties_set)
def __str__(self):
class_name = reflection.get_class_name(self, fully_qualified=False)
if self.stack.id is not None:
if self.resource_id is not None:
text = '%s "%s" [%s] %s' % (class_name, self.name,
self.resource_id,
six.text_type(self.stack))
else:
text = '%s "%s" %s' % (class_name, self.name,
six.text_type(self.stack))
else:
text = '%s "%s"' % (class_name, self.name)
return six.text_type(text)
def add_explicit_dependencies(self, deps):
"""Add all dependencies explicitly specified in the template.
The deps parameter is a Dependencies object to which dependency pairs
are added.
"""
for dep in self.t.dependencies(self.stack):
deps += (self, dep)
deps += (self, None)
def add_dependencies(self, deps):
"""Add implicit dependencies specific to the resource type.
Some resource types may have implicit dependencies on other resources
in the same stack that are not linked by a property value (that would
be set using get_resource or get_attr for example, thus creating an
explicit dependency). Such dependencies are opaque to the user and
should be avoided wherever possible, however in some circumstances they
are required due to magic in the underlying API.
The deps parameter is a Dependencies object to which dependency pairs
may be added.
"""
return
def required_by(self):
"""List of resources that require this one as a dependency.
Returns a list of names of resources that depend on this resource
directly.
"""
try:
reqd_by = self.stack.dependencies.required_by(self)
except KeyError:
if self.stack.convergence:
# for convergence, fall back to building from node graph
needed_by_ids = self.stack.dependent_resource_ids(self.id)
reqd_by = [r for r in self.stack.resources.values()
if r.id in needed_by_ids]
else:
LOG.error('Getting required_by list for Resource not in '
'dependency graph.')
return []
return [r.name for r in reqd_by]
def client(self, name=None, version=None):
client_name = name or self.default_client_name
assert client_name, "Must specify client name"
return self.stack.clients.client(client_name, version)
def client_plugin(self, name=None):
client_name = name or self.default_client_name
assert client_name, "Must specify client name"
return self.stack.clients.client_plugin(client_name)
def _default_client_plugin(self):
"""Always return a client plugin.
This will be the client_plugin if the resource has defined a
default_client_name, or a no-op plugin if it does not. Thus, the
result of this call always has e.g. is_not_found() and is_conflict()
methods.
"""
cp = None
if self.default_client_name:
cp = self.client_plugin()
if cp is None:
cp = default_client_plugin.DefaultClientPlugin(self.context)
return cp
@classmethod
def is_service_available(cls, context):
# NOTE(kanagaraj-manickam): return True to satisfy the cases like
# resource does not have endpoint, such as RandomString, OS::Heat
# resources as they are implemented within the engine.
if cls.default_client_name is None:
return (True, None)
client_plugin = clients.Clients(context).client_plugin(
cls.default_client_name)
if not client_plugin:
raise exception.ClientNotAvailable(
client_name=cls.default_client_name)
service_types = client_plugin.service_types
if not service_types:
return (True, None)
# NOTE(kanagaraj-manickam): if one of the service_type does
# exist in the keystone, then considered it as available.
for service_type in service_types:
endpoint_exists = client_plugin.does_endpoint_exist(
service_type=service_type,
service_name=cls.default_client_name)
if endpoint_exists:
req_extension = cls.required_service_extension
is_ext_available = (
not req_extension or client_plugin.has_extension(
req_extension))
if is_ext_available:
return (True, None)
else:
reason = _('Required extension {0} in {1} service '
'is not available.')
reason = reason.format(req_extension,
cls.default_client_name)
else:
reason = _('{0} {1} endpoint is not in service catalog.')
reason = reason.format(cls.default_client_name, service_type)
return (False, reason)
def keystone(self):
return self.client('keystone')
def nova(self):
return self.client('nova')
def swift(self):
return self.client('swift')
def neutron(self):
return self.client('neutron')
def cinder(self):
return self.client('cinder')
def trove(self):
return self.client('trove')
def heat(self):
return self.client('heat')
def glance(self):
return self.client('glance')
def _incr_atomic_key(self, last_key):
if last_key is None:
self._atomic_key = 1
else:
self._atomic_key = last_key + 1
def _should_lock_on_action(self, action):
"""Return whether we should take a resource-level lock for an action.
In the legacy path, we always took a lock at the Stack level and never
at the Resource level. In convergence, we lock at the Resource level
for most operations. However, there are currently some exceptions:
the SUSPEND, RESUME, SNAPSHOT, and CHECK actions, and stack abandon.
"""
return (self.stack.convergence and
not self.abandon_in_progress and
action in {self.ADOPT,
self.CREATE,
self.UPDATE,
self.ROLLBACK,
self.DELETE})
@contextlib.contextmanager
def _action_recorder(self, action, expected_exceptions=tuple()):
"""Return a context manager to record the progress of an action.
Upon entering the context manager, the state is set to IN_PROGRESS.
Upon exiting, the state will be set to COMPLETE if no exception was
raised, or FAILED otherwise. Non-exit exceptions will be translated
to ResourceFailure exceptions.
Expected exceptions are re-raised, with the Resource moved to the
COMPLETE state.
"""
attempts = 1
first_iter = [True] # work around no nonlocal in py27
if self.stack.convergence:
if self._should_lock_on_action(action):
lock_acquire = self.LOCK_ACQUIRE
lock_release = self.LOCK_RELEASE
else:
lock_acquire = lock_release = self.LOCK_RESPECT
if action != self.CREATE:
attempts += max(cfg.CONF.client_retry_limit, 0)
else:
lock_acquire = lock_release = self.LOCK_NONE
# retry for convergence DELETE or UPDATE if we get the usual
# lock-acquire exception of exception.UpdateInProgress
@tenacity.retry(
stop=tenacity.stop_after_attempt(attempts),
retry=tenacity.retry_if_exception_type(
exception.UpdateInProgress),
wait=tenacity.wait_random(max=2),
reraise=True)
def set_in_progress():
if not first_iter[0]:
res_obj = resource_objects.Resource.get_obj(
self.context, self.id)
self._atomic_key = res_obj.atomic_key
else:
first_iter[0] = False
self.state_set(action, self.IN_PROGRESS, lock=lock_acquire)
try:
set_in_progress()
yield
except exception.UpdateInProgress as ex:
with excutils.save_and_reraise_exception():
LOG.info('Update in progress for %s', self.name)
except expected_exceptions as ex:
with excutils.save_and_reraise_exception():
self.state_set(action, self.COMPLETE, six.text_type(ex),
lock=lock_release)
LOG.debug('%s', six.text_type(ex))
except Exception as ex:
LOG.info('%(action)s: %(info)s',
{"action": action,
"info": six.text_type(self)},
exc_info=True)
failure = exception.ResourceFailure(ex, self, action)
self.state_set(action, self.FAILED, six.text_type(failure),
lock=lock_release)
raise failure
except BaseException as exc:
with excutils.save_and_reraise_exception():
try:
reason = six.text_type(exc)
msg = '%s aborted' % action
if reason:
msg += ' (%s)' % reason
self.state_set(action, self.FAILED, msg,
lock=lock_release)
except Exception:
LOG.exception('Error marking resource as failed')
else:
self.state_set(action, self.COMPLETE, lock=lock_release)
def action_handler_task(self, action, args=None, action_prefix=None):
"""A task to call the Resource subclass's handler methods for action.
