def test_resolve(self):
sot = schema.Boolean()
res = sot.resolve(True)
self.assertTrue(res)
res = sot.resolve(False)
self.assertFalse(res)
res = sot.resolve('Yes')
self.assertTrue(res)
def test_validate(self):
sot = schema.Boolean()
res = sot.validate(True)
self.assertIsNone(res)
res = sot.validate('No')
self.assertIsNone(res)
ex = self.assertRaises(exc.ESchema, sot.validate, 'bogus')
self.assertEqual("The value 'bogus' is not a valid Boolean",
six.text_type(ex))
def test_to_schema_type(self):
sot = schema.Boolean('desc')
res = sot.to_schema_type(True)
self.assertTrue(res)
res = sot.to_schema_type('true')
self.assertTrue(res)
res = sot.to_schema_type('trUE')
self.assertTrue(res)
res = sot.to_schema_type('False')
self.assertFalse(res)
res = sot.to_schema_type('FALSE')
self.assertFalse(res)
ex = self.assertRaises(exc.ESchema, sot.to_schema_type, 'bogus')
self.assertEqual("The value 'bogus' is not a valid Boolean",
six.text_type(ex))
class ServerProfile(base.KubeBaseProfile):
"""Profile for an kubernetes master server."""
VERSIONS = {'1.0': [{'status': consts.EXPERIMENTAL, 'since': '2017.10'}]}
KEYS = (
CONTEXT,
FLAVOR,
IMAGE,
KEY_NAME,
PUBLIC_NETWORK,
BLOCK_DEVICE_MAPPING_V2,
) = (
'context',
'flavor',
'image',
'key_name',
'public_network',
'block_device_mapping_v2',
)
INTERNAL_KEYS = (
KUBEADM_TOKEN,
KUBE_MASTER_IP,
SECURITY_GROUP,
PRIVATE_NETWORK,
PRIVATE_SUBNET,
PRIVATE_ROUTER,
KUBE_MASTER_FLOATINGIP,
KUBE_MASTER_FLOATINGIP_ID,
SCALE_OUT_RECV_ID,
SCALE_OUT_URL,
) = (
'kubeadm_token',
'kube_master_ip',
'security_group',
'private_network',
'private_subnet',
'private_router',
'kube_master_floatingip',
'kube_master_floatingip_id',
'scale_out_recv_id',
'scale_out_url',
)
NETWORK_KEYS = (
PORT,
FIXED_IP,
NETWORK,
PORT_SECURITY_GROUPS,
FLOATING_NETWORK,
FLOATING_IP,
) = (
'port',
'fixed_ip',
'network',
'security_groups',
'floating_network',
'floating_ip',
)
BDM2_KEYS = (
BDM2_UUID,
BDM2_SOURCE_TYPE,
BDM2_DESTINATION_TYPE,
BDM2_DISK_BUS,
BDM2_DEVICE_NAME,
BDM2_VOLUME_SIZE,
BDM2_GUEST_FORMAT,
BDM2_BOOT_INDEX,
BDM2_DEVICE_TYPE,
BDM2_DELETE_ON_TERMINATION,
) = (
'uuid',
'source_type',
'destination_type',
'disk_bus',
'device_name',
'volume_size',
'guest_format',
'boot_index',
'device_type',
'delete_on_termination',
)
properties_schema = {
CONTEXT:
schema.Map(_('Customized security context for operating servers.'), ),
FLAVOR:
schema.String(
_('ID of flavor used for the server.'),
required=True,
updatable=True,
),
IMAGE:
schema.String(
# IMAGE is not required, because there could be BDM or BDMv2
# support and the corresponding settings effective
_('ID of image to be used for the new server.'),
updatable=True,
),
KEY_NAME:
schema.String(_('Name of Nova keypair to be injected to server.'), ),
PUBLIC_NETWORK:
schema.String(
_('Public network for kubernetes.'),
required=True,
),
BLOCK_DEVICE_MAPPING_V2:
schema.List(
_('A list specifying the properties of block devices to be used '
'for this server.'),
schema=schema.Map(
_('A map specifying the properties of a block device to be '
'used by the server.'),
schema={
BDM2_UUID:
schema.String(
_('ID of the source image, snapshot or volume'), ),
BDM2_SOURCE_TYPE:
schema.String(
_("Volume source type, must be one of 'image', "
"'snapshot', 'volume' or 'blank'"),
required=True,
),
BDM2_DESTINATION_TYPE:
schema.String(
_("Volume destination type, must be 'volume' or "
"'local'"),
required=True,
),
BDM2_DISK_BUS:
schema.String(_('Bus of the device.'), ),
BDM2_DEVICE_NAME:
schema.String(
_('Name of the device(e.g. vda, xda, ....).'), ),
BDM2_VOLUME_SIZE:
schema.Integer(
_('Size of the block device in MB(for swap) and '
'in GB(for other formats)'),
required=True,
),
BDM2_GUEST_FORMAT:
schema.String(
_('Specifies the disk file system format(e.g. swap, '
'ephemeral, ...).'), ),
BDM2_BOOT_INDEX:
schema.Integer(_('Define the boot order of the device'), ),
BDM2_DEVICE_TYPE:
schema.String(
_('Type of the device(e.g. disk, cdrom, ...).'), ),
BDM2_DELETE_ON_TERMINATION:
schema.Boolean(
_('Whether to delete the volume when the server '
'stops.'), ),
}),
),
}
def __init__(self, type_name, name, **kwargs):
super(ServerProfile, self).__init__(type_name, name, **kwargs)
self.server_id = None
def do_cluster_create(self, obj):
self._generate_kubeadm_token(obj)
self._create_security_group(obj)
self._create_network(obj)
def do_cluster_delete(self, obj):
if obj.dependents and 'kube-node' in obj.dependents:
msg = ("Cluster %s delete failed, "
"Node clusters %s must be deleted first." %
(obj.id, obj.dependents['kube-node']))
raise exc.EResourceDeletion(type='kubernetes.master',
id=obj.id,
message=msg)
self._delete_network(obj)
self._delete_security_group(obj)
def do_create(self, obj):
"""Create a server for the node object.
:param obj: The node object for which a server will be created.
"""
kwargs = {}
for key in self.KEYS:
if self.properties[key] is not None:
kwargs[key] = self.properties[key]
image_ident = self.properties[self.IMAGE]
if image_ident is not None:
image = self._validate_image(obj, image_ident, 'create')
kwargs.pop(self.IMAGE)
kwargs['imageRef'] = image.id
flavor_ident = self.properties[self.FLAVOR]
flavor = self._validate_flavor(obj, flavor_ident, 'create')
kwargs.pop(self.FLAVOR)
kwargs['flavorRef'] = flavor.id
keypair_name = self.properties[self.KEY_NAME]
if keypair_name:
keypair = self._validate_keypair(obj, keypair_name, 'create')
kwargs['key_name'] = keypair.name
kwargs['name'] = obj.name
metadata = self._build_metadata(obj, {})
kwargs['metadata'] = metadata
jj_vars = {}
cluster_data = self._get_cluster_data(obj)
kwargs['networks'] = [{'uuid': cluster_data[self.PRIVATE_NETWORK]}]
# Get user_data parameters from metadata
jj_vars['KUBETOKEN'] = cluster_data[self.KUBEADM_TOKEN]
jj_vars['MASTER_FLOATINGIP'] = cluster_data[
self.KUBE_MASTER_FLOATINGIP]
block_device_mapping_v2 = self.properties[self.BLOCK_DEVICE_MAPPING_V2]
if block_device_mapping_v2 is not None:
kwargs['block_device_mapping_v2'] = self._resolve_bdm(
obj, block_device_mapping_v2, 'create')
# user_data = self.properties[self.USER_DATA]
user_data = base.loadScript('./scripts/master.sh')
if user_data is not None:
# Use jinja2 to replace variables defined in user_data
try:
jj_t = jinja2.Template(user_data)
user_data = jj_t.render(**jj_vars)
except (jinja2.exceptions.UndefinedError, ValueError) as ex:
# TODO(anyone) Handle jinja2 error
pass
ud = encodeutils.safe_encode(user_data)
kwargs['user_data'] = encodeutils.safe_decode(base64.b64encode(ud))
sgid = self._get_security_group(obj)
kwargs['security_groups'] = [{'name': sgid}]
server = None
resource_id = None
try:
server = self.compute(obj).server_create(**kwargs)
self.compute(obj).wait_for_server(server.id)
server = self.compute(obj).server_get(server.id)
self._update_master_ip(obj, server.addresses[''][0]['addr'])
self._associate_floatingip(obj, server)
LOG.info("Created master node: %s" % server.id)
return server.id
except exc.InternalError as ex:
if server and server.id:
resource_id = server.id
raise exc.EResourceCreation(type='server',
message=six.text_type(ex),
resource_id=resource_id)
def do_delete(self, obj, **params):
"""Delete the physical resource associated with the specified node.
:param obj: The node object to operate on.
:param kwargs params: Optional keyword arguments for the delete
operation.
:returns: This operation always return True unless exception is
caught.
:raises: `EResourceDeletion` if interaction with compute service fails.
"""
if not obj.physical_id:
return True
server_id = obj.physical_id
ignore_missing = params.get('ignore_missing', True)
internal_ports = obj.data.get('internal_ports', [])
force = params.get('force', False)
try:
self._disassociate_floatingip(obj, server_id)
driver = self.compute(obj)
if force:
driver.server_force_delete(server_id, ignore_missing)
else:
driver.server_delete(server_id, ignore_missing)
driver.wait_for_server_delete(server_id)
if internal_ports:
ex = self._delete_ports(obj, internal_ports)
if ex:
raise ex
return True
except exc.InternalError as ex:
raise exc.EResourceDeletion(type='server',
id=server_id,
message=six.text_type(ex))
class StackProfile(base.Profile):
'''Profile for an OpenStack Heat stack.
When this profile is used, the whole cluster is a collection of Heat
stacks.
'''
KEYS = (
CONTEXT, TEMPLATE, TEMPLATE_URL, PARAMETERS,
FILES, TIMEOUT, DISABLE_ROLLBACK, ENVIRONMENT,
) = (
'context', 'template', 'template_url', 'parameters',
'files', 'timeout', 'disable_rollback', 'environment',
)
properties_schema = {
CONTEXT: schema.Map(
_('A dictionary for specifying the customized context for '
'stack operations'),
default={},
),
TEMPLATE: schema.Map(
_('Heat stack template.'),
default={},
updatable=True,
),
TEMPLATE_URL: schema.String(
_('Heat stack template url.'),
default='',
updatable=True,
),
PARAMETERS: schema.Map(
_('Parameters to be passed to Heat for stack operations.'),
default={},
updatable=True,
),
FILES: schema.Map(
_('Contents of files referenced by the template, if any.'),
default={},
updatable=True,
),
TIMEOUT: schema.Integer(
_('A integer that specifies the number of minutes that a '
'stack operation times out.'),
updatable=True,
),
DISABLE_ROLLBACK: schema.Boolean(
_('A boolean specifying whether a stack operation can be '
'rolled back.'),
default=True,
updatable=True,
),
ENVIRONMENT: schema.Map(
_('A map that specifies the environment used for stack '
'operations.'),
default={},
updatable=True,
)
}
OP_NAMES = (
OP_ABANDON,
) = (
'abandon',
)
OPERATIONS = {
OP_ABANDON: schema.Map(
_('Abandon a heat stack node.'),
)
}
def __init__(self, type_name, name, **kwargs):
super(StackProfile, self).__init__(type_name, name, **kwargs)
self.stack_id = None
def validate(self, validate_props=False):
'''Validate the schema and the data provided.'''
# general validation
self.spec_data.validate()
self.properties.validate()
# validate template
template = self.properties[self.TEMPLATE]
template_url = self.properties[self.TEMPLATE_URL]
if not template and not template_url:
msg = _("Both template and template_url are not specified "
"for profile '%s'.") % self.name
raise exc.InvalidSpec(message=msg)
if validate_props:
self.do_validate(obj=self)
def do_validate(self, obj):
"""Validate the stack template used by a node.
:param obj: Node object to operate.
:returns: True if validation succeeds.
:raises: `InvalidSpec` exception is raised if template is invalid.
"""
kwargs = {
'stack_name': utils.random_name(),
'template': self.properties[self.TEMPLATE],
'template_url': self.properties[self.TEMPLATE_URL],
'parameters': self.properties[self.PARAMETERS],
'files': self.properties[self.FILES],
'environment': self.properties[self.ENVIRONMENT],
'preview': True,
}
try:
self.orchestration(obj).stack_create(**kwargs)
except exc.InternalError as ex:
msg = _('Failed in validating template: %s') % six.text_type(ex)
raise exc.InvalidSpec(message=msg)
return True
def do_create(self, obj):
"""Create a heat stack using the given node object.
:param obj: The node object to operate on.
:returns: The UUID of the heat stack created.
"""
kwargs = {
'stack_name': obj.name + '-' + utils.random_name(8),
'template': self.properties[self.TEMPLATE],
'template_url': self.properties[self.TEMPLATE_URL],
'timeout_mins': self.properties[self.TIMEOUT],
'disable_rollback': self.properties[self.DISABLE_ROLLBACK],
'parameters': self.properties[self.PARAMETERS],
'files': self.properties[self.FILES],
'environment': self.properties[self.ENVIRONMENT],
}
try:
stack = self.orchestration(obj).stack_create(**kwargs)
# Timeout = None means we will use the 'default_action_timeout'
# It can be overridden by the TIMEOUT profile propertie
timeout = None
if self.properties[self.TIMEOUT]:
timeout = self.properties[self.TIMEOUT] * 60
self.orchestration(obj).wait_for_stack(stack.id, 'CREATE_COMPLETE',
timeout=timeout)
return stack.id
except exc.InternalError as ex:
raise exc.EResourceCreation(type='stack', message=ex.message)
def do_delete(self, obj, **params):
"""Delete the physical stack behind the node object.
:param obj: The node object to operate on.
:param kwargs params: Optional keyword arguments for the delete
operation.
:returns: This operation always return True unless exception is
caught.
:raises: `EResourceDeletion` if interaction with heat fails.
"""
stack_id = obj.physical_id
ignore_missing = params.get('ignore_missing', True)
try:
self.orchestration(obj).stack_delete(stack_id, ignore_missing)
self.orchestration(obj).wait_for_stack_delete(stack_id)
except exc.InternalError as ex:
raise exc.EResourceDeletion(type='stack', id=stack_id,
message=six.text_type(ex))
return True
def do_update(self, obj, new_profile, **params):
"""Perform update on object.
:param obj: the node object to operate on
:param new_profile: the new profile used for updating
:param params: other parameters for the update request.
:returns: A boolean indicating whether the operation is successful.
"""
self.stack_id = obj.physical_id
if not self.stack_id:
return False
if not self.validate_for_update(new_profile):
return False
fields = {}
new_template = new_profile.properties[new_profile.TEMPLATE]
if new_template != self.properties[self.TEMPLATE]:
fields['template'] = new_template
new_params = new_profile.properties[new_profile.PARAMETERS]
if new_params != self.properties[self.PARAMETERS]:
fields['parameters'] = new_params
new_timeout = new_profile.properties[new_profile.TIMEOUT]
if new_timeout != self.properties[self.TIMEOUT]:
fields['timeout_mins'] = new_timeout
new_dr = new_profile.properties[new_profile.DISABLE_ROLLBACK]
if new_dr != self.properties[self.DISABLE_ROLLBACK]:
fields['disable_rollback'] = new_dr
new_files = new_profile.properties[new_profile.FILES]
if new_files != self.properties[self.FILES]:
fields['files'] = new_files
new_environment = new_profile.properties[new_profile.ENVIRONMENT]
if new_environment != self.properties[self.ENVIRONMENT]:
fields['environment'] = new_environment
if not fields:
return True
try:
hc = self.orchestration(obj)
# Timeout = None means we will use the 'default_action_timeout'
# It can be overridden by the TIMEOUT profile propertie
timeout = None
if self.properties[self.TIMEOUT]:
timeout = self.properties[self.TIMEOUT] * 60
hc.stack_update(self.stack_id, **fields)
hc.wait_for_stack(self.stack_id, 'UPDATE_COMPLETE',
timeout=timeout)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='stack', id=self.stack_id,
message=ex.message)
return True
def do_check(self, obj):
"""Check stack status.
:param obj: Node object to operate.
:returns: True if check succeeded, or False otherwise.
"""
stack_id = obj.physical_id
if stack_id is None:
return False
hc = self.orchestration(obj)
try:
# Timeout = None means we will use the 'default_action_timeout'
# It can be overridden by the TIMEOUT profile propertie
timeout = None
if self.properties[self.TIMEOUT]:
timeout = self.properties[self.TIMEOUT] * 60
hc.stack_check(stack_id)
hc.wait_for_stack(stack_id, 'CHECK_COMPLETE', timeout=timeout)
except exc.InternalError as ex:
LOG.error(_LE('Failed in checking stack: %s.'), ex)
return False
return True
def do_get_details(self, obj):
if not obj.physical_id:
return {}
try:
stack = self.orchestration(obj).stack_get(obj.physical_id)
return stack.to_dict()
except exc.InternalError as ex:
return {
'Error': {
'code': ex.code,
'message': six.text_type(ex)
}
}
def handle_abandon(self, obj, **options):
"""Handler for abandoning a heat stack node."""
pass
class HealthPolicy(base.Policy):
"""Policy for health management of a cluster."""
VERSION = '1.1'
VERSIONS = {
'1.0': [
{
'status': consts.EXPERIMENTAL,
'since': '2017.02'
},
{
'status': consts.SUPPORTED,
'since': '2018.06'
},
],
'1.1': [{
'status': consts.SUPPORTED,
'since': '2018.09'
}],
}
PRIORITY = 600
TARGET = [
('BEFORE', consts.CLUSTER_RECOVER),
('BEFORE', consts.CLUSTER_DEL_NODES),
('BEFORE', consts.CLUSTER_SCALE_IN),
('BEFORE', consts.CLUSTER_RESIZE),
('BEFORE', consts.NODE_DELETE),
('AFTER', consts.CLUSTER_DEL_NODES),
('AFTER', consts.CLUSTER_SCALE_IN),
('AFTER', consts.CLUSTER_RESIZE),
('AFTER', consts.NODE_DELETE),
]
# Should be ANY if profile provides health check support?
PROFILE_TYPE = [
'os.nova.server',
'os.heat.stack',
]
KEYS = (DETECTION, RECOVERY) = ('detection', 'recovery')
_DETECTION_KEYS = (DETECTION_MODES, DETECTION_TYPE, DETECTION_OPTIONS,
DETECTION_INTERVAL, NODE_UPDATE_TIMEOUT,
RECOVERY_CONDITIONAL) = ('detection_modes', 'type',
'options', 'interval',
'node_update_timeout',
'recovery_conditional')
_DETECTION_OPTIONS = (
POLL_URL,
POLL_URL_SSL_VERIFY,
POLL_URL_CONN_ERROR_AS_UNHEALTHY,
POLL_URL_HEALTHY_RESPONSE,
POLL_URL_RETRY_LIMIT,
POLL_URL_RETRY_INTERVAL,
) = ('poll_url', 'poll_url_ssl_verify', 'poll_url_conn_error_as_unhealthy',
'poll_url_healthy_response', 'poll_url_retry_limit',
'poll_url_retry_interval')
_RECOVERY_KEYS = (
RECOVERY_ACTIONS,
RECOVERY_FENCING,
RECOVERY_DELETE_TIMEOUT,
RECOVERY_FORCE_RECREATE,
) = (
'actions',
'fencing',
'node_delete_timeout',
'node_force_recreate',
)
FENCING_OPTION_VALUES = (
COMPUTE,
# STORAGE, NETWORK,
) = (
'COMPUTE',
# 'STORAGE', 'NETWORK'
)
ACTION_KEYS = (
ACTION_NAME,
ACTION_PARAMS,
) = (
'name',
'params',
)
properties_schema = {
DETECTION:
schema.Map(
_('Policy aspect for node failure detection.'),
schema={
DETECTION_INTERVAL:
schema.Integer(
_("Number of seconds between pollings. Only "
"required when type is 'NODE_STATUS_POLLING' or "
"'NODE_STATUS_POLL_URL'."),
default=60,
),
NODE_UPDATE_TIMEOUT:
schema.Integer(
_("Number of seconds since last node update to "
"wait before checking node health."),
default=300,
),
RECOVERY_CONDITIONAL:
schema.String(
_("The conditional that determines when recovery should be"
" performed in case multiple detection modes are "
"specified. 'ALL_FAILED' means that all "
"detection modes have to return failed health checks "
"before a node is recovered. 'ANY_FAILED'"
" means that a failed health check with a single "
"detection mode triggers a node recovery."),
constraints=[
constraints.AllowedValues(consts.RECOVERY_CONDITIONAL),
],
default=consts.ANY_FAILED,
required=False,
),
DETECTION_MODES:
schema.List(
_('List of node failure detection modes.'),
schema=schema.Map(
_('Node failure detection mode to try'),
schema={
DETECTION_TYPE:
schema.String(
_('Type of node failure detection.'),
constraints=[
constraints.AllowedValues(
consts.DETECTION_TYPES),
],
required=True,
),
DETECTION_OPTIONS:
schema.Map(schema={
POLL_URL:
schema.String(
_("URL to poll for node status. See "
"documentation for valid expansion "
"parameters. Only required "
"when type is "
"'NODE_STATUS_POLL_URL'."),
default='',
),
POLL_URL_SSL_VERIFY:
schema.Boolean(
_("Whether to verify SSL when calling "
"URL to poll for node status. Only "
"required when type is "
"'NODE_STATUS_POLL_URL'."),
default=True,
),
POLL_URL_CONN_ERROR_AS_UNHEALTHY:
schema.Boolean(
_("Whether to treat URL connection "
"errors as an indication of an "
"unhealthy node. Only required "
"when type is "
"'NODE_STATUS_POLL_URL'."),
default=True,
),
POLL_URL_HEALTHY_RESPONSE:
schema.String(
_("String pattern in the poll URL "
"response body that indicates a "
"healthy node. Required when type "
"is 'NODE_STATUS_POLL_URL'."),
default='',
),
POLL_URL_RETRY_LIMIT:
schema.Integer(
_("Number of times to retry URL "
"polling when its return body is "
"missing POLL_URL_HEALTHY_RESPONSE "
"string before a node is considered "
"down. Required when type is "
"'NODE_STATUS_POLL_URL'."),
default=3,
),
POLL_URL_RETRY_INTERVAL:
schema.Integer(
_("Number of seconds between URL "
"polling retries before a node is "
"considered down. Required when "
"type is 'NODE_STATUS_POLL_URL'."),
default=3,
),
},
default={}),
}))
},
required=True,
),
RECOVERY:
schema.Map(
_('Policy aspect for node failure recovery.'),
schema={
RECOVERY_ACTIONS:
schema.List(_('List of actions to try for node recovery.'),
schema=schema.Map(
_('Action to try for node recovery.'),
schema={
ACTION_NAME:
schema.String(
_("Name of action to execute."),
constraints=[
constraints.AllowedValues(
consts.RECOVERY_ACTIONS),
],
required=True),
ACTION_PARAMS:
schema.Map(_("Parameters for the action")),
})),
RECOVERY_FENCING:
schema.List(
_('List of services to be fenced.'),
schema=schema.String(
_('Service to be fenced.'),
constraints=[
constraints.AllowedValues(FENCING_OPTION_VALUES),
],
required=True,
),
),
RECOVERY_DELETE_TIMEOUT:
schema.Integer(
_("Number of seconds to wait for node deletion to "
"finish and start node creation for recreate "
"recovery option. Required when type is "
"'NODE_STATUS_POLL_URL and recovery action "
"is RECREATE'."),
default=20,
),
RECOVERY_FORCE_RECREATE:
schema.Boolean(
_("Whether to create node even if node deletion "
"failed. Required when type is "
"'NODE_STATUS_POLL_URL' and action recovery "
"action is RECREATE."),
default=False,
),
},
required=True,
),
}
def __init__(self, name, spec, **kwargs):
super(HealthPolicy, self).__init__(name, spec, **kwargs)
self.interval = self.properties[self.DETECTION].get(
self.DETECTION_INTERVAL, 60)
self.node_update_timeout = self.properties[self.DETECTION].get(
self.NODE_UPDATE_TIMEOUT, 300)
self.recovery_conditional = self.properties[self.DETECTION].get(
self.RECOVERY_CONDITIONAL, consts.ANY_FAILED)
DetectionMode = namedtuple('DetectionMode', [self.DETECTION_TYPE] +
list(self._DETECTION_OPTIONS))
self.detection_modes = []
raw_modes = self.properties[self.DETECTION][self.DETECTION_MODES]
for mode in raw_modes:
options = mode[self.DETECTION_OPTIONS]
self.detection_modes.append(
DetectionMode(
mode[self.DETECTION_TYPE], options.get(self.POLL_URL, ''),
options.get(self.POLL_URL_SSL_VERIFY, True),
options.get(self.POLL_URL_CONN_ERROR_AS_UNHEALTHY, True),
options.get(self.POLL_URL_HEALTHY_RESPONSE, ''),
options.get(self.POLL_URL_RETRY_LIMIT, ''),
options.get(self.POLL_URL_RETRY_INTERVAL, '')))
recover_settings = self.properties[self.RECOVERY]
self.recover_actions = recover_settings[self.RECOVERY_ACTIONS]
self.fencing_types = recover_settings[self.RECOVERY_FENCING]
self.node_delete_timeout = recover_settings.get(
self.RECOVERY_DELETE_TIMEOUT, None)
self.node_force_recreate = recover_settings.get(
self.RECOVERY_FORCE_RECREATE, False)
def validate(self, context, validate_props=False):
super(HealthPolicy, self).validate(context,
validate_props=validate_props)
if len(self.recover_actions) > 1:
message = _(
"Only one '%s' is supported for now.") % self.RECOVERY_ACTIONS
raise exc.ESchema(message=message)
if self.interval < cfg.CONF.health_check_interval_min:
message = _("Specified interval of %(interval)d seconds has to be "
"larger than health_check_interval_min of "
"%(min_interval)d seconds set in configuration.") % {
"interval": self.interval,
"min_interval": cfg.CONF.health_check_interval_min
}
raise exc.InvalidSpec(message=message)
# check valid detection types
polling_types = [
consts.NODE_STATUS_POLLING, consts.NODE_STATUS_POLL_URL
]
has_valid_polling_types = all(d.type in polling_types
for d in self.detection_modes)
has_valid_lifecycle_type = (len(self.detection_modes) == 1
and self.detection_modes[0].type
== consts.LIFECYCLE_EVENTS)
if not has_valid_polling_types and not has_valid_lifecycle_type:
message = ("Invalid detection modes in health policy: %s" %
', '.join([d.type for d in self.detection_modes]))
raise exc.InvalidSpec(message=message)
if len(self.detection_modes) != len(set(self.detection_modes)):
message = ("Duplicate detection modes are not allowed in "
"health policy: %s" %
', '.join([d.type for d in self.detection_modes]))
raise exc.InvalidSpec(message=message)
# TODO(Qiming): Add detection of duplicated action names when
# support to list of actions is implemented.
def attach(self, cluster, enabled=True):
""""Hook for policy attach.
Register the cluster for health management.
:param cluster: The cluster to which the policy is being attached to.
:param enabled: The attached cluster policy is enabled or disabled.
:return: A tuple comprising execution result and policy data.
"""
p_type = cluster.rt['profile'].type_name
action_names = [a['name'] for a in self.recover_actions]
if p_type != 'os.nova.server':
if consts.RECOVER_REBUILD in action_names:
err_msg = _("Recovery action REBUILD is only applicable to "
"os.nova.server clusters.")
return False, err_msg
if consts.RECOVER_REBOOT in action_names:
err_msg = _("Recovery action REBOOT is only applicable to "
"os.nova.server clusters.")
return False, err_msg
kwargs = {
'interval': self.interval,
'node_update_timeout': self.node_update_timeout,
'params': {
'recover_action': self.recover_actions,
'node_delete_timeout': self.node_delete_timeout,
'node_force_recreate': self.node_force_recreate,
'recovery_conditional': self.recovery_conditional,
},
'enabled': enabled
}
converted_detection_modes = [d._asdict() for d in self.detection_modes]
detection_mode = {'detection_modes': converted_detection_modes}
kwargs['params'].update(detection_mode)
health_manager.register(cluster.id, engine_id=None, **kwargs)
data = {
'interval': self.interval,
'node_update_timeout': self.node_update_timeout,
'recovery_conditional': self.recovery_conditional,
'node_delete_timeout': self.node_delete_timeout,
'node_force_recreate': self.node_force_recreate,
}
data.update(detection_mode)
return True, self._build_policy_data(data)
def detach(self, cluster):
"""Hook for policy detach.
Unregister the cluster for health management.
:param cluster: The target cluster.
:returns: A tuple comprising the execution result and reason.
"""
ret = health_manager.unregister(cluster.id)
if not ret:
LOG.warning(
'Unregistering health manager for cluster %s '
'timed out.', cluster.id)
return True, ''
def pre_op(self, cluster_id, action, **args):
"""Hook before action execution.
One of the task for this routine is to disable health policy if the
action is a request that will shrink the cluster. The reason is that
the policy may attempt to recover nodes that are to be deleted.
:param cluster_id: The ID of the target cluster.
:param action: The action to be examined.
:param kwargs args: Other keyword arguments to be checked.
:returns: Boolean indicating whether the checking passed.
"""
if action.action in (consts.CLUSTER_SCALE_IN, consts.CLUSTER_DEL_NODES,
consts.NODE_DELETE):
health_manager.disable(cluster_id)
return True
if action.action == consts.CLUSTER_RESIZE:
deletion = action.data.get('deletion', None)
if deletion:
health_manager.disable(cluster_id)
return True
cluster = action.entity
current = len(cluster.nodes)
res, reason = scaleutils.parse_resize_params(
action, cluster, current)
if res == base.CHECK_ERROR:
action.data['status'] = base.CHECK_ERROR
action.data['reason'] = reason
return False
if action.data.get('deletion', None):
health_manager.disable(cluster_id)
return True
pd = {
'recover_action': self.recover_actions,
'fencing': self.fencing_types,
}
action.data.update({'health': pd})
action.store(action.context)
return True
def post_op(self, cluster_id, action, **args):
"""Hook before action execution.
One of the task for this routine is to re-enable health policy if the
action is a request that will shrink the cluster thus the policy has
been temporarily disabled.
:param cluster_id: The ID of the target cluster.
:param action: The action to be examined.
:param kwargs args: Other keyword arguments to be checked.
:returns: Boolean indicating whether the checking passed.
"""
if action.action in (consts.CLUSTER_SCALE_IN, consts.CLUSTER_DEL_NODES,
consts.NODE_DELETE):
health_manager.enable(cluster_id)
return True
if action.action == consts.CLUSTER_RESIZE:
deletion = action.data.get('deletion', None)
if deletion:
health_manager.enable(cluster_id)
return True
cluster = action.entity
current = len(cluster.nodes)
res, reason = scaleutils.parse_resize_params(
action, cluster, current)
if res == base.CHECK_ERROR:
action.data['status'] = base.CHECK_ERROR
action.data['reason'] = reason
return False
if action.data.get('deletion', None):
health_manager.enable(cluster_id)
return True
return True
class AffinityPolicy(base.Policy):
"""Policy for placing members of a cluster based on server groups.
This policy is expected to be enforced before new member(s) added to an
existing cluster.
"""
VERSION = '1.0'
PRIORITY = 300
TARGET = [
('BEFORE', consts.CLUSTER_SCALE_OUT),
('BEFORE', consts.CLUSTER_RESIZE),
]
PROFILE_TYPE = [
'os.nova.server-1.0',
]
KEYS = (
SERVER_GROUP,
AVAILABILITY_ZONE,
ENABLE_DRS_EXTENSION,
) = (
'servergroup',
'availability_zone',
'enable_drs_extension',
)
_GROUP_KEYS = (
GROUP_NAME,
GROUP_POLICIES,
) = (
'name',
'policies',
)
_POLICIES_VALUES = (
AFFINITY,
ANTI_AFFINITY,
) = (
'affinity',
'anti-affinity',
)
properties_schema = {
SERVER_GROUP:
schema.Map(
_('Properties of the VM server group'),
schema={
GROUP_NAME:
schema.String(_('The name of the server group'), ),
GROUP_POLICIES:
schema.String(
_('The server group policies.'),
default=ANTI_AFFINITY,
constraints=[
constraints.AllowedValues(_POLICIES_VALUES),
],
),
},
),
AVAILABILITY_ZONE:
schema.String(
_('Name of the availability zone to place the nodes.'), ),
ENABLE_DRS_EXTENSION:
schema.Boolean(
_('Enable vSphere DRS extension.'),
default=False,
),
}
def __init__(self, name, spec, **kwargs):
super(AffinityPolicy, self).__init__(name, spec, **kwargs)
self.enable_drs = self.properties.get(self.ENABLE_DRS_EXTENSION)
self._novaclient = None
def nova(self, obj):
"""Construct nova client based on object.
:param obj: Object for which the client is created. It is expected to
be None when retrieving an existing client. When creating
a client, it conatins the user and project to be used.
"""
if self._novaclient is not None:
return self._novaclient
params = self._build_conn_params(obj)
self._novaclient = driver.SenlinDriver().compute(params)
return self._novaclient
def attach(self, cluster):
"""Routine to be invoked when policy is to be attached to a cluster.
:para cluster: The target cluster to attach to;
:returns: When the operation was successful, returns a tuple (True,
message); otherwise, return a tuple (False, error).