Calls the handle_<ACTION>() method for the given action and then calls
the check_<ACTION>_complete() method with the result in a loop until it
returns True. If the methods are not provided, the call is omitted.
Any args provided are passed to the handler.
If a prefix is supplied, the handler method handle_<PREFIX>_<ACTION>()
is called instead.
"""
args = args or []
handler_action = action.lower()
check = getattr(self, 'check_%s_complete' % handler_action, None)
if action_prefix:
handler_action = '%s_%s' % (action_prefix.lower(), handler_action)
handler = getattr(self, 'handle_%s' % handler_action, None)
if callable(handler):
try:
handler_data = handler(*args)
except StopIteration:
raise RuntimeError('Plugin method raised StopIteration')
yield
if callable(check):
try:
while True:
try:
done = check(handler_data)
except PollDelay as delay:
yield delay.period
else:
if done:
break
else:
yield
except StopIteration:
raise RuntimeError('Plugin method raised StopIteration')
except Exception:
raise
except: # noqa
with excutils.save_and_reraise_exception():
canceller = getattr(
self,
'handle_%s_cancel' % handler_action,
None
)
if callable(canceller):
try:
canceller(handler_data)
except Exception:
LOG.exception(
'Error cancelling resource %s',
action
)
@scheduler.wrappertask
def _do_action(self, action, pre_func=None, resource_data=None):
"""Perform a transition to a new state via a specified action.
Action should be e.g self.CREATE, self.UPDATE etc, we set
status based on this, the transition is handled by calling the
corresponding handle_* and check_*_complete functions
Note pre_func is an optional function reference which will
be called before the handle_<action> function
If the resource does not declare a check_$action_complete function,
we declare COMPLETE status as soon as the handle_$action call has
finished, and if no handle_$action function is declared, then we do
nothing, useful e.g if the resource requires no action for a given
state transition
"""
assert action in self.ACTIONS, 'Invalid action %s' % action
with self._action_recorder(action):
if callable(pre_func):
pre_func()
handler_args = [resource_data] if resource_data is not None else []
yield self.action_handler_task(action, args=handler_args)
def _update_stored_properties(self):
old_props = self._stored_properties_data
self._stored_properties_data = function.resolve(self.properties.data)
if self._stored_properties_data != old_props:
self._rsrc_prop_data_id = None
self.attributes.reset_resolved_values()
def referenced_attrs(self, stk_defn=None,
in_resources=True, in_outputs=True,
load_all=False):
"""Return the set of all attributes referenced in the template.
This enables the resource to calculate which of its attributes will
be used. By default, attributes referenced in either other resources
or outputs will be included. Either can be excluded by setting the
`in_resources` or `in_outputs` parameters to False. To limit to a
subset of outputs, pass an iterable of the output names to examine
for the `in_outputs` parameter.
The set of referenced attributes is calculated from the
StackDefinition object provided, or from the stack's current
definition if none is passed.
"""
if stk_defn is None:
stk_defn = self.stack.defn
def get_dep_attrs(source):
return set(itertools.chain.from_iterable(s.dep_attrs(self.name,
load_all)
for s in source))
refd_attrs = set()
if in_resources:
enabled_resources = stk_defn.enabled_rsrc_names()
refd_attrs |= get_dep_attrs(stk_defn.resource_definition(r_name)
for r_name in enabled_resources)
subset_outputs = isinstance(in_outputs, collections.Iterable)
if subset_outputs or in_outputs:
if not subset_outputs:
in_outputs = stk_defn.enabled_output_names()
refd_attrs |= get_dep_attrs(stk_defn.output_definition(op_name)
for op_name in in_outputs)
if attributes.ALL_ATTRIBUTES in refd_attrs:
refd_attrs.remove(attributes.ALL_ATTRIBUTES)
refd_attrs |= (set(self.attributes) - {self.SHOW})
return refd_attrs
def node_data(self, stk_defn=None, for_resources=True, for_outputs=False):
"""Return a NodeData object representing the resource.
The NodeData object returned contains basic data about the resource,
including its name, ID and state, as well as its reference ID and any
attribute values that are used.
By default, those attribute values that are referenced by other
resources are included. These can be ignored by setting the
for_resources parameter to False. If the for_outputs parameter is
True, those attribute values that are referenced by stack outputs are
included. If the for_outputs parameter is an iterable of output names,
only those attribute values referenced by the specified stack outputs
are included.
The set of referenced attributes is calculated from the
StackDefinition object provided, or from the stack's current
definition if none is passed.
After calling this method, the resource's attribute cache is
populated with any cacheable attribute values referenced by stack
outputs, even if they are not also referenced by other resources.
"""
def get_attrs(attrs, cacheable_only=False):
for attr in attrs:
path = (attr,) if isinstance(attr, six.string_types) else attr
if (cacheable_only and
(self.attributes.get_cache_mode(path[0]) ==
attributes.Schema.CACHE_NONE)):
continue
if self.action == self.INIT:
if (path[0] in self.attributes or
(type(self).get_attribute != Resource.get_attribute or
type(self).FnGetAtt != Resource.FnGetAtt)):
# TODO(ricolin) make better placeholder values here
yield attr, None
else:
try:
yield attr, self.FnGetAtt(*path)
except exception.InvalidTemplateAttribute as ita:
# Attribute doesn't exist, so don't store it. Whatever
# tries to access it will get another
# InvalidTemplateAttribute exception at that point
LOG.info('%s', ita)
except Exception as exc:
# Store the exception that occurred. It will be
# re-raised when something tries to access it, or when
# we try to serialise the NodeData.
yield attr, exc
load_all = not self.stack.in_convergence_check
dep_attrs = self.referenced_attrs(stk_defn,
in_resources=for_resources,
in_outputs=for_outputs,
load_all=load_all)
# Ensure all attributes referenced in outputs get cached
if for_outputs is False and self.stack.convergence:
out_attrs = self.referenced_attrs(stk_defn, in_resources=False,
load_all=load_all)
for e in get_attrs(out_attrs - dep_attrs, cacheable_only=True):
pass
# Calculate attribute values *before* reference ID, to potentially
# save an extra RPC call in TemplateResource
attribute_values = dict(get_attrs(dep_attrs))
return node_data.NodeData(self.id, self.name, self.uuid,
self.FnGetRefId(), attribute_values,
self.action, self.status)
def preview(self):
"""Default implementation of Resource.preview.