"""
res, data = super(AffinityPolicy, self).attach(cluster)
if res is False:
return False, data
data = {'inherited_group': False}
nc = self.nova(cluster)
group = self.properties.get(self.SERVER_GROUP)
# guess servergroup name
group_name = group.get(self.GROUP_NAME, None)
if group_name is None:
profile = cluster.rt['profile']
if 'scheduler_hints' in profile.spec:
hints = profile.spec['scheduler_hints']
group_name = hints.get('group', None)
if group_name:
try:
server_group = nc.find_server_group(group_name, True)
except exception.InternalError as ex:
msg = _("Failed in retrieving servergroup '%s'.") % group_name
LOG.exception(
_LE('%(msg)s: %(ex)s') % {
'msg': msg,
'ex': six.text_type(ex)
})
return False, msg
if server_group:
# Check if the policies match
policies = group.get(self.GROUP_POLICIES)
if policies and policies != server_group.policies[0]:
msg = _(
"Policies specified (%(specified)s) doesn't match "
"that of the existing servergroup (%(existing)s).") % {
'specified': policies,
'existing': server_group.policies[0]
}
return False, msg
data['servergroup_id'] = server_group.id
data['inherited_group'] = True
if not data['inherited_group']:
# create a random name if necessary
if not group_name:
group_name = 'server_group_%s' % utils.random_name()
try:
server_group = nc.create_server_group(
name=group_name, policies=[group.get(self.GROUP_POLICIES)])
except Exception as ex:
msg = _('Failed in creating servergroup.')
LOG.exception(
_LE('%(msg)s: %(ex)s') % {
'msg': msg,
'ex': six.text_type(ex)
})
return False, msg
data['servergroup_id'] = server_group.id
policy_data = self._build_policy_data(data)
return True, policy_data
def detach(self, cluster):
"""Routine to be called when the policy is detached from a cluster.
:param cluster: The cluster from which the policy is to be detached.
:returns: When the operation was successful, returns a tuple of
(True, data) where the data contains references to the
resources created; otherwise returns a tuple of (False,
error) where the err contains a error message.
"""
reason = _('Servergroup resource deletion succeeded.')
ctx = context.get_admin_context()
binding = cpo.ClusterPolicy.get(ctx, cluster.id, self.id)
if not binding or not binding.data:
return True, reason
policy_data = self._extract_policy_data(binding.data)
if not policy_data:
return True, reason
group_id = policy_data.get('servergroup_id', None)
inherited_group = policy_data.get('inherited_group', False)
if group_id and not inherited_group:
try:
self.nova(cluster).delete_server_group(group_id)
except Exception as ex:
msg = _('Failed in deleting servergroup.')
LOG.exception(
_LE('%(msg)s: %(ex)s') % {
'msg': msg,
'ex': six.text_type(ex)
})
return False, msg
return True, reason
def pre_op(self, cluster_id, action):
"""Routine to be called before an 'CLUSTER_SCALE_OUT' action.
For this particular policy, we take this chance to intelligently
select the most proper hypervisor/vsphere cluster to create nodes.
In order to realize the function, we need to create construct meta
to handle affinity/anti-affinity then update the profile with the
specific parameters at first
:param cluster_id: ID of the cluster on which the relevant action
is to be executed.
:param action: The action object that triggered this operation.
:returns: Nothing.
"""
zone_name = self.properties.get(self.AVAILABILITY_ZONE)
if not zone_name and self.enable_drs:
# we make a reasonable guess of the zone name for vSphere
# support because the zone name is required in that case.
zone_name = 'nova'
# we respect other policies decisions (if any) and fall back to the
# action inputs if no hints found.
pd = action.data.get('creation', None)
if pd is not None:
count = pd.get('count', 1)
elif action.action == consts.CLUSTER_SCALE_OUT:
count = action.inputs.get('count', 1)
else: # CLUSTER_RESIZE
db_cluster = co.Cluster.get(action.context, cluster_id)
su.parse_resize_params(action, db_cluster)
if 'creation' not in action.data:
return
count = action.data['creation']['count']
cp = cpo.ClusterPolicy.get(action.context, cluster_id, self.id)
policy_data = self._extract_policy_data(cp.data)
pd_entry = {'servergroup': policy_data['servergroup_id']}
# special handling for vSphere DRS case where we need to find out
# the name of the vSphere host which has DRS enabled.
if self.enable_drs:
cluster_obj = co.Cluster.get(action.context, cluster_id)
nc = self.nova(cluster_obj)
hypervisors = nc.hypervisor_list()
hv_id = ''
pattern = re.compile(r'.*drs*', re.I)
for hypervisor in hypervisors:
match = pattern.match(hypervisor.hypervisor_hostname)
if match:
hv_id = hypervisor.id
break
if not hv_id:
action.data['status'] = base.CHECK_ERROR
action.data['status_reason'] = _('No suitable vSphere host '
'is available.')
action.store(action.context)
return
hv_info = nc.hypervisor_get(hv_id)
hostname = hv_info['service']['host']
pd_entry['zone'] = ":".join([zone_name, hostname])
elif zone_name:
pd_entry['zone'] = zone_name
pd = {
'count': count,
'placements': [pd_entry] * count,
}
action.data.update({'placement': pd})
action.store(action.context)
return
class DeletionPolicy(base.Policy):
"""Policy for choosing victim node(s) from a cluster for deletion.
This policy is enforced when nodes are to be removed from a cluster.
It will yield an ordered list of candidates for deletion based on user
specified criteria.
"""
VERSION = '1.0'
VERSIONS = {
'1.0': [
{'status': consts.SUPPORTED, 'since': '2016.04'}
]
}
PRIORITY = 400
KEYS = (
CRITERIA, DESTROY_AFTER_DELETION, GRACE_PERIOD,
REDUCE_DESIRED_CAPACITY,
) = (
'criteria', 'destroy_after_deletion', 'grace_period',
'reduce_desired_capacity',
)
CRITERIA_VALUES = (
OLDEST_FIRST, OLDEST_PROFILE_FIRST, YOUNGEST_FIRST, RANDOM,
) = (
'OLDEST_FIRST', 'OLDEST_PROFILE_FIRST', 'YOUNGEST_FIRST', 'RANDOM',
)
TARGET = [
('BEFORE', consts.CLUSTER_SCALE_IN),
('BEFORE', consts.CLUSTER_DEL_NODES),
('BEFORE', consts.CLUSTER_RESIZE),
('BEFORE', consts.NODE_DELETE),
]
PROFILE_TYPE = [
'ANY'
]
properties_schema = {
CRITERIA: schema.String(
_('Criteria used in selecting candidates for deletion'),
default=RANDOM,
constraints=[
constraints.AllowedValues(CRITERIA_VALUES),
]
),
DESTROY_AFTER_DELETION: schema.Boolean(
_('Whether a node should be completely destroyed after '
'deletion. Default to True'),
default=True,
),
GRACE_PERIOD: schema.Integer(
_('Number of seconds before real deletion happens.'),
default=0,
),
REDUCE_DESIRED_CAPACITY: schema.Boolean(
_('Whether the desired capacity of the cluster should be '
'reduced along the deletion. Default to False.'),
default=True,
)
}
def __init__(self, name, spec, **kwargs):
super(DeletionPolicy, self).__init__(name, spec, **kwargs)
self.criteria = self.properties[self.CRITERIA]
self.grace_period = self.properties[self.GRACE_PERIOD]
self.destroy_after_deletion = self.properties[
self.DESTROY_AFTER_DELETION]
self.reduce_desired_capacity = self.properties[
self.REDUCE_DESIRED_CAPACITY]
def _victims_by_regions(self, cluster, regions):
victims = []
for region in sorted(regions.keys()):
count = regions[region]
nodes = cluster.nodes_by_region(region)
if self.criteria == self.RANDOM:
candidates = scaleutils.nodes_by_random(nodes, count)
elif self.criteria == self.OLDEST_PROFILE_FIRST:
candidates = scaleutils.nodes_by_profile_age(nodes, count)
elif self.criteria == self.OLDEST_FIRST:
candidates = scaleutils.nodes_by_age(nodes, count, True)
else:
candidates = scaleutils.nodes_by_age(nodes, count, False)
victims.extend(candidates)
return victims
def _victims_by_zones(self, cluster, zones):
victims = []
for zone in sorted(zones.keys()):
count = zones[zone]
nodes = cluster.nodes_by_zone(zone)
if self.criteria == self.RANDOM:
candidates = scaleutils.nodes_by_random(nodes, count)
elif self.criteria == self.OLDEST_PROFILE_FIRST:
candidates = scaleutils.nodes_by_profile_age(nodes, count)
elif self.criteria == self.OLDEST_FIRST:
candidates = scaleutils.nodes_by_age(nodes, count, True)
else:
candidates = scaleutils.nodes_by_age(nodes, count, False)
victims.extend(candidates)
return victims
def _update_action(self, action, victims):
pd = action.data.get('deletion', {})
pd['count'] = len(victims)
pd['candidates'] = victims
pd['destroy_after_deletion'] = self.destroy_after_deletion
pd['grace_period'] = self.grace_period
pd['reduce_desired_capacity'] = self.reduce_desired_capacity
action.data.update({
'status': base.CHECK_OK,
'reason': _('Candidates generated'),
'deletion': pd
})
action.store(action.context)
def pre_op(self, cluster_id, action):
"""Choose victims that can be deleted.
:param cluster_id: ID of the cluster to be handled.
:param action: The action object that triggered this policy.
"""
victims = action.inputs.get('candidates', [])
if len(victims) > 0:
self._update_action(action, victims)
return
if action.action == consts.NODE_DELETE:
self._update_action(action, [action.node.id])
return
db_cluster = None
regions = None
zones = None
deletion = action.data.get('deletion', {})
if deletion:
# there are policy decisions
count = deletion['count']
regions = deletion.get('regions', None)
zones = deletion.get('zones', None)
# No policy decision, check action itself: SCALE_IN
elif action.action == consts.CLUSTER_SCALE_IN:
count = action.inputs.get('count', 1)
# No policy decision, check action itself: RESIZE
else:
db_cluster = co.Cluster.get(action.context, cluster_id)
current = no.Node.count_by_cluster(action.context, cluster_id)
res, reason = scaleutils.parse_resize_params(action, db_cluster,
current)
if res == base.CHECK_ERROR:
action.data['status'] = base.CHECK_ERROR
action.data['reason'] = reason
LOG.error(reason)
return
if 'deletion' not in action.data:
return
count = action.data['deletion']['count']
cluster = cm.Cluster.load(action.context, dbcluster=db_cluster,
cluster_id=cluster_id)
# Cross-region
if regions:
victims = self._victims_by_regions(cluster, regions)
self._update_action(action, victims)
return
# Cross-AZ
if zones:
victims = self._victims_by_zones(cluster, zones)
self._update_action(action, victims)
return
if count > len(cluster.nodes):
count = len(cluster.nodes)
if self.criteria == self.RANDOM:
victims = scaleutils.nodes_by_random(cluster.nodes, count)
elif self.criteria == self.OLDEST_PROFILE_FIRST:
victims = scaleutils.nodes_by_profile_age(cluster.nodes, count)
elif self.criteria == self.OLDEST_FIRST:
victims = scaleutils.nodes_by_age(cluster.nodes, count, True)
else:
victims = scaleutils.nodes_by_age(cluster.nodes, count, False)
self._update_action(action, victims)
return
class LoadBalancingPolicy(base.Policy):
'''Policy for load balancing among members of a cluster.
This policy is expected to be enforced after the member list of a cluster
is changed. We need to reload the load-balancer specified (or internally
created) when these actions are performed.
'''
VERSION = '1.0'
TARGET = [
('AFTER', consts.CLUSTER_ADD_NODES),
('AFTER', consts.CLUSTER_DEL_NODES),
('AFTER', consts.CLUSTER_SCALE_OUT),
('AFTER', consts.CLUSTER_SCALE_IN),
('AFTER', consts.CLUSTER_RESIZE),
]
PROFILE_TYPE = [
'os.nova.server-1.0',
]
KEYS = (
POOL, VIP,
) = (
'pool', 'vip',
)
_POOL_KEYS = (
POOL_PROTOCOL, POOL_PROTOCOL_PORT, POOL_SUBNET,
POOL_LB_METHOD, POOL_ADMIN_STATE_UP, POOL_SESSION_PERSISTENCE,
) = (
'protocol', 'protocol_port', 'subnet',
'lb_method', 'admin_state_up', 'session_persistence',
)
PROTOCOLS = (
HTTP, HTTPS, TCP,
) = (
'HTTP', 'HTTPS', 'TCP',
)
LB_METHODS = (
ROUND_ROBIN, LEAST_CONNECTIONS, SOURCE_IP,
) = (
'ROUND_ROBIN', 'LEAST_CONNECTIONS', 'SOURCE_IP',
)
_VIP_KEYS = (
VIP_SUBNET, VIP_ADDRESS, VIP_CONNECTION_LIMIT, VIP_PROTOCOL,
VIP_PROTOCOL_PORT, VIP_ADMIN_STATE_UP,
) = (
'subnet', 'address', 'connection_limit', 'protocol',
'protocol_port', 'admin_state_up',
)
_SESSION_PERSISTENCE_KEYS = (
PERSISTENCE_TYPE, COOKIE_NAME,
) = (
'type', 'cookie_name',
)
PERSISTENCE_TYPES = (
PERSIST_SOURCE_IP, PERSIST_HTTP_COOKIE, PERSIST_APP_COOKIE,
) = (
'SOURCE_IP', 'HTTP_COOKIE', 'APP_COOKIE',
)
properties_schema = {
POOL: schema.Map(
_('LB pool properties.'),
schema={
POOL_PROTOCOL: schema.String(
_('Protocol used for load balancing.'),
constraints=[
constraints.AllowedValues(PROTOCOLS),
],
default=HTTP,
),
POOL_PROTOCOL_PORT: schema.Integer(
_('Port on which servers are running on the nodes.'),
default=80,
),
POOL_SUBNET: schema.String(
_('Name or ID of subnet for the port on which nodes can '
'be connected.'),
required=True,
),
POOL_LB_METHOD: schema.String(
_('Load balancing algorithm.'),
constraints=[
constraints.AllowedValues(LB_METHODS),
],
default=ROUND_ROBIN,
),
POOL_ADMIN_STATE_UP: schema.Boolean(
_('Administrative state of the pool.'),
default=True,
),
POOL_SESSION_PERSISTENCE: schema.Map(
_('Session pesistence configuration.'),
schema={
PERSISTENCE_TYPE: schema.String(
_('Type of session persistence implementation.'),
constraints=[
constraints.AllowedValues(PERSISTENCE_TYPES),
],
),
COOKIE_NAME: schema.String(
_('Name of cookie if type set to APP_COOKIE.'),
),
},
default={},
),
},
),
VIP: schema.Map(
_('VIP address and port of the pool.'),
schema={
VIP_SUBNET: schema.String(
_('Name or ID of Subnet on which the VIP address will be '
'allocated.'),
required=True,
),
VIP_ADDRESS: schema.String(
_('IP address of the VIP.'),
default=None,
),
VIP_CONNECTION_LIMIT: schema.Integer(
_('Maximum number of connections per second allowed for '
'this VIP'),
default=-1,
),
VIP_PROTOCOL: schema.String(
_('Protocol used for VIP.'),
constraints=[
constraints.AllowedValues(PROTOCOLS),
],
default=HTTP,
),
VIP_PROTOCOL_PORT: schema.Integer(
_('TCP port to listen on.'),
default=80,
),
VIP_ADMIN_STATE_UP: schema.Boolean(
_('Administrative state of the VIP.'),
default=True,
),
},
),
}
def __init__(self, name, spec, **kwargs):
super(LoadBalancingPolicy, self).__init__(name, spec, **kwargs)
self.pool_spec = self.properties.get(self.POOL, {})
self.vip_spec = self.properties.get(self.VIP, {})
self.validate()
self.lb = None
def validate(self):
super(LoadBalancingPolicy, self).validate()
# validate subnet's exists
# subnet = self.nc.subnet_get(vip[self.VIP_SUBNET])
def attach(self, cluster):
"""Routine to be invoked when policy is to be attached to a cluster.
:param cluster: The target cluster to be attached to;
:returns: When the operation was successful, returns a tuple (True,
message); otherwise, return a tuple (False, error).
"""
res, data = super(LoadBalancingPolicy, self).attach(cluster)
if res is False:
return False, data
nodes = node_mod.Node.load_all(oslo_context.get_current(),
cluster_id=cluster.id)
params = self._build_conn_params(cluster)
lb_driver = driver_base.SenlinDriver().loadbalancing(params)
res, data = lb_driver.lb_create(self.vip_spec, self.pool_spec)
if res is False:
return False, data
port = self.pool_spec.get(self.POOL_PROTOCOL_PORT)
subnet = self.pool_spec.get(self.POOL_SUBNET)
for node in nodes:
member_id = lb_driver.member_add(node, data['loadbalancer'],
data['pool'], port, subnet)
if member_id is None:
# When failed in adding member, remove all lb resources that
# were created and return the failure reason.
# TODO(Yanyan Hu): Maybe we should tolerate member adding
# failure and allow policy attaching to succeed without
# all nodes being added into lb pool?
lb_driver.lb_delete(**data)
return False, 'Failed in adding node into lb pool'
node.data.update({'lb_member': member_id})
node.store(oslo_context.get_current())
policy_data = self._build_policy_data(data)
return True, policy_data
def detach(self, cluster):
"""Routine to be called when the policy is detached from a cluster.
:param cluster: The cluster from which the policy is to be detached.
:returns: When the operation was successful, returns a tuple of
(True, data) where the data contains references to the resources
created; otherwise returns a tuple of (False, err) where the err
contains a error message.
"""
reason = _('LB resources deletion succeeded.')
params = self._build_conn_params(cluster)
lb_driver = driver_base.SenlinDriver().loadbalancing(params)
cp = cluster_policy.ClusterPolicy.load(oslo_context.get_current(),
cluster.id, self.id)
policy_data = self._extract_policy_data(cp.data)
if policy_data is None:
return True, reason
res, reason = lb_driver.lb_delete(**policy_data)
if res is False:
return False, reason
nodes = node_mod.Node.load_all(oslo_context.get_current(),
cluster_id=cluster.id)
for node in nodes:
if 'lb_member' in node.data:
node.data.pop('lb_member')
node.store(oslo_context.get_current())
return True, reason
def post_op(self, cluster_id, action):
"""Routine to be called after an action has been executed.
For this particular policy, we take this chance to update the pool
maintained by the load-balancer.
:param cluster_id: The ID of the cluster on which a relevant action
has been executed.
:param action: The action object that triggered this operation.
:returns: Nothing.
"""
nodes_added = action.outputs.get('nodes_added', [])
nodes_removed = action.outputs.get('nodes_removed', [])
if ((len(nodes_added) == 0) and (len(nodes_removed) == 0)):
return
db_cluster = db_api.cluster_get(action.context, cluster_id)
params = self._build_conn_params(db_cluster)
lb_driver = driver_base.SenlinDriver().loadbalancing(params)
cp = cluster_policy.ClusterPolicy.load(action.context, cluster_id,
self.id)
policy_data = self._extract_policy_data(cp.data)
lb_id = policy_data['loadbalancer']
pool_id = policy_data['pool']
port = self.pool_spec.get(self.POOL_PROTOCOL_PORT)
subnet = self.pool_spec.get(self.POOL_SUBNET)
# Remove nodes that have been deleted from lb pool
for node_id in nodes_removed:
node = node_mod.Node.load(action.context, node_id=node_id,
show_deleted=True)
member_id = node.data.get('lb_member', None)
if member_id is None:
LOG.warning(_LW('Node %(n)s not found in lb pool %(p)s.'),
{'n': node_id, 'p': pool_id})
continue
res = lb_driver.member_remove(lb_id, pool_id, member_id)
if res is not True:
action.data['status'] = base.CHECK_ERROR
action.data['reason'] = _('Failed in removing deleted '
'node(s) from lb pool.')
return
# Add new nodes to lb pool
for node_id in nodes_added:
node = node_mod.Node.load(action.context, node_id=node_id,
show_deleted=True)
member_id = node.data.get('lb_member', None)
if member_id:
LOG.warning(_LW('Node %(n)s already in lb pool %(p)s.'),
{'n': node_id, 'p': pool_id})
continue
member_id = lb_driver.member_add(node, lb_id, pool_id, port,
subnet)
if member_id is None:
action.data['status'] = base.CHECK_ERROR
action.data['reason'] = _('Failed in adding new node(s) '
'into lb pool.')
return
node.data.update({'lb_member': member_id})
node.store(action.context)
return
class StackProfile(base.Profile):
'''Profile for an OpenStack Heat stack.
When this profile is used, the whole cluster is a collection of Heat
stacks.
'''
KEYS = (
TEMPLATE,
CONTEXT,
PARAMETERS,
FILES,
TIMEOUT,
DISABLE_ROLLBACK,
ENVIRONMENT,
) = (
'template',
'context',
'parameters',
'files',
'timeout',
'disable_rollback',
'environment',
)
properties_schema = {
CONTEXT:
schema.Map(
_('A dictionary for specifying the customized context for '
'stack operations'),
default={},
),
TEMPLATE:
schema.Map(
_('Heat stack template.'),
required=True,
updatable=True,
),
PARAMETERS:
schema.Map(
_('Parameters to be passed to Heat for stack operations.'),
default={},
updatable=True,
),
FILES:
schema.Map(
_('Contents of files referenced by the template, if any.'),
default={},
updatable=True,
),
TIMEOUT:
schema.Integer(
_('A integer that specifies the number of minutes that a '
'stack operation times out.'),
updatable=True,
),
DISABLE_ROLLBACK:
schema.Boolean(
_('A boolean specifying whether a stack operation can be '
'rolled back.'),
default=True,
updatable=True,
),
ENVIRONMENT:
schema.Map(
_('A map that specifies the environment used for stack '
'operations.'),
default={},
updatable=True,
)
}
OP_NAMES = (OP_ABANDON, ) = ('abandon', )
OPERATIONS = {OP_ABANDON: schema.Map(_('Abandon a heat stack node.'), )}
def __init__(self, type_name, name, **kwargs):
super(StackProfile, self).__init__(type_name, name, **kwargs)
self.hc = None
self.stack_id = None
def heat(self, obj):
'''Construct heat client using the combined context.'''
if self.hc:
return self.hc
params = self._build_conn_params(obj.user, obj.project)
self.hc = driver_base.SenlinDriver().orchestration(params)
return self.hc
def do_validate(self, obj):
'''Validate if the spec has provided info for stack creation.'''
kwargs = {
'stack_name': obj.name,
'template': self.properties[self.TEMPLATE],
'timeout_mins': self.properties[self.TIMEOUT],
'disable_rollback': self.properties[self.DISABLE_ROLLBACK],
'parameters': self.properties[self.PARAMETERS],
'files': self.properties[self.FILES],
'environment': self.properties[self.ENVIRONMENT],
}
try:
self.heat(obj).stacks.validate(**kwargs)
except Exception as ex:
msg = _('Failed validate stack template due to '
'"%s"') % six.text_type(ex)
raise exception.InvalidSpec(message=msg)
return True
def _check_action_complete(self, obj, action):
stack = self.heat(obj).stack_get(self.stack_id)
status = stack.status.split('_', 1)
if status[0] == action:
if status[1] == 'IN_PROGRESS':
return False
if status[1] == 'COMPLETE':
return True
raise exception.ResourceStatusError(resource_id=self.stack_id,
status=stack.status,
reason=stack.status_reason)
else:
return False
def do_create(self, obj):
'''Create a stack using the given profile.'''
kwargs = {
'stack_name': obj.name + '-' + utils.random_name(8),
'template': self.properties[self.TEMPLATE],
'timeout_mins': self.properties[self.TIMEOUT],
'disable_rollback': self.properties[self.DISABLE_ROLLBACK],
'parameters': self.properties[self.PARAMETERS],
'files': self.properties[self.FILES],
'environment': self.properties[self.ENVIRONMENT],
}
LOG.info('Creating stack: %s' % kwargs)
stack = self.heat(obj).stack_create(**kwargs)
self.stack_id = stack.id
# Wait for action to complete/fail
while not self._check_action_complete(obj, 'CREATE'):
scheduler.sleep(1)
return stack.id
def do_delete(self, obj):
self.stack_id = obj.physical_id
try:
self.heat(obj).stack_delete(self.stack_id, True)
self.heat(obj).wait_for_stack_delete(self.stack_id)
except Exception as ex:
LOG.error('Error: %s' % six.text_type(ex))
raise ex
return True
def do_update(self, obj, new_profile, **params):
'''Perform update on object.
:param obj: the node object to operate on
:param new_profile: the new profile used for updating
:param params: other parametes for the update request.
'''
self.stack_id = obj.physical_id
if not self.stack_id:
return True
if not self.validate_for_update(new_profile):
return False
fields = {}
new_template = new_profile.properties[new_profile.TEMPLATE]
if new_template != self.properties[self.TEMPLATE]:
fields['template'] = new_template
new_params = new_profile.properties[new_profile.PARAMETERS]
if new_params != self.properties[self.PARAMETERS]:
fields['parameters'] = new_params
new_timeout = new_profile.properties[new_profile.TIMEOUT]
if new_timeout != self.properties[self.TIMEOUT]:
fields['timeout_mins'] = new_timeout
new_dr = new_profile.properties[new_profile.DISABLE_ROLLBACK]
if new_dr != self.properties[self.DISABLE_ROLLBACK]:
fields['disable_rollback'] = new_dr
new_files = new_profile.properties[new_profile.FILES]
if new_files != self.properties[self.FILES]:
fields['files'] = new_files
new_environment = new_profile.properties[new_profile.ENVIRONMENT]
if new_environment != self.properties[self.ENVIRONMENT]:
fields['environment'] = new_environment
if fields:
try:
self.heat(obj).stack_update(self.stack_id, **fields)
except Exception as ex:
LOG.exception(_('Failed in updating stack: %s'),
six.text_type(ex))
return False
# Wait for action to complete/fail
while not self._check_action_complete(obj, 'UPDATE'):
scheduler.sleep(1)
return True
def do_check(self, obj):
"""Check stack status."""
hc = self.heat(obj)
try:
stack = hc.stack_get(obj.physical_id)
except Exception as ex:
raise ex
# When the stack is in a status which can't be checked(
# CREATE_IN_PROGRESS, DELETE_IN_PROGRESS, etc), return False.
try:
stack.check(hc.session)
except Exception:
return False
status = stack.status
while status == 'CHECK_IN_PROGRESS':
status = hc.stack_get(obj.physical_id).status
if status == 'CHECK_COMPLETE':
return True
else:
return False
def do_get_details(self, obj):
if obj.physical_id is None or obj.physical_id == '':
return {}
return self.heat(obj).stack_get(obj.physical_id)
def handle_abandon(self, obj, **options):
"""Handler for abandoning a heat stack node."""
pass
class ServerProfile(base.Profile):
"""Profile for an OpenStack Nova server."""
KEYS = (
CONTEXT,
ADMIN_PASS,
AUTO_DISK_CONFIG,
AVAILABILITY_ZONE,
BLOCK_DEVICE_MAPPING_V2,
CONFIG_DRIVE,
FLAVOR,
IMAGE,
KEY_NAME,
METADATA,
NAME,
NETWORKS,
PERSONALITY,
SECURITY_GROUPS,
USER_DATA,
SCHEDULER_HINTS,
) = (
'context',
'admin_pass',
'auto_disk_config',
'availability_zone',
'block_device_mapping_v2',
'config_drive',
'flavor',
'image',
'key_name',
'metadata',
'name',
'networks',
'personality',
'security_groups',
'user_data',
'scheduler_hints',
)
BDM2_KEYS = (
BDM2_UUID,
BDM2_SOURCE_TYPE,
BDM2_DESTINATION_TYPE,
BDM2_DISK_BUS,
BDM2_DEVICE_NAME,
BDM2_VOLUME_SIZE,
BDM2_GUEST_FORMAT,
BDM2_BOOT_INDEX,
BDM2_DEVICE_TYPE,
BDM2_DELETE_ON_TERMINATION,
) = (
'uuid',
'source_type',
'destination_type',
'disk_bus',
'device_name',
'volume_size',
'guest_format',
'boot_index',
'device_type',
'delete_on_termination',
)
NETWORK_KEYS = (
PORT,
FIXED_IP,
NETWORK,
) = (
'port',
'fixed_ip',
'network',
)
PERSONALITY_KEYS = (
PATH,
CONTENTS,
) = (
'path',
'contents',
)
SCHEDULER_HINTS_KEYS = (GROUP, ) = ('group', )
properties_schema = {
CONTEXT:
schema.Map(_('Customized security context for operating servers.'), ),
ADMIN_PASS:
schema.String(_('Password for the administrator account.'), ),
AUTO_DISK_CONFIG:
schema.Boolean(
_('Whether the disk partition is done automatically.'),
default=True,
),
AVAILABILITY_ZONE:
schema.String(
_('Name of availability zone for running the server.'), ),
BLOCK_DEVICE_MAPPING_V2:
schema.List(
_('A list specifying the properties of block devices to be used '
'for this server.'),
schema=schema.Map(
_('A map specifying the properties of a block device to be '
'used by the server.'),
schema={
BDM2_UUID:
schema.String(
_('ID of the source image, snapshot or volume'), ),
BDM2_SOURCE_TYPE:
schema.String(
_('Volume source type, should be image, snapshot, '
'volume or blank'),
required=True,
),
BDM2_DESTINATION_TYPE:
schema.String(
_('Volume destination type, should be volume or '
'local'),
required=True,
),
BDM2_DISK_BUS:
schema.String(_('Bus of the device.'), ),
BDM2_DEVICE_NAME:
schema.String(
_('Name of the device(e.g. vda, xda, ....).'), ),
BDM2_VOLUME_SIZE:
schema.Integer(
_('Size of the block device in MB(for swap) and '
'in GB(for other formats)'),
required=True,
),
BDM2_GUEST_FORMAT:
schema.String(
_('Specifies the disk file system format(e.g. swap, '
'ephemeral, ...).'), ),
BDM2_BOOT_INDEX:
schema.Integer(_('Define the boot order of the device'), ),
BDM2_DEVICE_TYPE:
schema.String(
_('Type of the device(e.g. disk, cdrom, ...).'), ),
BDM2_DELETE_ON_TERMINATION:
schema.Boolean(
_('Whether to delete the volume when the server '
'stops.'), ),
}),
),
CONFIG_DRIVE:
schema.Boolean(
_('Whether config drive should be enabled for the server.'), ),
FLAVOR:
schema.String(
_('ID of flavor used for the server.'),
required=True,
updatable=True,
),
IMAGE:
schema.String(
# IMAGE is not required, because there could be BDM or BDMv2
# support and the corresponding settings effective
_('ID of image to be used for the new server.'),
updatable=True,
),
KEY_NAME:
schema.String(_('Name of Nova keypair to be injected to server.'), ),
METADATA:
schema.Map(
_('A collection of key/value pairs to be associated with the '
'server created. Both key and value should be <=255 chars.'),
updatable=True,
),
NAME:
schema.String(
_('Name of the server. When omitted, the node name will be used.'),
updatable=True,
),
NETWORKS:
schema.List(
_('List of networks for the server.'),
schema=schema.Map(
_('A map specifying the properties of a network for uses.'),
schema={
NETWORK:
schema.String(
_('Name or ID of network to create a port on.'), ),
PORT:
schema.String(_('Port ID to be used by the network.'), ),
FIXED_IP:
schema.String(_('Fixed IP to be used by the network.'), ),
},
),
updatable=True,
),
PERSONALITY:
schema.List(
_('List of files to be injected into the server, where each.'),
schema=schema.Map(
_('A map specifying the path & contents for an injected '
'file.'),
schema={
PATH:
schema.String(
_('In-instance path for the file to be injected.'),
required=True,
),
CONTENTS:
schema.String(
_('Contents of the file to be injected.'),
required=True,
),
},
),
),
SCHEDULER_HINTS:
schema.Map(
_('A collection of key/value pairs to be associated with the '
'Scheduler hints. Both key and value should be <=255 chars.'), ),
SECURITY_GROUPS:
schema.List(
_('List of security groups.'),
schema=schema.String(
_('Name of a security group'),
required=True,
),
),
USER_DATA:
schema.String(_('User data to be exposed by the metadata server.'), ),
}
OP_NAMES = (
OP_REBOOT,
OP_CHANGE_PASSWORD,
) = (
'reboot',
'change_password',
)
REBOOT_TYPE = 'type'
REBOOT_TYPES = (REBOOT_SOFT, REBOOT_HARD) = ('SOFT', 'HARD')
ADMIN_PASSWORD = '******'
OPERATIONS = {
OP_REBOOT:
schema.Operation(
_("Reboot the nova server."),
schema={
REBOOT_TYPE:
schema.StringParam(
_("Type of reboot which can be 'SOFT' or 'HARD'."),
default=REBOOT_SOFT,
constraints=[
constraints.AllowedValues(REBOOT_TYPES),
])
}),
OP_CHANGE_PASSWORD:
schema.Operation(_("Change the administrator password."),
schema={
ADMIN_PASSWORD:
schema.StringParam(
_("New password for the administrator."))