This method should be overridden by child classes for specific
behavior.
"""
return self
def create_convergence(self, template_id, requires, engine_id,
timeout, progress_callback=None):
"""Creates the resource by invoking the scheduler TaskRunner."""
self._calling_engine_id = engine_id
self.requires = requires
self.current_template_id = template_id
if self.stack.adopt_stack_data is None:
runner = scheduler.TaskRunner(self.create)
else:
adopt_data = self.stack._adopt_kwargs(self)
runner = scheduler.TaskRunner(self.adopt, **adopt_data)
runner(timeout=timeout, progress_callback=progress_callback)
def validate_external(self):
if self.external_id is not None:
try:
self.resource_id = self.external_id
self._show_resource()
except Exception as ex:
if self._default_client_plugin().is_not_found(ex):
error_message = (_("Invalid external resource: Resource "
"%(external_id)s (%(type)s) can not "
"be found.") %
{'external_id': self.external_id,
'type': self.type()})
raise exception.StackValidationFailed(
message="%s" % error_message)
raise
@scheduler.wrappertask
def create(self):
"""Create the resource.
Subclasses should provide a handle_create() method to customise
creation.
"""
action = self.CREATE
if (self.action, self.status) != (self.INIT, self.COMPLETE):
exc = exception.Error(_('State %s invalid for create')
% six.text_type(self.state))
raise exception.ResourceFailure(exc, self, action)
if self.external_id is not None:
yield self._do_action(self.ADOPT,
resource_data={
'resource_id': self.external_id})
yield self.check()
return
# This method can be called when we replace a resource, too. In that
# case, a hook has already been dealt with in `Resource.update` so we
# shouldn't do it here again:
if self.stack.action == self.stack.CREATE:
yield self._break_if_required(
self.CREATE, environment.HOOK_PRE_CREATE)
LOG.info('creating %s', self)
# Re-resolve the template, since if the resource Ref's
# the StackId pseudo parameter, it will change after
# the parser.Stack is stored (which is after the resources
# are __init__'d, but before they are create()'d). We also
# do client lookups for RESOLVE translation rules here.
self.reparse()
self._update_stored_properties()
count = {self.CREATE: 0, self.DELETE: 0}
retry_limit = max(cfg.CONF.action_retry_limit, 0)
first_failure = None
while (count[self.CREATE] <= retry_limit and
count[self.DELETE] <= retry_limit):
pre_func = None
if count[action] > 0:
delay = timeutils.retry_backoff_delay(count[action],
jitter_max=2.0)
waiter = scheduler.TaskRunner(self.pause)
yield waiter.as_task(timeout=delay)
elif action == self.CREATE:
# Only validate properties in first create call.
pre_func = self.properties.validate
try:
yield self._do_action(action, pre_func)
if action == self.CREATE:
first_failure = None
break
else:
action = self.CREATE
except exception.ResourceFailure as failure:
exc = failure.exc
if isinstance(exc, exception.StackValidationFailed):
path = [self.t.name]
path.extend(exc.path)
raise exception.ResourceFailure(
exception_or_error=exception.StackValidationFailed(
error=exc.error,
path=path,
message=exc.error_message
),
resource=failure.resource,
action=failure.action
)
if not (isinstance(exc, exception.ResourceInError) or
self._default_client_plugin().is_conflict(exc)):
raise failure
count[action] += 1
if action == self.CREATE:
action = self.DELETE
count[action] = 0
if first_failure is None:
# Save the first exception
first_failure = failure
if first_failure:
raise first_failure
if self.stack.action == self.stack.CREATE:
yield self._break_if_required(
self.CREATE, environment.HOOK_POST_CREATE)
@staticmethod
def pause():
try:
while True:
yield
except scheduler.Timeout:
return
def prepare_abandon(self):
self.abandon_in_progress = True
return {
'name': self.name,
'resource_id': self.resource_id,
'type': self.type(),
'action': self.action,
'status': self.status,
'metadata': self.metadata_get(),
'resource_data': self.data()
}
def adopt(self, resource_data):
"""Adopt the existing resource.
Resource subclasses can provide a handle_adopt() method to customise
adopt.
"""
self._update_stored_properties()
return self._do_action(self.ADOPT, resource_data=resource_data)
def handle_adopt(self, resource_data=None):
resource_id, data, metadata = self._get_resource_info(resource_data)
if not resource_id:
exc = Exception(_('Resource ID was not provided.'))
failure = exception.ResourceFailure(exc, self)
raise failure
# set resource id
self.resource_id_set(resource_id)
# save the resource data
if data and isinstance(data, dict):
for key, value in six.iteritems(data):
self.data_set(key, value)
# save the resource metadata
self.metadata_set(metadata)
def translation_rules(self, properties):
"""Return specified rules for resource."""
return []
def translate_properties(self, properties,
client_resolve=True, ignore_resolve_error=False):
"""Set resource specific rules for properties translation.
The properties parameter is a properties object and the
optional client_resolve flag is to specify whether to
do 'RESOLVE' translation with client lookup.
"""
rules = self.translation_rules(properties) or []
properties.update_translation(
rules, client_resolve=client_resolve,
ignore_resolve_error=ignore_resolve_error)
def cancel_grace_period(self):
canceller = getattr(self,
'handle_%s_cancel' % self.action.lower(),
None)
if callable(canceller):
return None
return cfg.CONF.error_wait_time
def _get_resource_info(self, resource_data):
if not resource_data:
return None, None, None
return (resource_data.get('resource_id'),
resource_data.get('resource_data'),
resource_data.get('metadata'))
def needs_replace(self, after_props):
"""Mandatory replace based on certain properties."""
return False
def needs_replace_with_prop_diff(self, changed_properties_set,
after_props, before_props):
"""Needs replace based on prop_diff."""
return False
def needs_replace_with_tmpl_diff(self, tmpl_diff):
"""Needs replace based on tmpl_diff."""
return False
def needs_replace_failed(self):
"""Needs replace if resource is in ``*_FAILED``."""
return True
def _needs_update(self, after, before, after_props, before_props,
prev_resource, check_init_complete=True):
if self.status == self.FAILED:
# always replace when a resource is in CHECK_FAILED
if ((self.action == self.CHECK and
self.always_replace_on_check_failed) or (
self.needs_replace_failed())):
raise UpdateReplace(self)
if self.state == (self.DELETE, self.COMPLETE):
raise UpdateReplace(self)
if (check_init_complete and
self.state == (self.INIT, self.COMPLETE)):
raise UpdateReplace(self)
if self.needs_replace(after_props):
raise UpdateReplace(self)
if before != after.freeze():
return True
try:
return before_props != after_props
except ValueError:
return True
def _check_for_convergence_replace(self, restricted_actions):
if 'replace' in restricted_actions:
ex = exception.ResourceActionRestricted(action='replace')
failure = exception.ResourceFailure(ex, self, self.UPDATE)
self._add_event(self.UPDATE, self.FAILED, six.text_type(ex))
raise failure
else:
raise UpdateReplace(self.name)
def update_convergence(self, template_id, new_requires, engine_id,
timeout, new_stack, progress_callback=None):
"""Update the resource synchronously.
Persist the resource's current_template_id to template_id and
resource's requires to list of the required resource ids from the given
resource_data and existing resource's requires, then updates the
resource by invoking the scheduler TaskRunner.
"""
self._calling_engine_id = engine_id
# Check that the resource type matches. If the type has changed by a
# legitimate substitution, the load()ed resource will already be of
# the new type.
registry = new_stack.env.registry
new_res_def = new_stack.defn.resource_definition(self.name)
new_res_type = registry.get_class_to_instantiate(
new_res_def.resource_type, resource_name=self.name)
if type(self) is not new_res_type:
restrictions = registry.get_rsrc_restricted_actions(self.name)
self._check_for_convergence_replace(restrictions)
action_rollback = self.stack.action == self.stack.ROLLBACK
status_in_progress = self.stack.status == self.stack.IN_PROGRESS
if action_rollback and status_in_progress and self.replaced_by:
try:
self.restore_prev_rsrc(convergence=True)
except Exception as e:
failure = exception.ResourceFailure(e, self, self.action)
self.state_set(self.UPDATE, self.FAILED,
six.text_type(failure))
raise failure
self.replaced_by = None
runner = scheduler.TaskRunner(self.update, new_res_def,
new_template_id=template_id,
new_requires=new_requires)
runner(timeout=timeout, progress_callback=progress_callback)
def preview_update(self, after, before, after_props, before_props,
prev_resource, check_init_complete=False):
"""Simulates update without actually updating the resource.
Raises UpdateReplace, if replacement is required or returns True,
if in-place update is required.
"""
if self._needs_update(after, before, after_props, before_props,
prev_resource, check_init_complete):
tmpl_diff = self.update_template_diff(after.freeze(), before)
if tmpl_diff and self.needs_replace_with_tmpl_diff(tmpl_diff):
raise UpdateReplace(self)
self.update_template_diff_properties(after_props, before_props)
return True
else:
return False
def _check_restricted_actions(self, actions, after, before,
after_props, before_props,
prev_resource):
"""Checks for restricted actions.
Raises ResourceActionRestricted, if the resource requires update
or replace and the required action is restricted.
Else, Raises UpdateReplace, if replacement is required or returns
True, if in-place update is required.
"""
try:
if self.preview_update(after, before, after_props, before_props,
prev_resource, check_init_complete=True):
if 'update' in actions:
raise exception.ResourceActionRestricted(action='update')
return True
except UpdateReplace:
if 'replace' in actions:
raise exception.ResourceActionRestricted(action='replace')
raise
return False
def _prepare_update_props(self, after, before):
before_props = before.properties(self.properties_schema,
self.context)
# Regenerate the schema, else validation would fail
self.regenerate_info_schema(after)
after.set_translation_rules(self.translation_rules(self.properties))
after_props = after.properties(self.properties_schema,
self.context)
self.translate_properties(after_props)
self.translate_properties(before_props, ignore_resolve_error=True)
if (cfg.CONF.observe_on_update or self.converge) and before_props:
if not self.resource_id:
raise UpdateReplace(self)
try:
resource_reality = self.get_live_state(before_props)
if resource_reality:
self._update_properties_with_live_state(before_props,
resource_reality)
except exception.EntityNotFound:
raise UpdateReplace(self)
except Exception as ex:
LOG.warning("Resource cannot be updated with it's "
"live state in case of next "
"error: %s", ex)
return after_props, before_props
def _prepare_update_replace_handler(self, action):
"""Return the handler method for preparing to replace a resource.
This may be either restore_prev_rsrc() (in the case of a legacy
rollback) or, more typically, prepare_for_replace().
If the plugin has not overridden the method, then None is returned in
place of the default method (which is empty anyway).
"""
if (self.stack.action == 'ROLLBACK' and
self.stack.status == 'IN_PROGRESS' and
not self.stack.convergence):
# handle case, when it's rollback and we should restore
# old resource
if self.restore_prev_rsrc != Resource.restore_prev_rsrc:
return self.restore_prev_rsrc
else:
if self.prepare_for_replace != Resource.prepare_for_replace:
return self.prepare_for_replace
return None
def _prepare_update_replace(self, action):
handler = self._prepare_update_replace_handler(action)
if handler is None:
return
try:
handler()
except Exception as e:
# if any exception happen, we should set the resource to
# FAILED, then raise ResourceFailure
failure = exception.ResourceFailure(e, self, action)
self.state_set(action, self.FAILED, six.text_type(failure))
raise failure
@classmethod
def check_is_substituted(cls, new_res_type):
support_status = getattr(cls, 'support_status', None)
if support_status:
is_substituted = support_status.is_substituted(new_res_type)
return is_substituted
return False
def _persist_update_no_change(self, new_template_id):
"""Persist an update where the resource is unchanged."""
if new_template_id is not None:
self.current_template_id = new_template_id
lock = (self.LOCK_RESPECT if self.stack.convergence
else self.LOCK_NONE)
if self.status == self.FAILED:
status_reason = _('Update status to COMPLETE for '
'FAILED resource neither update '
'nor replace.')
self.state_set(self.action, self.COMPLETE,
status_reason, lock=lock)
elif new_template_id is not None:
self.store(lock=lock)
@scheduler.wrappertask
def update(self, after, before=None, prev_resource=None,
new_template_id=None, new_requires=None):
"""Return a task to update the resource.