}),
}
def __init__(self, type_name, name, **kwargs):
super(ServerProfile, self).__init__(type_name, name, **kwargs)
self.server_id = None
def _validate_az(self, obj, az_name, reason=None):
try:
res = self.compute(obj).validate_azs([az_name])
except exc.InternalError as ex:
if reason == 'create':
raise exc.EResourceCreation(type='server',
message=six.text_type(ex))
else:
raise
if not res:
msg = _("The specified %(key)s '%(value)s' could not be found") % {
'key': self.AVAILABILITY_ZONE,
'value': az_name
}
if reason == 'create':
raise exc.EResourceCreation(type='server', message=msg)
else:
raise exc.InvalidSpec(message=msg)
return az_name
def _validate_flavor(self, obj, name_or_id, reason=None):
flavor = None
msg = ''
try:
flavor = self.compute(obj).flavor_find(name_or_id, False)
except exc.InternalError as ex:
msg = six.text_type(ex)
if reason is None: # reaons is 'validate'
if ex.code == 404:
msg = _(
"The specified %(k)s '%(v)s' could not be found.") % {
'k': self.FLAVOR,
'v': name_or_id
}
raise exc.InvalidSpec(message=msg)
else:
raise
if flavor is not None:
if not flavor.is_disabled:
return flavor
msg = _("The specified %(k)s '%(v)s' is disabled") % {
'k': self.FLAVOR,
'v': name_or_id
}
if reason == 'create':
raise exc.EResourceCreation(type='server', message=msg)
elif reason == 'update':
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=msg)
else:
raise exc.InvalidSpec(message=msg)
def _validate_image(self, obj, name_or_id, reason=None):
try:
return self.compute(obj).image_find(name_or_id, False)
except exc.InternalError as ex:
if reason == 'create':
raise exc.EResourceCreation(type='server',
message=six.text_type(ex))
elif reason == 'update':
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
elif ex.code == 404:
msg = _("The specified %(k)s '%(v)s' could not be found.") % {
'k': self.IMAGE,
'v': name_or_id
}
raise exc.InvalidSpec(message=msg)
else:
raise
def _validate_keypair(self, obj, name_or_id, reason=None):
try:
return self.compute(obj).keypair_find(name_or_id, False)
except exc.InternalError as ex:
if reason == 'create':
raise exc.EResourceCreation(type='server',
message=six.text_type(ex))
elif reason == 'update':
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
elif ex.code == 404:
msg = _("The specified %(k)s '%(v)s' could not be found.") % {
'k': self.KEY_NAME,
'v': name_or_id
}
raise exc.InvalidSpec(message=msg)
else:
raise
def do_validate(self, obj):
"""Validate if the spec has provided valid info for server creation.
:param obj: The node object.
"""
# validate availability_zone
az_name = self.properties[self.AVAILABILITY_ZONE]
if az_name is not None:
self._validate_az(obj, az_name)
# validate flavor
flavor = self.properties[self.FLAVOR]
self._validate_flavor(obj, flavor)
# validate image
image = self.properties[self.IMAGE]
if image is not None:
self._validate_image(obj, image)
# validate key_name
keypair = self.properties[self.KEY_NAME]
if keypair is not None:
self._validate_keypair(obj, keypair)
# validate networks
networks = self.properties[self.NETWORKS]
for net in networks:
self._validate_network(obj, net)
return True
def _resolve_bdm(self, bdm):
for bd in bdm:
for key in self.BDM2_KEYS:
if bd[key] is None:
del bd[key]
return bdm
def _validate_network(self, obj, network, reason=None):
result = {}
error = None
# check network
net_ident = network.get(self.NETWORK)
if net_ident:
try:
net = self.network(obj).network_get(net_ident)
if reason == 'update':
result['net_id'] = net.id
else:
result['uuid'] = net.id
except exc.InternalError as ex:
error = six.text_type(ex)
# check port
port_ident = network.get(self.PORT)
if not error and port_ident:
try:
port = self.network(obj).port_find(port_ident)
if port.status != 'DOWN':
error = _(
"The status of the port %(port)s must be DOWN") % {
'port': port_ident
}
if reason == 'update':
result['port_id'] = port.id
else:
result['port'] = port.id
except exc.InternalError as ex:
error = six.text_type(ex)
elif port_ident is None and net_ident is None:
error = _("'%(port)s' is required if '%(net)s' is omitted") % {
'port': self.PORT,
'net': self.NETWORK
}
fixed_ip = network.get(self.FIXED_IP)
if not error and fixed_ip:
if port_ident is not None:
error = _("The '%(port)s' property and the '%(fixed_ip)s' "
"property cannot be specified at the same time") % {
'port': self.PORT,
'fixed_ip': self.FIXED_IP
}
else:
if reason == 'update':
result['fixed_ips'] = [{'ip_address': fixed_ip}]
else:
result['fixed_ip'] = fixed_ip
if error:
if reason == 'create':
raise exc.EResourceCreation(type='server', message=error)
elif reason == 'update':
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=error)
else:
raise exc.InvalidSpec(message=error)
return result
def _build_metadata(self, obj, usermeta):
"""Build custom metadata for server.
:param obj: The node object to operate on.
:return: A dictionary containing the new metadata.
"""
metadata = usermeta or {}
metadata['cluster_node_id'] = obj.id
if obj.cluster_id:
metadata['cluster_id'] = obj.cluster_id
metadata['cluster_node_index'] = six.text_type(obj.index)
return metadata
def do_create(self, obj):
"""Create a server for the node object.
:param obj: The node object for which a server will be created.
"""
kwargs = {}
for key in self.KEYS:
# context is treated as connection parameters
if key == self.CONTEXT:
continue
if self.properties[key] is not None:
kwargs[key] = self.properties[key]
admin_pass = self.properties[self.ADMIN_PASS]
if admin_pass:
kwargs.pop(self.ADMIN_PASS)
kwargs['adminPass'] = admin_pass
auto_disk_config = self.properties[self.AUTO_DISK_CONFIG]
kwargs.pop(self.AUTO_DISK_CONFIG)
kwargs['OS-DCF:diskConfig'] = 'AUTO' if auto_disk_config else 'MANUAL'
image_ident = self.properties[self.IMAGE]
if image_ident is not None:
image = self._validate_image(obj, image_ident, 'create')
kwargs.pop(self.IMAGE)
kwargs['imageRef'] = image.id
flavor_ident = self.properties[self.FLAVOR]
flavor = self._validate_flavor(obj, flavor_ident, 'create')
kwargs.pop(self.FLAVOR)
kwargs['flavorRef'] = flavor.id
keypair_name = self.properties[self.KEY_NAME]
if keypair_name:
keypair = self._validate_keypair(obj, keypair_name, 'create')
kwargs['key_name'] = keypair.name
kwargs['name'] = self.properties[self.NAME] or obj.name
metadata = self._build_metadata(obj, self.properties[self.METADATA])
kwargs['metadata'] = metadata
block_device_mapping_v2 = self.properties[self.BLOCK_DEVICE_MAPPING_V2]
if block_device_mapping_v2 is not None:
kwargs['block_device_mapping_v2'] = self._resolve_bdm(
block_device_mapping_v2)
user_data = self.properties[self.USER_DATA]
if user_data is not None:
ud = encodeutils.safe_encode(user_data)
kwargs['user_data'] = encodeutils.safe_decode(base64.b64encode(ud))
networks = self.properties[self.NETWORKS]
if networks is not None:
kwargs['networks'] = []
for net_spec in networks:
net = self._validate_network(obj, net_spec, 'create')
kwargs['networks'].append(net)
secgroups = self.properties[self.SECURITY_GROUPS]
if secgroups:
kwargs['security_groups'] = [{'name': sg} for sg in secgroups]
if 'placement' in obj.data:
if 'zone' in obj.data['placement']:
kwargs['availability_zone'] = obj.data['placement']['zone']
if 'servergroup' in obj.data['placement']:
group_id = obj.data['placement']['servergroup']
hints = self.properties.get(self.SCHEDULER_HINTS, {})
hints.update({'group': group_id})
kwargs['scheduler_hints'] = hints
server = None
resource_id = 'UNKNOWN'
try:
server = self.compute(obj).server_create(**kwargs)
self.compute(obj).wait_for_server(server.id)
return server.id
except exc.InternalError as ex:
if server and server.id:
resource_id = server.id
raise exc.EResourceCreation(type='server',
message=ex.message,
resource_id=resource_id)
def do_delete(self, obj, **params):
"""Delete the physical resource associated with the specified node.
:param obj: The node object to operate on.
:param kwargs params: Optional keyword arguments for the delete
operation.
:returns: This operation always return True unless exception is
caught.
:raises: `EResourceDeletion` if interaction with compute service fails.
"""
if not obj.physical_id:
return True
server_id = obj.physical_id
ignore_missing = params.get('ignore_missing', True)
force = params.get('force', False)
try:
driver = self.compute(obj)
if force:
driver.server_force_delete(server_id, ignore_missing)
else:
driver.server_delete(server_id, ignore_missing)
driver.wait_for_server_delete(server_id)
return True
except exc.InternalError as ex:
raise exc.EResourceDeletion(type='server',
id=server_id,
message=six.text_type(ex))
def _check_server_name(self, obj, profile):
"""Check if there is a new name to be assigned to the server.
:param obj: The node object to operate on.
:param new_profile: The new profile which may contain a name for
the server instance.
:return: A tuple consisting a boolean indicating whether the name
needs change and the server name determined.
"""
old_name = self.properties[self.NAME] or obj.name
new_name = profile.properties[self.NAME] or obj.name
if old_name == new_name:
return False, new_name
return True, new_name
def _update_name(self, obj, new_name):
"""Update the name of the server.
:param obj: The node object to operate.
:param new_name: The new name for the server instance.
:return: ``None``.
:raises: ``EResourceUpdate``.
"""
try:
self.compute(obj).server_update(obj.physical_id, name=new_name)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
def _check_password(self, obj, new_profile):
"""Check if the admin password has been changed in the new profile.
:param obj: The server node to operate, not used currently.
:param new_profile: The new profile which may contain a new password
for the server instance.
:return: A tuple consisting a boolean indicating whether the password
needs a change and the password determined which could be
'' if new password is not set.
"""
old_passwd = self.properties.get(self.ADMIN_PASS) or ''
new_passwd = new_profile.properties[self.ADMIN_PASS] or ''
if old_passwd == new_passwd:
return False, new_passwd
return True, new_passwd
def _update_password(self, obj, new_password):
"""Update the admin password for the server.
:param obj: The node object to operate.
:param new_password: The new password for the server instance.
:return: ``None``.
:raises: ``EResourceUpdate``.
"""
try:
self.compute(obj).server_change_password(obj.physical_id,
new_password)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
def _update_metadata(self, obj, new_profile):
"""Update the server metadata.
:param obj: The node object to operate on.
:param new_profile: The new profile that may contain some changes to
the metadata.
:returns: ``None``
:raises: `EResourceUpdate`.
"""
old_meta = self._build_metadata(obj, self.properties[self.METADATA])
new_meta = self._build_metadata(obj,
new_profile.properties[self.METADATA])
if new_meta == old_meta:
return
try:
self.compute(obj).server_metadata_update(obj.physical_id, new_meta)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
def _update_flavor(self, obj, new_profile):
"""Update server flavor.
:param obj: The node object to operate on.
:param old_flavor: The identity of the current flavor.
:param new_flavor: The identity of the new flavor.
:returns: ``None``.
:raises: `EResourceUpdate` when operation was a failure.
"""
old_flavor = self.properties[self.FLAVOR]
new_flavor = new_profile.properties[self.FLAVOR]
cc = self.compute(obj)
oldflavor = self._validate_flavor(obj, old_flavor, 'update')
newflavor = self._validate_flavor(obj, new_flavor, 'update')
if oldflavor.id == newflavor.id:
return
try:
cc.server_resize(obj.physical_id, newflavor.id)
cc.wait_for_server(obj.physical_id, 'VERIFY_RESIZE')
except exc.InternalError as ex:
msg = six.text_type(ex)
try:
cc.server_resize_revert(obj.physical_id)
cc.wait_for_server(obj.physical_id, 'ACTIVE')
except exc.InternalError as ex1:
msg = six.text_type(ex1)
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=msg)
try:
cc.server_resize_confirm(obj.physical_id)
cc.wait_for_server(obj.physical_id, 'ACTIVE')
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
def _update_image(self, obj, new_profile, new_name, new_password):
"""Update image used by server node.
:param obj: The node object to operate on.
:param new_profile: The profile which may contain a new image name or
ID to use.
:param new_name: The name for the server node.
:param newn_password: The new password for the administrative account
if provided.
:returns: A boolean indicating whether the image needs an update.
:raises: ``InternalError`` if operation was a failure.
"""
old_image = self.properties[self.IMAGE]
new_image = new_profile.properties[self.IMAGE]
if not new_image:
msg = _("Updating Nova server with image set to None is not "
"supported by Nova")
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=msg)
# check the new image first
img_new = self._validate_image(obj, new_image, reason='update')
new_image_id = img_new.id
driver = self.compute(obj)
if old_image:
img_old = self._validate_image(obj, old_image, reason='update')
old_image_id = img_old.id
else:
try:
server = driver.server_get(obj.physical_id)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
# Still, this 'old_image_id' could be empty, but it doesn't matter
# because the comparison below would fail if that is the case
old_image_id = server.image.get('id', None)
if new_image_id == old_image_id:
return False
try:
driver.server_rebuild(obj.physical_id, new_image_id, new_name,
new_password)
driver.wait_for_server(obj.physical_id, 'ACTIVE')
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
return True
def _create_interfaces(self, obj, networks):
"""Create new interfaces for the server node.
:param obj: The node object to operate.
:param networks: A list containing information about new network
interfaces to be created.
:returns: ``None``.
:raises: ``EResourceUpdate`` if interaction with drivers failed.
"""
cc = self.compute(obj)
try:
server = cc.server_get(obj.physical_id)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
for net_spec in networks:
net_attrs = self._validate_network(obj, net_spec, 'update')
if net_attrs:
try:
cc.server_interface_create(server, **net_attrs)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
def _delete_interfaces(self, obj, networks):
"""Delete existing interfaces from the node.
:param obj: The node object to operate.
:param networks: A list containing information about network
interfaces to be created.
:returns: ``None``
:raises: ``EResourceUpdate``
"""
def _get_network(nc, net_id, server_id):
try:
net = nc.network_get(net_id)
return net.id
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=server_id,
message=six.text_type(ex))
def _do_delete(port_id, server_id):
try:
cc.server_interface_delete(port_id, server_id)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=server_id,
message=six.text_type(ex))
cc = self.compute(obj)
nc = self.network(obj)
try:
existing = list(cc.server_interface_list(obj.physical_id))
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='server',
id=obj.physical_id,
message=six.text_type(ex))
ports = []
for intf in existing:
fixed_ips = [addr['ip_address'] for addr in intf.fixed_ips]
ports.append({
'id': intf.port_id,
'net': intf.net_id,
'ips': fixed_ips
})
for n in networks:
network = n.get('network', None)
port = n.get('port', None)
fixed_ip = n.get('fixed_ip', None)
if port:
for p in ports:
if p['id'] == port:
ports.remove(p)
_do_delete(port, obj.physical_id)
elif fixed_ip:
net_id = _get_network(nc, network, obj.physical_id)
for p in ports:
if (fixed_ip in p['ips'] and net_id == p['net']):
ports.remove(p)
_do_delete(p['id'], obj.physical_id)
elif port is None and fixed_ip is None:
net_id = _get_network(nc, network, obj.physical_id)
for p in ports:
if p['net'] == net_id:
ports.remove(p)
_do_delete(p['id'], obj.physical_id)
def _update_network(self, obj, new_profile):
"""Updating server network interfaces.
:param obj: The node object to operate.
:param new_profile: The new profile which may contain new network
settings.
:return: ``None``
:raises: ``EResourceUpdate`` if there are driver failures.
"""
networks_current = self.properties[self.NETWORKS]
networks_create = new_profile.properties[self.NETWORKS]
networks_delete = copy.deepcopy(networks_current)
for network in networks_current:
if network in networks_create:
networks_create.remove(network)
networks_delete.remove(network)
# Detach some existing interfaces
if networks_delete:
self._delete_interfaces(obj, networks_delete)
# Attach new interfaces
if networks_create:
self._create_interfaces(obj, networks_create)
return
def do_update(self, obj, new_profile=None, **params):
"""Perform update on the server.
:param obj: the server to operate on
:param new_profile: the new profile for the server.
:param params: a dictionary of optional parameters.
:returns: True if update was successful or False otherwise.
:raises: `EResourceUpdate` if operation fails.