Subclasses should provide a handle_update() method to customise update,
the base-class handle_update will fail by default.
"""
action = self.UPDATE
assert isinstance(after, rsrc_defn.ResourceDefinition)
if before is None:
before = self.frozen_definition()
after_external_id = after.external_id()
if self.external_id != after_external_id:
msg = _("Update to property %(prop)s of %(name)s (%(res)s)"
) % {'prop': hot_tmpl.HOTemplate20161014.RES_EXTERNAL_ID,
'res': self.type(), 'name': self.name}
exc = exception.NotSupported(feature=msg)
raise exception.ResourceFailure(exc, self, action)
elif after_external_id is not None:
LOG.debug("Skip update on external resource.")
self._persist_update_no_change(new_template_id)
return
after_props, before_props = self._prepare_update_props(after, before)
yield self._break_if_required(
self.UPDATE, environment.HOOK_PRE_UPDATE)
try:
registry = self.stack.env.registry
restr_actions = registry.get_rsrc_restricted_actions(self.name)
if restr_actions:
needs_update = self._check_restricted_actions(restr_actions,
after, before,
after_props,
before_props,
prev_resource)
else:
needs_update = self._needs_update(after, before,
after_props, before_props,
prev_resource)
except UpdateReplace:
with excutils.save_and_reraise_exception():
if self._prepare_update_replace_handler(action) is not None:
with self.lock(self._calling_engine_id):
self._prepare_update_replace(action)
except exception.ResourceActionRestricted as ae:
failure = exception.ResourceFailure(ae, self, action)
self._add_event(action, self.FAILED, six.text_type(ae))
raise failure
if not needs_update:
self._persist_update_no_change(new_template_id)
return
if not self.stack.convergence:
if (self.action, self.status) in (
(self.CREATE, self.IN_PROGRESS),
(self.UPDATE, self.IN_PROGRESS),
(self.ADOPT, self.IN_PROGRESS)):
exc = Exception(_('Resource update already requested'))
raise exception.ResourceFailure(exc, self, action)
LOG.info('updating %s', self)
self.updated_time = datetime.utcnow()
if new_requires is not None:
self.requires = self.requires | new_requires
with self._action_recorder(action, UpdateReplace):
after_props.validate()
self.properties = before_props
tmpl_diff = self.update_template_diff(after.freeze(), before)
self.old_template_id = self.current_template_id
try:
if tmpl_diff and self.needs_replace_with_tmpl_diff(tmpl_diff):
raise UpdateReplace(self)
prop_diff = self.update_template_diff_properties(after_props,
before_props)
if new_template_id is not None:
self.current_template_id = new_template_id
yield self.action_handler_task(action,
args=[after, tmpl_diff,
prop_diff])
except UpdateReplace:
with excutils.save_and_reraise_exception():
self.current_template_id = self.old_template_id
self.old_template_id = None
self._prepare_update_replace(action)
self.t = after
self.reparse()
self._update_stored_properties()
if new_requires is not None:
self.requires = new_requires
yield self._break_if_required(
self.UPDATE, environment.HOOK_POST_UPDATE)
def prepare_for_replace(self):
"""Prepare resource for replacing.
Some resources requires additional actions before replace them.
If resource need to be changed before replacing, this method should
be implemented in resource class.
"""
pass
def restore_prev_rsrc(self, convergence=False):
"""Restore resource after rollback.
Some resources requires additional actions after rollback.
If resource need to be changed during rollback, this method should
be implemented in resource class.
"""
pass
def check(self):
"""Checks that the physical resource is in its expected state.
Gets the current status of the physical resource and updates the
database accordingly. If check is not supported by the resource,
default action is to fail and revert the resource's status to its
original state with the added message that check was not performed.
"""
action = self.CHECK
LOG.info('Checking %s', self)
if hasattr(self, 'handle_%s' % action.lower()):
if self.state == (self.INIT, self.COMPLETE):
reason = _('Can not check %s, resource not '
'created yet.') % self.name
self.state_set(action, self.FAILED, reason)
exc = Exception(_('Resource %s not created yet.') % self.name)
failure = exception.ResourceFailure(exc, self, action)
raise failure
with self.frozen_properties():
return self._do_action(action)
else:
if self.state == (self.INIT, self.COMPLETE):
# No need to store status; better to leave the resource in
# INIT_COMPLETE than imply that we've checked and it exists.
return
reason = '%s not supported for %s' % (action, self.type())
self.state_set(action, self.COMPLETE, reason)
def _verify_check_conditions(self, checks):
def valid(check):
if isinstance(check['expected'], list):
return check['current'] in check['expected']
else:
return check['current'] == check['expected']
msg = _("'%(attr)s': expected '%(expected)s', got '%(current)s'")
invalid_checks = [
msg % check
for check in checks
if not valid(check)
]
if invalid_checks:
raise exception.Error('; '.join(invalid_checks))
def suspend(self):
"""Return a task to suspend the resource.
Subclasses should provide a handle_suspend() method to implement
suspend.
"""
action = self.SUSPEND
# Don't try to suspend the resource unless it's in a stable state
# or if the previous suspend failed
if (self.action == self.DELETE or
(self.action != self.SUSPEND and
self.status != self.COMPLETE)):
exc = exception.Error(_('State %s invalid for suspend')
% six.text_type(self.state))
raise exception.ResourceFailure(exc, self, action)
LOG.info('suspending %s', self)
with self.frozen_properties():
return self._do_action(action)
def resume(self):
"""Return a task to resume the resource.
Subclasses should provide a handle_resume() method to implement resume.
"""
action = self.RESUME
# Allow resume a resource if it's SUSPEND_COMPLETE
# or RESUME_FAILED or RESUME_COMPLETE. Recommend to check
# the real state of physical resource in handle_resume()
if self.state not in ((self.SUSPEND, self.COMPLETE),
(self.RESUME, self.FAILED),
(self.RESUME, self.COMPLETE)):
exc = exception.Error(_('State %s invalid for resume')
% six.text_type(self.state))
raise exception.ResourceFailure(exc, self, action)
LOG.info('resuming %s', self)
with self.frozen_properties():
return self._do_action(action)
def snapshot(self):
"""Snapshot the resource and return the created data, if any."""