"""
self.server_id = obj.physical_id
if not self.server_id:
return False
if not new_profile:
return False
if not self.validate_for_update(new_profile):
return False
name_changed, new_name = self._check_server_name(obj, new_profile)
passwd_changed, new_passwd = self._check_password(obj, new_profile)
# Update server image: may have side effect of changing server name
# and/or admin password
image_changed = self._update_image(obj, new_profile, new_name,
new_passwd)
if not image_changed:
# we do this separately only when rebuild wasn't performed
if name_changed:
self._update_name(obj, new_name)
if passwd_changed:
self._update_password(obj, new_passwd)
# Update server flavor: note that flavor is a required property
self._update_flavor(obj, new_profile)
self._update_network(obj, new_profile)
# TODO(Yanyan Hu): Update block_device properties
# Update server metadata
self._update_metadata(obj, new_profile)
return True
def do_get_details(self, obj):
known_keys = {
'OS-DCF:diskConfig', 'OS-EXT-AZ:availability_zone',
'OS-EXT-STS:power_state', 'OS-EXT-STS:vm_state', 'accessIPv4',
'accessIPv6', 'config_drive', 'created', 'hostId', 'id',
'key_name', 'locked', 'metadata', 'name',
'os-extended-volumes:volumes_attached', 'progress', 'status',
'updated'
}
if obj.physical_id is None or obj.physical_id == '':
return {}
driver = self.compute(obj)
try:
server = driver.server_get(obj.physical_id)
except exc.InternalError as ex:
return {'Error': {'code': ex.code, 'message': six.text_type(ex)}}
if server is None:
return {}
server_data = server.to_dict()
details = {
'image': server_data['image']['id'],
'flavor': server_data['flavor']['id'],
}
for key in known_keys:
if key in server_data:
details[key] = server_data[key]
# process special keys like 'OS-EXT-STS:task_state': these keys have
# a default value '-' when not existing
special_keys = [
'OS-EXT-STS:task_state',
'OS-SRV-USG:launched_at',
'OS-SRV-USG:terminated_at',
]
for key in special_keys:
if key in server_data:
val = server_data[key]
details[key] = val if val else '-'
# process network addresses
details['addresses'] = copy.deepcopy(server_data['addresses'])
# process security groups
sgroups = []
if 'security_groups' in server_data:
for sg in server_data['security_groups']:
sgroups.append(sg['name'])
if len(sgroups) == 0:
details['security_groups'] = ''
elif len(sgroups) == 1:
details['security_groups'] = sgroups[0]
else:
details['security_groups'] = sgroups
return dict((k, details[k]) for k in sorted(details))
def do_join(self, obj, cluster_id):
if not obj.physical_id:
return False
driver = self.compute(obj)
metadata = driver.server_metadata_get(obj.physical_id) or {}
metadata['cluster_id'] = cluster_id
metadata['cluster_node_index'] = six.text_type(obj.index)
driver.server_metadata_update(obj.physical_id, metadata)
return super(ServerProfile, self).do_join(obj, cluster_id)
def do_leave(self, obj):
if not obj.physical_id:
return False
keys = ['cluster_id', 'cluster_node_index']
self.compute(obj).server_metadata_delete(obj.physical_id, keys)
return super(ServerProfile, self).do_leave(obj)
def do_rebuild(self, obj):
if not obj.physical_id:
return False
self.server_id = obj.physical_id
driver = self.compute(obj)
try:
server = driver.server_get(self.server_id)
except exc.InternalError as ex:
raise exc.EResourceOperation(op='rebuilding',
type='server',
id=self.server_id,
message=six.text_type(ex))
if server is None or server.image is None:
return False
image_id = server.image['id']
admin_pass = self.properties.get(self.ADMIN_PASS)
try:
driver.server_rebuild(self.server_id, image_id,
self.properties.get(self.NAME), admin_pass)
driver.wait_for_server(self.server_id, 'ACTIVE')
except exc.InternalError as ex:
raise exc.EResourceOperation(op='rebuilding',
type='server',
id=self.server_id,
message=six.text_type(ex))
return True
def do_check(self, obj):
if not obj.physical_id:
return False
try:
server = self.compute(obj).server_get(obj.physical_id)
except exc.InternalError as ex:
raise exc.EResourceOperation(op='checking',
type='server',
id=obj.physical_id,
message=six.text_type(ex))
if (server is None or server.status != 'ACTIVE'):
return False
return True
def do_recover(self, obj, **options):
# NOTE: We do a 'get' not a 'pop' here, because the operations may
# get fall back to the base class for handling
operation = options.get('operation', None)
if operation and not isinstance(operation, six.string_types):
operation = operation[0]
# TODO(Qiming): Handle the case that the operation contains other
# alternative recover operation
# Depends-On: https://review.openstack.org/#/c/359676/
if operation == 'REBUILD':
return self.do_rebuild(obj)
return super(ServerProfile, self).do_recover(obj, **options)
def handle_reboot(self, obj, **options):
"""Handler for the reboot operation."""
if not obj.physical_id:
return False
reboot_type = options.get(self.REBOOT_TYPE, self.REBOOT_SOFT)
if (not isinstance(reboot_type, six.string_types)
or reboot_type not in self.REBOOT_TYPES):
return False
self.compute(obj).server_reboot(obj.physical_id, reboot_type)
self.compute(obj).wait_for_server(obj.physical_id, 'ACTIVE')
return True
def handle_change_password(self, obj, **options):
"""Handler for the change_password operation."""
if not obj.physical_id:
return False
password = options.get(self.ADMIN_PASSWORD, None)
if (password is None or not isinstance(password, six.string_types)):
return False
self.compute(obj).server_change_password(obj.physical_id, password)
return True
class StackProfile(base.Profile):
"""Profile for an OpenStack Heat stack."""
VERSIONS = {
'1.0': [
{'status': consts.SUPPORTED, 'since': '2016.04'}
]
}
KEYS = (
CONTEXT, TEMPLATE, TEMPLATE_URL, PARAMETERS,
FILES, TIMEOUT, DISABLE_ROLLBACK, ENVIRONMENT,
) = (
'context', 'template', 'template_url', 'parameters',
'files', 'timeout', 'disable_rollback', 'environment',
)
properties_schema = {
CONTEXT: schema.Map(
_('A dictionary for specifying the customized context for '
'stack operations'),
default={},
),
TEMPLATE: schema.Map(
_('Heat stack template.'),
default={},
updatable=True,
),
TEMPLATE_URL: schema.String(
_('Heat stack template url.'),
default='',
updatable=True,
),
PARAMETERS: schema.Map(
_('Parameters to be passed to Heat for stack operations.'),
default={},
updatable=True,
),
FILES: schema.Map(
_('Contents of files referenced by the template, if any.'),
default={},
updatable=True,
),
TIMEOUT: schema.Integer(
_('A integer that specifies the number of minutes that a '
'stack operation times out.'),
updatable=True,
),
DISABLE_ROLLBACK: schema.Boolean(
_('A boolean specifying whether a stack operation can be '
'rolled back.'),
default=True,
updatable=True,
),
ENVIRONMENT: schema.Map(
_('A map that specifies the environment used for stack '
'operations.'),
default={},
updatable=True,
)
}
OP_NAMES = (
OP_ABANDON,
) = (
'abandon',
)
OPERATIONS = {
OP_ABANDON: schema.Map(
_('Abandon a heat stack node.'),
)
}
def __init__(self, type_name, name, **kwargs):
super(StackProfile, self).__init__(type_name, name, **kwargs)
self.stack_id = None
def validate(self, validate_props=False):
'''Validate the schema and the data provided.'''
# general validation
self.spec_data.validate()
self.properties.validate()
# validate template
template = self.properties[self.TEMPLATE]
template_url = self.properties[self.TEMPLATE_URL]
if not template and not template_url:
msg = _("Both template and template_url are not specified "
"for profile '%s'.") % self.name
raise exc.InvalidSpec(message=msg)
if validate_props:
self.do_validate(obj=self)
def do_validate(self, obj):
"""Validate the stack template used by a node.
:param obj: Node object to operate.
:returns: True if validation succeeds.
:raises: `InvalidSpec` exception is raised if template is invalid.
"""
kwargs = {
'stack_name': utils.random_name(),
'template': self.properties[self.TEMPLATE],
'template_url': self.properties[self.TEMPLATE_URL],
'parameters': self.properties[self.PARAMETERS],
'files': self.properties[self.FILES],
'environment': self.properties[self.ENVIRONMENT],
'preview': True,
}
try:
self.orchestration(obj).stack_create(**kwargs)
except exc.InternalError as ex:
msg = _('Failed in validating template: %s') % six.text_type(ex)
raise exc.InvalidSpec(message=msg)
return True
def do_create(self, obj):
"""Create a heat stack using the given node object.
:param obj: The node object to operate on.
:returns: The UUID of the heat stack created.
"""
tags = ["cluster_node_id=%s" % obj.id]
if obj.cluster_id:
tags.append('cluster_id=%s' % obj.cluster_id)
tags.append('cluster_node_index=%s' % obj.index)
kwargs = {
'stack_name': obj.name + '-' + utils.random_name(8),
'template': self.properties[self.TEMPLATE],
'template_url': self.properties[self.TEMPLATE_URL],
'timeout_mins': self.properties[self.TIMEOUT],
'disable_rollback': self.properties[self.DISABLE_ROLLBACK],
'parameters': self.properties[self.PARAMETERS],
'files': self.properties[self.FILES],
'environment': self.properties[self.ENVIRONMENT],
'tags': ",".join(tags)
}
try:
stack = self.orchestration(obj).stack_create(**kwargs)
# Timeout = None means we will use the 'default_action_timeout'
# It can be overridden by the TIMEOUT profile properties
timeout = None
if self.properties[self.TIMEOUT]:
timeout = self.properties[self.TIMEOUT] * 60
self.orchestration(obj).wait_for_stack(stack.id, 'CREATE_COMPLETE',
timeout=timeout)
return stack.id
except exc.InternalError as ex:
raise exc.EResourceCreation(type='stack',
message=six.text_type(ex))
def do_delete(self, obj, **params):
"""Delete the physical stack behind the node object.
:param obj: The node object to operate on.
:param kwargs params: Optional keyword arguments for the delete
operation.
:returns: This operation always returns True unless exception is
caught.
:raises: `EResourceDeletion` if interaction with heat fails.
"""
stack_id = obj.physical_id
if not stack_id:
return True
ignore_missing = params.get('ignore_missing', True)
try:
self.orchestration(obj).stack_delete(stack_id, ignore_missing)
self.orchestration(obj).wait_for_stack_delete(stack_id)
except exc.InternalError as ex:
raise exc.EResourceDeletion(type='stack', id=stack_id,
message=six.text_type(ex))
return True
def do_update(self, obj, new_profile, **params):
"""Perform update on object.
:param obj: the node object to operate on
:param new_profile: the new profile used for updating
:param params: other parameters for the update request.
:returns: A boolean indicating whether the operation is successful.
"""
self.stack_id = obj.physical_id
if not self.stack_id:
return False
if not self.validate_for_update(new_profile):
return False
fields = {}
new_template = new_profile.properties[new_profile.TEMPLATE]
if new_template != self.properties[self.TEMPLATE]:
fields['template'] = new_template
new_params = new_profile.properties[new_profile.PARAMETERS]
if new_params != self.properties[self.PARAMETERS]:
fields['parameters'] = new_params
new_timeout = new_profile.properties[new_profile.TIMEOUT]
if new_timeout != self.properties[self.TIMEOUT]:
fields['timeout_mins'] = new_timeout
new_dr = new_profile.properties[new_profile.DISABLE_ROLLBACK]
if new_dr != self.properties[self.DISABLE_ROLLBACK]:
fields['disable_rollback'] = new_dr
new_files = new_profile.properties[new_profile.FILES]
if new_files != self.properties[self.FILES]:
fields['files'] = new_files
new_environment = new_profile.properties[new_profile.ENVIRONMENT]
if new_environment != self.properties[self.ENVIRONMENT]:
fields['environment'] = new_environment
if not fields:
return True
try:
hc = self.orchestration(obj)
# Timeout = None means we will use the 'default_action_timeout'
# It can be overridden by the TIMEOUT profile properties
timeout = None
if self.properties[self.TIMEOUT]:
timeout = self.properties[self.TIMEOUT] * 60
hc.stack_update(self.stack_id, **fields)
hc.wait_for_stack(self.stack_id, 'UPDATE_COMPLETE',
timeout=timeout)
except exc.InternalError as ex:
raise exc.EResourceUpdate(type='stack', id=self.stack_id,
message=six.text_type(ex))
return True
def do_check(self, obj):
"""Check stack status.
:param obj: Node object to operate.
:returns: True if check succeeded, or False otherwise.
"""
stack_id = obj.physical_id
if stack_id is None:
return False
hc = self.orchestration(obj)
try:
# Timeout = None means we will use the 'default_action_timeout'
# It can be overridden by the TIMEOUT profile properties
timeout = None
if self.properties[self.TIMEOUT]:
timeout = self.properties[self.TIMEOUT] * 60
hc.stack_check(stack_id)
hc.wait_for_stack(stack_id, 'CHECK_COMPLETE', timeout=timeout)
except exc.InternalError as ex:
raise exc.EResourceOperation(op='checking', type='stack',
id=stack_id,
message=six.text_type(ex))
return True
def do_get_details(self, obj):
if not obj.physical_id:
return {}
try:
stack = self.orchestration(obj).stack_get(obj.physical_id)
return stack.to_dict()
except exc.InternalError as ex:
return {
'Error': {
'code': ex.code,
'message': six.text_type(ex)
}
}
def do_adopt(self, obj, overrides=None, snapshot=False):
"""Adopt an existing stack node for management.
:param obj: A node object for this operation. It could be a puppet
node that provides only 'user', 'project' and 'physical_id'
properties when doing a preview. It can be a real Node object for
node adoption.
:param overrides: A dict containing the properties that will be
overridden when generating a profile for the stack.
:param snapshot: A boolean flag indicating whether the profile should
attempt a snapshot operation before adopting the stack. If set to
True, the ID of the snapshot will be used as the image ID.
:returns: A dict containing the spec created from the stack object or
a dict containing error information if failure occurred.
"""
driver = self.orchestration(obj)
# TODO(Qiming): Add snapshot support
# snapshot = driver.snapshot_create(...)
try:
stack = driver.stack_get(obj.physical_id)
tmpl = driver.stack_get_template(obj.physical_id)
env = driver.stack_get_environment(obj.physical_id)
files = driver.stack_get_files(obj.physical_id)
except exc.InternalError as ex:
return {'Error': {'code': ex.code, 'message': six.text_type(ex)}}
spec = {
self.ENVIRONMENT: env.to_dict(),
self.FILES: files,
self.TEMPLATE: tmpl.to_dict(),
self.PARAMETERS: dict((k, v) for k, v in stack.parameters.items()
if k.find('OS::', 0) < 0),
self.TIMEOUT: stack.timeout_mins,
self.DISABLE_ROLLBACK: stack.is_rollback_disabled
}
if overrides:
spec.update(overrides)
return spec
def _refresh_tags(self, current, node, add=False):
"""Refresh tag list.
:param current: Current list of tags.
:param node: The node object.
:param add: Flag indicating whether new tags are added.
:returns: (tags, updated) where tags contains a new list of tags and
updated indicates whether new tag list differs from the old
one.
"""
tags = []
for tag in current:
if tag.find('cluster_id=') == 0:
continue
elif tag.find('cluster_node_id=') == 0:
continue
elif tag.find('cluster_node_index=') == 0:
continue
if tag.strip() != "":
tags.append(tag.strip())
if add:
tags.append('cluster_id=' + node.cluster_id)
tags.append('cluster_node_id=' + node.id)
tags.append('cluster_node_index=%s' % node.index)
tag_str = ",".join(tags)
return (tag_str, tags != current)
def do_join(self, obj, cluster_id):
if not obj.physical_id:
return False
hc = self.orchestration(obj)
try:
stack = hc.stack_get(obj.physical_id)
tags, updated = self._refresh_tags(stack.tags, obj, True)
field = {'tags': tags}
if updated:
hc.stack_update(obj.physical_id, **field)
except exc.InternalError as ex:
LOG.error('Failed in updating stack tags: %s.', ex)
return False
return True
def do_leave(self, obj):
if not obj.physical_id:
return False
hc = self.orchestration(obj)
try:
stack = hc.stack_get(obj.physical_id)
tags, updated = self._refresh_tags(stack.tags, obj, False)
field = {'tags': tags}
if updated:
hc.stack_update(obj.physical_id, **field)
except exc.InternalError as ex:
LOG.error('Failed in updating stack tags: %s.', ex)
return False
return True
def handle_abandon(self, obj, **options):
"""Handler for abandoning a heat stack node."""
pass
def test_basic(self):
sot = schema.Boolean('desc')
self.assertEqual('Boolean', sot['type'])
self.assertEqual('desc', sot['description'])
class DeletionPolicy(base.Policy):
'''Policy for choosing victim node(s) from a cluster for deletion.
This policy is enforced when nodes are to be removed from a cluster.
It will yield an ordered list of candidates for deletion based on user
specified criteria.
'''
VERSION = '1.0'
KEYS = (
CRITERIA,
DESTROY_AFTER_DELETION,
GRACE_PERIOD,
REDUCE_DESIRED_CAPACITY,
) = (
'criteria',
'destroy_after_deletion',
'grace_period',
'reduce_desired_capacity',
)
CRITERIA_VALUES = (
OLDEST_FIRST,
OLDEST_PROFILE_FIRST,
YOUNGEST_FIRST,
RANDOM,
) = (
'OLDEST_FIRST',
'OLDEST_PROFILE_FRIST',
'YOUNGEST_FIRST',
'RANDOM',
)
TARGET = [
('BEFORE', consts.CLUSTER_SCALE_IN),
('BEFORE', consts.CLUSTER_DEL_NODES),
('BEFORE', consts.CLUSTER_RESIZE),
]
PROFILE_TYPE = ['ANY']
properties_schema = {
CRITERIA:
schema.String(_('Criteria used in selecting candidates for deletion'),
default=RANDOM,
constraints=[
constraints.AllowedValues(CRITERIA_VALUES),
]),
DESTROY_AFTER_DELETION:
schema.Boolean(
_('Whethere a node should be completely destroyed after '
'deletion. Default to True'),
default=True,
),
GRACE_PERIOD:
schema.Integer(
_('Number of seconds before real deletion happens.'),
default=0,
),
REDUCE_DESIRED_CAPACITY:
schema.Boolean(
_('Whether the desired capacity of the cluster should be '
'reduced along the deletion. Default to False.'),
default=False,
)
}
def __init__(self, name, spec, **kwargs):
super(DeletionPolicy, self).__init__(name, spec, **kwargs)
self.criteria = self.properties[self.CRITERIA]
self.grace_period = self.properties[self.GRACE_PERIOD]
self.destroy_after_deletion = self.properties[
self.DESTROY_AFTER_DELETION]
self.reduce_desired_capacity = self.properties[
self.REDUCE_DESIRED_CAPACITY]
random.seed()
def _select_candidates(self, context, cluster_id, count):
candidates = []
nodes = db_api.node_get_all_by_cluster(context, cluster_id)
if count > len(nodes):
count = len(nodes)
err_nodes = [n for n in nodes if n.status == 'ERROR']
nodes = [n for n in nodes if n.status != 'ERROR']
if count <= len(err_nodes):
return [n.id for n in err_nodes[:count]]
candidates.extend([n.id for n in err_nodes])
count -= len(err_nodes)
# Random selection
if self.criteria == self.RANDOM:
i = count
while i > 0:
rand = random.randrange(i)
candidates.append(nodes[rand].id)
nodes.remove(nodes[rand])
i = i - 1
return candidates
# Node age based selection
if self.criteria in [self.OLDEST_FIRST, self.YOUNGEST_FIRST]:
sorted_list = sorted(nodes, key=lambda r: (r.created_time, r.name))
for i in range(count):
if self.criteria == self.OLDEST_FIRST:
candidates.append(sorted_list[i].id)
else: # YOUNGEST_FIRST
candidates.append(sorted_list[-1 - i].id)
return candidates
# Node profile based selection
node_map = []
for node in nodes:
profile = db_api.profile_get(context, node.profile_id)
created_at = profile.created_time
node_map.append({'id': node.id, 'created_at': created_at})
sorted_map = sorted(node_map, key=lambda m: m['created_at'])
for i in range(count):
candidates.append(sorted_map[i]['id'])
return candidates
def pre_op(self, cluster_id, action):
'''Choose victims that can be deleted.'''
if action.action == consts.CLUSTER_RESIZE:
cluster = db_api.cluster_get(action.context, cluster_id)
scaleutils.parse_resize_params(action, cluster)
if 'deletion' not in action.data:
return
count = action.data['deletion']['count']
else: # CLUSTER_SCALE_IN or CLUSTER_DEL_NODES
count = action.inputs.get('count', 1)
pd = action.data.get('deletion', {})
candidates = pd.get('candidates', [])
# For certain operations ( e.g. DEL_NODES), the candidates might
# have been specified
if len(candidates) == 0:
candidates = self._select_candidates(action.context, cluster_id,
count)
pd['candidates'] = candidates
pd['destroy_after_deletion'] = self.destroy_after_deletion
pd['grace_period'] = self.grace_period
action.data.update({
'status': base.CHECK_OK,
'reason': _('Candidates generated'),
'deletion': pd
})
action.store(action.context)
return
class ServerProfile(base.Profile):
'''Profile for an OpenStack Nova server.'''
KEYS = (
CONTEXT,
ADMIN_PASS,
AUTO_DISK_CONFIG,
AVAILABILITY_ZONE,
BLOCK_DEVICE_MAPPING,
BLOCK_DEVICE_MAPPING_V2,
CONFIG_DRIVE,
FLAVOR,
IMAGE,
KEY_NAME,
METADATA,
NAME,
NETWORKS,
PERSONALITY,
SECURITY_GROUPS,
USER_DATA,
SCHEDULER_HINTS,
) = (
'context',
'adminPass',
'auto_disk_config',
'availability_zone',
'block_device_mapping',
'block_device_mapping_v2',
'config_drive',
'flavor',
'image',
'key_name',
'metadata',
'name',
'networks',
'personality',
'security_groups',
'user_data',
'scheduler_hints',
)
BDM_KEYS = (
BDM_DEVICE_NAME,
BDM_VOLUME_SIZE,
) = (
'device_name',
'volume_size',
)
BDM2_KEYS = (
BDM2_UUID,
BDM2_SOURCE_TYPE,
BDM2_DESTINATION_TYPE,
BDM2_DISK_BUS,
BDM2_DEVICE_NAME,
BDM2_VOLUME_SIZE,
BDM2_GUEST_FORMAT,
BDM2_BOOT_INDEX,
BDM2_DEVICE_TYPE,
BDM2_DELETE_ON_TERMINATION,
) = (
'uuid',
'source_type',
'destination_type',
'disk_bus',
'device_name',
'volume_size',
'guest_format',
'boot_index',
'device_type',
'delete_on_termination',
)
NETWORK_KEYS = (
PORT,
FIXED_IP,
NETWORK,
) = (
'port',
'fixed-ip',
'network',
)
PERSONALITY_KEYS = (
PATH,
CONTENTS,
) = (
'path',
'contents',
)
SCHEDULER_HINTS_KEYS = (GROUP, ) = ('group', )
properties_schema = {
CONTEXT:
schema.Map(_('Customized security context for operating servers.'), ),
ADMIN_PASS:
schema.String(_('Password for the administrator account.'), ),
AUTO_DISK_CONFIG:
schema.Boolean(
_('Whether the disk partition is done automatically.'),
default=True,
),
AVAILABILITY_ZONE:
schema.String(
_('Name of availability zone for running the server.'), ),
BLOCK_DEVICE_MAPPING:
schema.List(
_('A list specifying the properties of block devices to be used '
'for this server.'),
schema=schema.Map(
_('A map specifying the properties of a block device to be '
'used by the server.'),
schema={
BDM_DEVICE_NAME:
schema.String(
_('Block device name, should be <=255 chars.'), ),
BDM_VOLUME_SIZE:
schema.Integer(_('Block device size in GB.'), ),
}),
),
BLOCK_DEVICE_MAPPING_V2:
schema.List(
_('A list specifying the properties of block devices to be used '
'for this server.'),
schema=schema.Map(
_('A map specifying the properties of a block device to be '
'used by the server.'),
schema={
BDM2_UUID:
schema.String(
_('ID of the source image, snapshot or volume'), ),
BDM2_SOURCE_TYPE:
schema.String(
_('Volume source type, should be image, snapshot, '
'volume or blank'),
required=True,
),
BDM2_DESTINATION_TYPE:
schema.String(
_('Volume destination type, should be volume or '
'local'),
required=True,
),
BDM2_DISK_BUS:
schema.String(_('Bus of the device.'), ),
BDM2_DEVICE_NAME:
schema.String(
_('Name of the device(e.g. vda, xda, ....).'), ),
BDM2_VOLUME_SIZE:
schema.Integer(
_('Size of the block device in MB(for swap) and '
'in GB(for other formats)'),
required=True,
),
BDM2_GUEST_FORMAT:
schema.String(
_('Specifies the disk file system format(e.g. swap, '
'ephemeral, ...).'), ),
BDM2_BOOT_INDEX:
schema.Integer(_('Define the boot order of the device'), ),
BDM2_DEVICE_TYPE:
schema.String(
_('Type of the device(e.g. disk, cdrom, ...).'), ),
BDM2_DELETE_ON_TERMINATION:
schema.Boolean(
_('Whether to delete the volume when the server '
'stops.'), ),
}),
),
CONFIG_DRIVE:
schema.Boolean(
_('Whether config drive should be enabled for the server.'), ),
FLAVOR:
schema.String(
_('ID of flavor used for the server.'),
required=True,
updatable=True,
),
IMAGE:
schema.String(
# IMAGE is not required, because there could be BDM or BDMv2
# support and the corresponding settings effective
_('ID of image to be used for the new server.'),
updatable=True,
),
KEY_NAME:
schema.String(_('Name of Nova keypair to be injected to server.'), ),
METADATA:
schema.Map(
_('A collection of key/value pairs to be associated with the '
'server created. Both key and value should be <=255 chars.'),
updatable=True,
),
NAME:
schema.String(
_('Name of the server.'),
updatable=True,
),
NETWORKS:
schema.List(
_('List of networks for the server.'),
schema=schema.Map(
_('A map specifying the properties of a network for uses.'),
schema={
NETWORK:
schema.String(
_('Name or ID of network to create a port on.'), ),
PORT:
schema.String(_('Port ID to be used by the network.'), ),
FIXED_IP:
schema.String(_('Fixed IP to be used by the network.'), ),
},
),
updatable=True,
),
PERSONALITY:
schema.List(
_('List of files to be injected into the server, where each.'),
schema=schema.Map(
_('A map specifying the path & contents for an injected '
'file.'),
schema={
PATH:
schema.String(
_('In-instance path for the file to be injected.'),
required=True,
),
CONTENTS:
schema.String(
_('Contents of the file to be injected.'),
required=True,
),
},
),
),
SCHEDULER_HINTS:
schema.Map(
_('A collection of key/value pairs to be associated with the '
'Scheduler hints. Both key and value should be <=255 chars.'), ),
SECURITY_GROUPS:
schema.List(
_('List of security groups.'),
schema=schema.String(
_('Name of a security group'),
required=True,
),
),
USER_DATA:
schema.String(_('User data to be exposed by the metadata server.'), ),
}
OP_NAMES = (OP_REBOOT, ) = ('reboot', )
REBOOT_TYPE = 'type'
REBOOT_TYPES = (REBOOT_SOFT, REBOOT_HARD) = ('SOFT', 'HARD')
OPERATIONS = {
OP_REBOOT:
schema.Operation(
_("Reboot the nova server."),
schema={
REBOOT_TYPE:
schema.String(
_("Type of reboot which can be 'SOFT' or 'HARD'."),
default=REBOOT_SOFT,
constraints=[
constraints.AllowedValues(REBOOT_TYPES),
])
})
}
def __init__(self, type_name, name, **kwargs):
super(ServerProfile, self).__init__(type_name, name, **kwargs)
self._novaclient = None
self._neutronclient = None
self.server_id = None
def nova(self, obj):
'''Construct nova client based on object.
:param obj: Object for which the client is created. It is expected to
be None when retrieving an existing client. When creating
a client, it contains the user and project to be used.
'''
if self._novaclient is not None:
return self._novaclient
params = self._build_conn_params(obj.user, obj.project)
self._novaclient = driver_base.SenlinDriver().compute(params)
return self._novaclient
def neutron(self, obj):
'''Construct neutron client based on object.
:param obj: Object for which the client is created. It is expected to
be None when retrieving an existing client. When creating
a client, it contains the user and project to be used.
'''
if self._neutronclient is not None:
return self._neutronclient
params = self._build_conn_params(obj.user, obj.project)
self._neutronclient = driver_base.SenlinDriver().network(params)
return self._neutronclient
def do_validate(self, obj):
'''Validate if the spec has provided valid info for server creation.'''
return True
def _resolve_bdm(self, bdm):
for bd in bdm:
for key in self.BDM2_KEYS:
if bd[key] is None:
del bd[key]
return bdm
def _resolve_network(self, networks, client):
for network in networks:
net_name_id = network.get(self.NETWORK)
if net_name_id:
res = client.network_get(net_name_id)
network['uuid'] = res.id
del network[self.NETWORK]
if network['port'] is None:
del network['port']
if network['fixed-ip'] is None:
del network['fixed-ip']
return networks
def do_create(self, obj):
'''Create a server using the given profile.'''
kwargs = {}
for key in self.KEYS:
# context is treated as connection parameters
if key == self.CONTEXT:
continue
if self.properties[key] is not None:
kwargs[key] = self.properties[key]
name_or_id = self.properties[self.IMAGE]
if name_or_id is not None:
image = self.nova(obj).image_find(name_or_id)
# wait for new version of openstacksdk to fix this
kwargs.pop(self.IMAGE)
kwargs['imageRef'] = image.id
flavor_id = self.properties[self.FLAVOR]
flavor = self.nova(obj).flavor_find(flavor_id, False)
# wait for new verson of openstacksdk to fix this
kwargs.pop(self.FLAVOR)
kwargs['flavorRef'] = flavor.id
name = self.properties[self.NAME]
if name:
kwargs['name'] = name
else:
kwargs['name'] = obj.name
metadata = self.properties[self.METADATA] or {}
if obj.cluster_id:
metadata['cluster'] = obj.cluster_id
kwargs['metadata'] = metadata
block_device_mapping_v2 = self.properties[self.BLOCK_DEVICE_MAPPING_V2]
if block_device_mapping_v2 is not None:
kwargs['block_device_mapping_v2'] = self._resolve_bdm(
block_device_mapping_v2)
user_data = self.properties[self.USER_DATA]
if user_data is not None:
ud = encodeutils.safe_encode(user_data)
kwargs['user_data'] = encodeutils.safe_decode(base64.b64encode(ud))
networks = self.properties[self.NETWORKS]
if networks is not None:
kwargs['networks'] = self._resolve_network(networks,
self.neutron(obj))
secgroups = self.properties[self.SECURITY_GROUPS]
if secgroups:
kwargs['security_groups'] = [{'name': sg} for sg in secgroups]
if 'placement' in obj.data:
if 'zone' in obj.data['placement']:
kwargs['availability_zone'] = obj.data['placement']['zone']
if 'servergroup' in obj.data['placement']:
group_id = obj.data['placement']['servergroup']
hints = self.properties.get(self.SCHEDULER_HINTS, {})
hints.update({'group': group_id})
kwargs['scheduler_hints'] = hints
LOG.info('Creating server: %s' % kwargs)
server = self.nova(obj).server_create(**kwargs)
self.nova(obj).wait_for_server(server.id)
self.server_id = server.id
return server.id
def do_delete(self, obj):
self.server_id = obj.physical_id
if not obj.physical_id:
return True
try:
self.nova(obj).server_delete(self.server_id)
self.nova(obj).wait_for_server_delete(self.server_id)
except Exception as ex:
LOG.error('Error: %s' % six.text_type(ex))
return False
return True
def do_update(self, obj, new_profile=None, **params):
'''Perform update on the server.
:param obj: the server to operate on
:param new_profile: the new profile for the server.
:param params: a dictionary of optional parameters.
'''
self.server_id = obj.physical_id
if not self.server_id:
return True
if not new_profile:
return True
if not self.validate_for_update(new_profile):
return False
# TODO(Yanyan Hu): Update block_device properties
# Update basic properties of server
if not self._update_basic_properties(obj, new_profile):
return False
# Update server flavor
flavor = self.properties[self.FLAVOR]
new_flavor = new_profile.properties[self.FLAVOR]
if new_flavor != flavor:
try:
self._update_flavor(obj, flavor, new_flavor)
except Exception as ex:
LOG.exception(_('Failed in updating server flavor: %s'),
six.text_type(ex))
return False
# Update server image
old_passwd = self.properties.get(self.ADMIN_PASS)
passwd = old_passwd
if new_profile.properties[self.ADMIN_PASS] is not None:
passwd = new_profile.properties[self.ADMIN_PASS]
image = self.properties[self.IMAGE]
new_image = new_profile.properties[self.IMAGE]
if new_image != image:
try:
self._update_image(obj, image, new_image, passwd)
except Exception as ex:
LOG.exception(_('Failed in updating server image: %s'),
six.text_type(ex))
return False
elif old_passwd != passwd:
# TODO(Jun Xu): update server admin password
pass
# Update server network
networks_current = self.properties[self.NETWORKS]
networks_create = new_profile.properties[self.NETWORKS]
networks_delete = copy.deepcopy(networks_current)
for network in networks_current:
if network in networks_create:
networks_create.remove(network)
networks_delete.remove(network)
if networks_create or networks_delete:
# We have network interfaces to be deleted and/or created
try:
self._update_network(obj, networks_create, networks_delete)
except Exception as ex:
LOG.exception(_('Failed in updating server network: %s'),
six.text_type(ex))
return False
return True
def _update_basic_properties(self, obj, new_profile):
'''Updating basic server properties including name, metadata'''
# Update server metadata
metadata = self.properties[self.METADATA]
new_metadata = new_profile.properties[self.METADATA]
if new_metadata != metadata:
if new_metadata is None:
new_metadata = {}
try:
self.nova(obj).server_metadata_update(self.server_id,
new_metadata)
except Exception as ex:
LOG.exception(_('Failed in updating server metadata: %s'),
six.text_type(ex))
return False
# Update server name
name = self.properties[self.NAME]
new_name = new_profile.properties[self.NAME]
if new_name != name:
attrs = {'name': new_name if new_name else obj.name}
try:
self.nova(obj).server_update(self.server_id, **attrs)
except Exception as ex:
LOG.exception(_('Failed in updating server name: %s'),
six.text_type(ex))
return False
return True
def _update_flavor(self, obj, old_flavor, new_flavor):
'''Updating server flavor'''
res = self.nova(obj).flavor_find(old_flavor)
old_flavor_id = res.id
res = self.nova(obj).flavor_find(new_flavor)
new_flavor_id = res.id
if new_flavor_id == old_flavor_id:
return
try:
self.nova(obj).server_resize(obj.physical_id, new_flavor_id)
self.nova(obj).wait_for_server(obj.physical_id, 'VERIFY_RESIZE')
except Exception as ex:
LOG.error(_("Server resizing failed, revert it: %s"),
six.text_type(ex))
self.nova(obj).server_resize_revert(obj.physical_id)
self.nova(obj).wait_for_server(obj.physical_id, 'ACTIVE')
raise exception.ResourceUpdateFailure(resource=obj.physical_id)
self.nova(obj).server_resize_confirm(obj.physical_id)
self.nova(obj).wait_for_server(obj.physical_id, 'ACTIVE')
def _update_image(self, obj, old_image, new_image, admin_password):
'''Updating server image'''
if old_image:
res = self.nova(obj).image_find(old_image)
image_id = res.id
else:
server = self.nova(obj).server_get(obj.physical_id)
image_id = server.image['id']
if new_image:
res = self.nova(obj).image_find(new_image)
new_image_id = res.id
if new_image_id != image_id:
# (Jun Xu): Not update name here if name changed,
# it should be updated in do_update
self.nova(obj).server_rebuild(obj.physical_id, new_image_id,
self.properties.get(self.NAME),
admin_password)
self.nova(obj).wait_for_server(obj.physical_id, 'ACTIVE')
else:
# TODO(Yanyan Hu): Allow server update with new_image
# set to None if Nova service supports it
LOG.error(
_("Updating Nova server with image set to None is "
"not supported by Nova."))