LOG.info('snapshotting %s', self)
with self.frozen_properties():
return self._do_action(self.SNAPSHOT)
@scheduler.wrappertask
def delete_snapshot(self, data):
yield self.action_handler_task('delete_snapshot', args=[data])
def physical_resource_name(self):
if self.id is None or self.action == self.INIT:
return None
name = '%s-%s-%s' % (self.stack.name.rstrip('*'),
self.name,
short_id.get_id(self.uuid))
if self.physical_resource_name_limit:
name = self.reduce_physical_resource_name(
name, self.physical_resource_name_limit)
return name
@staticmethod
def reduce_physical_resource_name(name, limit):
"""Reduce length of physical resource name to a limit.
The reduced name will consist of the following:
* the first 2 characters of the name
* a hyphen
* the end of the name, truncated on the left to bring
the name length within the limit
:param name: The name to reduce the length of
:param limit: The max length limit
:returns: A name whose length is less than or equal to the limit
"""
if len(name) <= limit:
return name
if limit < 4:
raise ValueError(_('limit cannot be less than 4'))
postfix_length = limit - 3
return name[0:2] + '-' + name[-postfix_length:]
def validate(self):
"""Validate the resource.
This may be overridden by resource plugins to add extra
validation logic specific to the resource implementation.
"""
LOG.info('Validating %s', self)
return self.validate_template()
def validate_template(self):
"""Validate structural/syntax aspects of the resource definition.
Resource plugins should not override this, because this interface
is expected to be called pre-create so things normally valid
in an overridden validate() such as accessing properties
may not work.
"""
self._validate_service_availability(
self.stack.context,
self.t.resource_type
)
try:
self.t.validate()
self.validate_deletion_policy(self.t.deletion_policy())
self.t.update_policy(self.update_policy_schema,
self.context).validate()
validate = self.properties.validate(
with_value=self.stack.strict_validate)
except exception.StackValidationFailed as ex:
path = [self.stack.t.RESOURCES, self.t.name]
if ex.path:
path.append(self.stack.t.get_section_name(ex.path[0]))
path.extend(ex.path[1:])
raise exception.StackValidationFailed(
error=ex.error,
path=path,
message=ex.error_message)
return validate
@classmethod
def validate_deletion_policy(cls, policy):
path = rsrc_defn.DELETION_POLICY
if policy not in rsrc_defn.ResourceDefinition.DELETION_POLICIES:
msg = _('Invalid deletion policy "%s"') % policy
raise exception.StackValidationFailed(message=msg, path=path)
if policy == rsrc_defn.ResourceDefinition.SNAPSHOT:
if not callable(getattr(cls, 'handle_snapshot_delete', None)):
msg = _('"%s" deletion policy not supported') % policy
raise exception.StackValidationFailed(message=msg, path=path)
def _update_replacement_data(self, template_id):
# Update the replacement resource's replaces field.
# Make sure that the replacement belongs to the given
# template and there is no engine working on it.
if self.replaced_by is None:
return
try:
db_res = resource_objects.Resource.get_obj(
self.context, self.replaced_by,
fields=('current_template_id', 'atomic_key'))
except exception.NotFound:
LOG.info("Could not find replacement of resource %(name)s "
"with id %(id)s while updating replaces.",
{'name': self.name, 'id': self.replaced_by})
return
if (db_res.current_template_id == template_id):
# Following update failure is ignorable; another
# update might have locked/updated the resource.
db_res.select_and_update({'replaces': None},
atomic_key=db_res.atomic_key,
expected_engine_id=None)
def delete_convergence(self, template_id, engine_id, timeout,
progress_callback=None):
"""Destroys the resource if it doesn't belong to given template.
The given template is suppose to be the current template being
provisioned.
Also, since this resource is visited as part of clean-up phase,
the needed_by should be updated. If this resource was
replaced by more recent resource, then delete this and update
the replacement resource's replaces field.
"""
self._calling_engine_id = engine_id
if self.current_template_id != template_id:
# just delete the resources in INIT state
if self.action == self.INIT:
try:
resource_objects.Resource.delete(self.context, self.id)
except exception.NotFound:
pass
else:
runner = scheduler.TaskRunner(self.delete)
runner(timeout=timeout,
progress_callback=progress_callback)
self._update_replacement_data(template_id)
def handle_delete(self):
"""Default implementation; should be overridden by resources."""
if self.entity and self.resource_id is not None:
with self._default_client_plugin().ignore_not_found:
obj = getattr(self.client(), self.entity)
obj.delete(self.resource_id)
return self.resource_id
return None
@scheduler.wrappertask
def delete(self):
"""A task to delete the resource.
Subclasses should provide a handle_delete() method to customise
deletion.
"""
@excutils.exception_filter
def should_retry(exc):
if count >= retry_limit:
return False
return (self._default_client_plugin().is_conflict(exc) or
isinstance(exc, exception.PhysicalResourceExists))
action = self.DELETE
if (self.action, self.status) == (self.DELETE, self.COMPLETE):
return
# No need to delete if the resource has never been created
if self.action == self.INIT:
return
initial_state = self.state
# This method can be called when we replace a resource, too. In that
# case, a hook has already been dealt with in `Resource.update` so we
# shouldn't do it here again:
if self.stack.action == self.stack.DELETE:
yield self._break_if_required(
self.DELETE, environment.HOOK_PRE_DELETE)
LOG.info('deleting %s', self)
if self._stored_properties_data is not None:
# On delete we can't rely on re-resolving the properties
# so use the stored frozen_definition instead
self.properties = self.frozen_definition().properties(
self.properties_schema, self.context)
self.translate_properties(self.properties,
ignore_resolve_error=True)
with self._action_recorder(action):
if self.abandon_in_progress:
deletion_policy = self.t.RETAIN
else:
deletion_policy = self.t.deletion_policy()
if deletion_policy != self.t.RETAIN:
if deletion_policy == self.t.SNAPSHOT:
action_args = [[initial_state], 'snapshot']
else:
action_args = []
count = -1
retry_limit = max(cfg.CONF.action_retry_limit, 0)
while True:
count += 1
LOG.info('delete %(name)s attempt %(attempt)d' %
{'name': six.text_type(self), 'attempt': count+1})
if count:
delay = timeutils.retry_backoff_delay(count,
jitter_max=2.0)
waiter = scheduler.TaskRunner(self.pause)
yield waiter.as_task(timeout=delay)
with excutils.exception_filter(should_retry):
yield self.action_handler_task(action,
*action_args)
break
if self.stack.action == self.stack.DELETE:
yield self._break_if_required(
self.DELETE, environment.HOOK_POST_DELETE)
@scheduler.wrappertask
def destroy(self):
"""A task to delete the resource and remove it from the database."""
yield self.delete()
if self.id is None:
return
try:
resource_objects.Resource.delete(self.context, self.id)
except exception.NotFound:
# Don't fail on delete if the db entry has
# not been created yet.
pass
self.id = None
def resource_id_set(self, inst):
self.resource_id = inst
if self.id is not None:
try:
resource_objects.Resource.update_by_id(
self.context,
self.id,
{'physical_resource_id': self.resource_id})
except Exception as ex:
LOG.warning('db error %s', ex)
def store(self, set_metadata=False, lock=LOCK_NONE):
"""Create the resource in the database.
If self.id is set, we update the existing stack.
"""
if not self.root_stack_id:
self.root_stack_id = self.stack.root_stack_id()
rs = {'action': self.action,
'status': self.status,
'status_reason': six.text_type(self.status_reason),
'stack_id': self.stack.id,
'physical_resource_id': self.resource_id,
'name': self.name,
'rsrc_prop_data_id':
self._create_or_replace_rsrc_prop_data(),
'needed_by': [],
'requires': sorted(self.requires, reverse=True),
'replaces': self.replaces,
'replaced_by': self.replaced_by,
'current_template_id': self.current_template_id,
'root_stack_id': self.root_stack_id,
'updated_at': self.updated_time,
'properties_data': None}
if set_metadata:
metadata = self.t.metadata()
rs['rsrc_metadata'] = metadata
self._rsrc_metadata = metadata
if self.id is not None:
if (lock == self.LOCK_NONE or
(lock in {self.LOCK_ACQUIRE, self.LOCK_RELEASE} and
self._calling_engine_id is None)):
resource_objects.Resource.update_by_id(
self.context, self.id, rs)
if lock != self.LOCK_NONE:
LOG.error('No calling_engine_id in store() %s',
six.text_type(rs))
else:
self._store_with_lock(rs, lock)
else:
new_rs = resource_objects.Resource.create(self.context, rs)
self.id = new_rs.id
self.uuid = new_rs.uuid
self.created_time = new_rs.created_at
def _store_with_lock(self, rs, lock):
if lock == self.LOCK_ACQUIRE:
rs['engine_id'] = self._calling_engine_id
expected_engine_id = None
elif lock == self.LOCK_RESPECT:
expected_engine_id = None
elif lock == self.LOCK_RELEASE:
expected_engine_id = self._calling_engine_id
rs['engine_id'] = None
else:
assert False, "Invalid lock action: %s" % lock
if resource_objects.Resource.select_and_update_by_id(
self.context, self.id, rs, expected_engine_id,
self._atomic_key):
self._incr_atomic_key(self._atomic_key)
else:
LOG.info('Resource %s is locked or does not exist',
six.text_type(self))
LOG.debug('Resource id:%(resource_id)s locked or does not exist. '
'Expected atomic_key:%(atomic_key)s, '
'accessing from engine_id:%(engine_id)s',
{'resource_id': self.id,
'atomic_key': self._atomic_key,
'engine_id': self._calling_engine_id})
raise exception.UpdateInProgress(self.name)
def _add_event(self, action, status, reason):
"""Add a state change event to the database."""
physical_res_id = self.resource_id or self.physical_resource_name()
ev = event.Event(self.context, self.stack, action, status, reason,
physical_res_id, self._rsrc_prop_data_id,
self._stored_properties_data, self.name, self.type())
ev.store()
self.stack.dispatch_event(ev)
@contextlib.contextmanager
def lock(self, engine_id):
self._calling_engine_id = engine_id
try:
if engine_id is not None:
self._store_with_lock({}, self.LOCK_ACQUIRE)
yield
except exception.UpdateInProgress:
raise
except BaseException:
with excutils.save_and_reraise_exception():
if engine_id is not None:
self._store_with_lock({}, self.LOCK_RELEASE)
else:
if engine_id is not None:
self._store_with_lock({}, self.LOCK_RELEASE)
def _resolve_any_attribute(self, attr):
"""Method for resolving any attribute, including base attributes.
This method uses basic _resolve_attribute method for resolving
plugin-specific attributes. Base attributes will be resolved with
corresponding method, which should be defined in each resource
class.
:param attr: attribute name, which will be resolved
:returns: method of resource class, which resolve base attribute
"""
if attr in self.base_attributes_schema:
# check resource_id, because usually it is required for getting
# information about resource
if self.resource_id is not None:
with self._default_client_plugin().ignore_not_found:
return getattr(self, '_{0}_resource'.format(attr))()
else:
with self._default_client_plugin().ignore_not_found:
return self._resolve_attribute(attr)
return None
def _show_resource(self):
"""Default implementation; should be overridden by resources.
:returns: the map of resource information or None
"""
if self.entity and self.default_client_name is not None:
try:
obj = getattr(self.client(), self.entity)
resource = obj.get(self.resource_id)
if isinstance(resource, dict):
return resource
else:
return resource.to_dict()
except AttributeError as ex:
LOG.warning("Resolving 'show' attribute has failed : %s",
ex)
return None
def get_live_resource_data(self):
"""Default implementation; can be overridden by resources.
Get resource data and handle it with exceptions.
"""
try:
resource_data = self._show_resource()
except Exception as ex:
if self._default_client_plugin().is_not_found(ex):
raise exception.EntityNotFound(
entity='Resource', name=self.name)
raise
return resource_data
def parse_live_resource_data(self, resource_properties, resource_data):
"""Default implementation; can be overridden by resources.
Parse resource data for using it in updating properties with live
state.
:param resource_properties: properties of stored resource plugin.
:param resource_data: data from current live state of a resource.
"""
resource_result = {}
for key in self._update_allowed_properties:
if key in resource_data:
if key == 'name' and resource_properties.get(key) is None:
# We use `physical_resource_name` for name property in some
# resources when name not provided during create, so we
# shouldn't add name in resource_data if it's None in
# property (might just the cases that we using
# `physical_resource_name`).
continue
resource_result[key] = resource_data.get(key)
return resource_result
def get_live_state(self, resource_properties):
"""Default implementation; should be overridden by resources.
:param resource_properties: resource's object of Properties class.
:returns: dict of resource's real state of properties.
"""
resource_data = self.get_live_resource_data()
if resource_data is None:
return {}
return self.parse_live_resource_data(resource_properties,
resource_data)
def _update_properties_with_live_state(self, resource_properties,
live_properties):
"""Update resource properties data with live state properties.
Note, that live_properties can contains None values, so there's next
situation: property equals to some value, but live state has no such
property, i.e. property equals to None, so during update property
should be updated with None.
"""
for key in resource_properties:
if key in live_properties:
if resource_properties.get(key) != live_properties.get(key):
resource_properties.data.update(
{key: live_properties.get(key)})
def _resolve_attribute(self, name):
"""Default implementation of resolving resource's attributes.
Should be overridden by resources, that expose attributes.
:param name: The attribute to resolve
:returns: the resource attribute named key
"""
# By default, no attributes resolve
pass
def regenerate_info_schema(self, definition):
"""Default implementation; should be overridden by resources.
Should be overridden by resources that would require schema refresh
during update, ex. TemplateResource.
:definition: Resource Definition
"""
# By default, do not regenerate
pass
def state_reset(self):
"""Reset state to (INIT, COMPLETE)."""
self.action = self.INIT
self.status = self.COMPLETE
def state_set(self, action, status, reason="state changed",
lock=LOCK_NONE):
if action not in self.ACTIONS:
raise ValueError(_("Invalid action %s") % action)
if status not in self.STATUSES:
raise ValueError(_("Invalid status %s") % status)
old_state = (self.action, self.status)
new_state = (action, status)
set_metadata = self.action == self.INIT
self.action = action
self.status = status
self.status_reason = reason
self.store(set_metadata, lock=lock)
if new_state != old_state:
self._add_event(action, status, reason)
if status != self.COMPLETE:
self.clear_stored_attributes()
@property
def state(self):
"""Returns state, tuple of action, status."""
return (self.action, self.status)
def store_attributes(self):
assert self.id is not None
if self.status != self.COMPLETE or self.action in (self.INIT,
self.DELETE):
return
if not self.attributes.has_new_cached_attrs():
return
try:
attr_data_id = resource_objects.Resource.store_attributes(
self.context, self.id, self._atomic_key,
self.attributes.cached_attrs, self._attr_data_id)
if attr_data_id is not None:
self._incr_atomic_key(self._atomic_key)
self._attr_data_id = attr_data_id
except Exception as ex:
LOG.error('store_attributes rsrc %(name)s %(id)s DB error %(ex)s',
{'name': self.name, 'id': self.id, 'ex': ex})
def clear_stored_attributes(self):
if self._attr_data_id:
resource_objects.Resource.attr_data_delete(
self.context, self.id, self._attr_data_id)
self.attributes.reset_resolved_values()
def get_reference_id(self):
"""Default implementation for function get_resource.
This may be overridden by resource plugins to add extra
logic specific to the resource implementation.
"""
if self.resource_id is not None:
return six.text_type(self.resource_id)
else:
return six.text_type(self.name)
def FnGetRefId(self):
"""For the intrinsic function Ref.
:results: the id or name of the resource.
"""
return self.get_reference_id()
def physical_resource_name_or_FnGetRefId(self):
res_name = self.physical_resource_name()
if res_name is not None:
return six.text_type(res_name)
else:
return Resource.get_reference_id(self)
def get_attribute(self, key, *path):
"""Default implementation for function get_attr and Fn::GetAtt.
This may be overridden by resource plugins to add extra
logic specific to the resource implementation.
"""
try:
attribute = self.attributes[key]
except KeyError:
raise exception.InvalidTemplateAttribute(resource=self.name,
key=key)
return attributes.select_from_attribute(attribute, path)
def FnGetAtt(self, key, *path):
"""For the intrinsic function Fn::GetAtt.
:param key: the attribute key.
:param path: a list of path components to select from the attribute.
:returns: the attribute value.
"""
cache_custom = ((self.attributes.get_cache_mode(key) !=
attributes.Schema.CACHE_NONE) and
(type(self).get_attribute != Resource.get_attribute))
if cache_custom:
if path:
full_key = sync_point.str_pack_tuple((key,) + path)
else:
full_key = key
if full_key in self.attributes.cached_attrs:
return self.attributes.cached_attrs[full_key]
attr_val = self.get_attribute(key, *path)
if cache_custom:
self.attributes.set_cached_attr(full_key, attr_val)
return attr_val
def _signal_check_action(self):
if self.action in self.no_signal_actions:
self._add_event(self.action, self.status,
'Cannot signal resource during %s' % self.action)
msg = _('Signal resource during %s') % self.action
raise exception.NotSupported(feature=msg)
def _signal_check_hook(self, details):
if details and 'unset_hook' in details:
hook = details['unset_hook']
if not environment.valid_hook_type(hook):
msg = (_('Invalid hook type "%(hook)s" for %(resource)s') %
{'hook': hook, 'resource': six.text_type(self)})
raise exception.InvalidBreakPointHook(message=msg)
if not self.has_hook(hook):
msg = (_('The "%(hook)s" hook is not defined '
'on %(resource)s') %
{'hook': hook, 'resource': six.text_type(self)})
raise exception.InvalidBreakPointHook(message=msg)
def _unset_hook(self, details):
# Clear the hook without interfering with resources'
# `handle_signal` callbacks:
hook = details['unset_hook']
self.clear_hook(hook)
LOG.info('Clearing %(hook)s hook on %(resource)s',
{'hook': hook, 'resource': six.text_type(self)})
self._add_event(self.action, self.status,
"Hook %s is cleared" % hook)
def _handle_signal(self, details):
if not callable(getattr(self, 'handle_signal', None)):
raise exception.ResourceActionNotSupported(action='signal')
def get_string_details():
if details is None:
return 'No signal details provided'
if isinstance(details, six.string_types):
return details
if isinstance(details, dict):
if all(k in details for k in ('previous', 'current',
'reason')):
# this is from Ceilometer.
auto = '%(previous)s to %(current)s (%(reason)s)' % details
return 'alarm state changed from %s' % auto
return 'Unknown'
try:
signal_result = self.handle_signal(details)
if signal_result:
reason_string = "Signal: %s" % signal_result
else:
reason_string = get_string_details()
self._add_event('SIGNAL', self.status, reason_string)
except NoActionRequired:
<