raise exception.ResourceUpdateFailure(resource=obj.physical_id)
def _update_network(self, obj, networks_create, networks_delete):
'''Updating server network interfaces'''
server = self.nova(obj).server_get(self.server_id)
ports_existing = list(self.nova(obj).server_interface_list(server))
ports = []
for p in ports_existing:
fixed_ips = []
for addr in p['fixed_ips']:
fixed_ips.append(addr['ip_address'])
ports.append({
'port_id': p['port_id'],
'net_id': p['net_id'],
'fixed_ips': fixed_ips
})
# Detach some existing ports
# Step1. Accurately search port with port_id or fixed-ip/net_id
for n in networks_delete:
if n['port'] is not None:
for p in ports:
if p['port_id'] == n['port']:
ports.remove(p)
break
res = self.nova(obj).server_interface_delete(n['port'], server)
elif n['fixed-ip'] is not None:
res = self.neutron(obj).network_get(n['network'])
net_id = res.id
for p in ports:
if (n['fixed-ip'] in p['fixed_ips']) and (p['net_id']
== net_id):
res = self.nova(obj).server_interface_delete(
p['port_id'], server)
ports.remove(p)
break
# Step2. Fuzzy search port with net_id
for n in networks_delete:
if n['port'] is None and n['fixed-ip'] is None:
res = self.neutron(obj).network_get(n['network'])
net_id = res.id
for p in ports:
if p['net_id'] == net_id:
res = self.nova(obj).server_interface_delete(
p['port_id'], server)
ports.remove(p)
break
# Attach new ports added in new network definition
for n in networks_create:
net_name_id = n.get(self.NETWORK, None)
if net_name_id:
res = self.neutron(obj).network_get(net_name_id)
n['net_id'] = res.id
if n['fixed-ip'] is not None:
n['fixed_ips'] = [{'ip_address': n['fixed-ip']}]
if n['port'] is not None:
n['port_id'] = n['port']
del n['network']
del n['port']
del n['fixed-ip']
self.nova(obj).server_interface_create(server, **n)
return
def do_check(self, obj):
if not obj.physical_id:
return False
self.server_id = obj.physical_id
try:
server = self.nova(obj).server_get(self.server_id)
except Exception as ex:
LOG.error('Error: %s' % six.text_type(ex))
return False
if (server is None or server.status != 'ACTIVE'):
return False
return True
def do_get_details(self, obj):
known_keys = {
'OS-DCF:diskConfig', 'OS-EXT-AZ:availability_zone',
'OS-EXT-STS:power_state', 'OS-EXT-STS:vm_state', 'accessIPv4',
'accessIPv6', 'config_drive', 'created', 'hostId', 'id',
'key_name', 'locked', 'metadata', 'name',
'os-extended-volumes:volumes_attached', 'progress', 'status',
'updated'
}
if obj.physical_id is None or obj.physical_id == '':
return {}
try:
server = self.nova(obj).server_get(obj.physical_id)
except exception.InternalError as ex:
return {'Error': {'code': ex.code, 'message': six.text_type(ex)}}
if server is None:
return {}
server_data = server.to_dict()
details = {
'image': server_data['image']['id'],
'flavor': server_data['flavor']['id'],
}
for key in known_keys:
if key in server_data:
details[key] = server_data[key]
# process special keys like 'OS-EXT-STS:task_state': these keys have
# a default value '-' when not existing
special_keys = [
'OS-EXT-STS:task_state',
'OS-SRV-USG:launched_at',
'OS-SRV-USG:terminated_at',
]
for key in special_keys:
if key in server_data:
val = server_data[key]
details[key] = val if val else '-'
# process network addresses
details['addresses'] = {}
for net in server_data['addresses']:
addresses = []
for addr in server_data['addresses'][net]:
# Ignore IPv6 address
if addr['version'] == 4:
addresses.append(addr['addr'])
details['addresses'][net] = addresses
# process security groups
sgroups = []
if 'security_groups' in server_data:
for sg in server_data['security_groups']:
sgroups.append(sg['name'])
if len(sgroups) == 0:
details['security_groups'] = ''
elif len(sgroups) == 1:
details['security_groups'] = sgroups[0]
else:
details['security_groups'] = sgroups
return dict((k, details[k]) for k in sorted(details))
def do_join(self, obj, cluster_id):
if not obj.physical_id:
return False
metadata = self.nova(obj).server_metadata_get(obj.physical_id) or {}
metadata['cluster'] = cluster_id
self.nova(obj).server_metadata_update(obj.physical_id, metadata)
return super(ServerProfile, self).do_join(obj, cluster_id)
def do_leave(self, obj):
if not obj.physical_id:
return False
self.nova(obj).server_metadata_delete(obj.physical_id, ['cluster'])
return super(ServerProfile, self).do_leave(obj)
def do_rebuild(self, obj):
if not obj.physical_id:
return False
self.server_id = obj.physical_id
try:
server = self.nova(obj).server_get(self.server_id)
except Exception as ex:
LOG.exception(_('Failed at getting server: %s'), six.text_type(ex))
return False
if server is None or server.image is None:
return False
image_id = server.image['id']
admin_pass = self.properties.get(self.ADMIN_PASS)
try:
self.nova(obj).server_rebuild(self.server_id, image_id,
self.properties.get(self.NAME),
admin_pass)
self.nova(obj).wait_for_server(self.server_id, 'ACTIVE')
except Exception as ex:
LOG.exception(_('Failed at rebuilding server: %s'),
six.text_type(ex))
return False
return True
def do_recover(self, obj, **options):
if 'operation' in options:
if options['operation'] == 'REBUILD':
return self.do_rebuild(obj)
res = super(ServerProfile, self).do_recover(obj, **options)
return res
def handle_reboot(self, obj, **options):
"""Handler for the reboot operation."""
pass
class ScalingPolicy(base.Policy):
'''Policy for changing the size of a cluster.
This policy is expected to be enforced before the node count of a cluster
is changed.
'''
VERSION = '1.0'
TARGET = [
('BEFORE', consts.CLUSTER_SCALE_IN),
('BEFORE', consts.CLUSTER_SCALE_OUT),
]
PROFILE_TYPE = [
'ANY',
]
KEYS = (
EVENT,
ADJUSTMENT,
) = (
'event',
'adjustment',
)
_SUPPORTED_EVENTS = (
CLUSTER_SCALE_IN,
CLUSTER_SCALE_OUT,
) = (
consts.CLUSTER_SCALE_IN,
consts.CLUSTER_SCALE_OUT,
)
_ADJUSTMENT_KEYS = (
ADJUSTMENT_TYPE,
ADJUSTMENT_NUMBER,
MIN_STEP,
BEST_EFFORT,
) = (
'type',
'number',
'min_step',
'best_effort',
)
properties_schema = {
EVENT:
schema.String(
_('Event that will trigger this policy. Must be one of '
'CLUSTER_SCALE_IN and CLUSTER_SCALE_OUT.'),
constraints=[
constraints.AllowedValues(_SUPPORTED_EVENTS),
],
required=True,
),
ADJUSTMENT:
schema.Map(
_('Detailed specification for scaling adjustments.'),
schema={
ADJUSTMENT_TYPE:
schema.String(
_('Type of adjustment when scaling is triggered.'),
constraints=[
constraints.AllowedValues(consts.ADJUSTMENT_TYPES),
],
default=consts.CHANGE_IN_CAPACITY,
),
ADJUSTMENT_NUMBER:
schema.Number(
_('A number specifying the amount of adjustment.'),
default=1,
),
MIN_STEP:
schema.Integer(
_('When adjustment type is set to "CHANGE_IN_PERCENTAGE",'
' this specifies the cluster size will be decreased by '
'at least this number of nodes.'),
default=1,
),
BEST_EFFORT:
schema.Boolean(
_('Whether do best effort scaling when new size of '
'cluster will break the size limitation'),
default=False,
),
}),
}
def __init__(self, name, spec, **kwargs):
super(ScalingPolicy, self).__init__(name, spec, **kwargs)
self.event = self.properties[self.EVENT]
self.singleton = False
adjustment = self.properties[self.ADJUSTMENT]
self.adjustment_type = adjustment[self.ADJUSTMENT_TYPE]
self.adjustment_number = adjustment[self.ADJUSTMENT_NUMBER]
self.adjustment_min_step = adjustment[self.MIN_STEP]
self.best_effort = adjustment[self.BEST_EFFORT]
def _calculate_adjustment_count(self, current_size):
'''Calculate adjustment count based on current_size'''
if self.adjustment_type == consts.EXACT_CAPACITY:
if self.event == consts.CLUSTER_SCALE_IN:
count = current_size - self.adjustment_number
else:
count = self.adjustment_number - current_size
elif self.adjustment_type == consts.CHANGE_IN_CAPACITY:
count = self.adjustment_number
else: # consts.CHANGE_IN_PERCENTAGE:
count = int((self.adjustment_number * current_size) / 100.0)
if count < self.adjustment_min_step:
count = self.adjustment_min_step
return count
def pre_op(self, cluster_id, action):
status = base.CHECK_OK
reason = _('Scaling request validated.')
# Check if the action is expected by the policy
if self.event != action.action:
action.data.update({'status': status, 'reason': reason})
action.store(action.context)
return
cluster = db_api.cluster_get(action.context, cluster_id)
nodes = db_api.node_get_all_by_cluster(action.context, cluster_id)
current_size = len(nodes)
count = self._calculate_adjustment_count(current_size)
# Use action input if count is provided
count = action.inputs.get('count', count)
if count <= 0:
status = base.CHECK_ERROR
reason = _("Count (%(count)s) invalid for action %(action)s.") % {
'count': count,
'action': action.action
}
# Check size constraints
if action.action == consts.CLUSTER_SCALE_IN:
new_size = current_size - count
if (new_size < cluster.min_size):
if self.best_effort:
count = current_size - cluster.min_size
reason = _('Do best effort scaling.')
else:
status = base.CHECK_ERROR
reason = _('Attempted scaling below minimum size.')
else:
new_size = current_size + count
if (new_size > cluster.max_size):
if self.best_effort:
count = cluster.max_size - current_size
reason = _('Do best effort scaling.')
else:
status = base.CHECK_ERROR
reason = _('Attempted scaling above maximum size.')
pd = {'status': status, 'reason': reason}
if status == base.CHECK_OK:
if action.action == consts.CLUSTER_SCALE_IN:
pd['deletion'] = {'count': count}
else:
pd['creation'] = {'count': count}
action.data.update(pd)
action.store(action.context)
return
class AffinityPolicy(base.Policy):
"""Policy for placing members of a cluster based on server groups.
This policy is expected to be enforced before new member(s) added to an
existing cluster.
"""
VERSION = '1.0'
VERSIONS = {'1.0': [{'status': consts.SUPPORTED, 'since': '2016.10'}]}
PRIORITY = 300
TARGET = [
('BEFORE', consts.CLUSTER_SCALE_OUT),
('BEFORE', consts.CLUSTER_RESIZE),
('BEFORE', consts.NODE_CREATE),
]
PROFILE_TYPE = [
'os.nova.server-1.0',
]
KEYS = (
SERVER_GROUP,
AVAILABILITY_ZONE,
ENABLE_DRS_EXTENSION,
) = (
'servergroup',
'availability_zone',
'enable_drs_extension',
)
_GROUP_KEYS = (
GROUP_NAME,
GROUP_POLICIES,
) = (
'name',
'policies',
)
_POLICIES_VALUES = (
# NOTE: soft policies are supported from compute micro version 2.15
AFFINITY,
SOFT_AFFINITY,
ANTI_AFFINITY,
SOFT_ANTI_AFFINITY,
) = (
'affinity',
'soft-affinity',
'anti-affinity',
'soft-anti-affinity',
)
properties_schema = {
SERVER_GROUP:
schema.Map(
_('Properties of the VM server group'),
schema={
GROUP_NAME:
schema.String(_('The name of the server group'), ),
GROUP_POLICIES:
schema.String(
_('The server group policies.'),
default=ANTI_AFFINITY,
constraints=[
constraints.AllowedValues(_POLICIES_VALUES),
],
),
},
),
AVAILABILITY_ZONE:
schema.String(
_('Name of the availability zone to place the nodes.'), ),
ENABLE_DRS_EXTENSION:
schema.Boolean(
_('Enable vSphere DRS extension.'),
default=False,
),
}
def __init__(self, name, spec, **kwargs):
super(AffinityPolicy, self).__init__(name, spec, **kwargs)
self.enable_drs = self.properties.get(self.ENABLE_DRS_EXTENSION)
def validate(self, context, validate_props=False):
super(AffinityPolicy, self).validate(context, validate_props)
if not validate_props:
return True
az_name = self.properties.get(self.AVAILABILITY_ZONE)
if az_name:
nc = self.nova(context.user_id, context.project_id)
valid_azs = nc.validate_azs([az_name])
if not valid_azs:
msg = _("The specified %(key)s '%(value)s' could not be "
"found.") % {
'key': self.AVAILABILITY_ZONE,
'value': az_name
}
raise exc.InvalidSpec(message=msg)
return True
def attach(self, cluster, enabled=True):
"""Routine to be invoked when policy is to be attached to a cluster.
:para cluster: The cluster to which the policy is being attached to.
:param enabled: The attached cluster policy is enabled or disabled.
:returns: When the operation was successful, returns a tuple (True,
message); otherwise, return a tuple (False, error).
"""
res, data = super(AffinityPolicy, self).attach(cluster)
if res is False:
return False, data
data = {'inherited_group': False}
nc = self.nova(cluster.user, cluster.project)
group = self.properties.get(self.SERVER_GROUP)
# guess servergroup name
group_name = group.get(self.GROUP_NAME, None)
if group_name is None:
profile = cluster.rt['profile']
if 'scheduler_hints' in profile.spec:
hints = profile.spec['scheduler_hints']
group_name = hints.get('group', None)
if group_name:
try:
server_group = nc.server_group_find(group_name, True)
except exc.InternalError as ex:
msg = _("Failed in retrieving servergroup '%s'.") % group_name
LOG.exception('%(msg)s: %(ex)s', {'msg': msg, 'ex': ex})
return False, msg
if server_group:
# Check if the policies match
policies = group.get(self.GROUP_POLICIES)
if policies and policies != server_group.policies[0]:
msg = _(
"Policies specified (%(specified)s) doesn't match "
"that of the existing servergroup (%(existing)s).") % {
'specified': policies,
'existing': server_group.policies[0]
}
return False, msg
data['servergroup_id'] = server_group.id
data['inherited_group'] = True
if not data['inherited_group']:
# create a random name if necessary
if not group_name:
group_name = 'server_group_%s' % utils.random_name()
try:
server_group = nc.server_group_create(
name=group_name, policies=[group.get(self.GROUP_POLICIES)])
except Exception as ex:
msg = _('Failed in creating servergroup.')
LOG.exception('%(msg)s: %(ex)s', {'msg': msg, 'ex': ex})
return False, msg
data['servergroup_id'] = server_group.id
policy_data = self._build_policy_data(data)
return True, policy_data
def detach(self, cluster):
"""Routine to be called when the policy is detached from a cluster.
:param cluster: The cluster from which the policy is to be detached.
:returns: When the operation was successful, returns a tuple of
(True, data) where the data contains references to the
resources created; otherwise returns a tuple of (False,
error) where the err contains an error message.
"""
reason = _('Servergroup resource deletion succeeded.')
ctx = context.get_admin_context()
binding = cpo.ClusterPolicy.get(ctx, cluster.id, self.id)
if not binding or not binding.data:
return True, reason
policy_data = self._extract_policy_data(binding.data)
if not policy_data:
return True, reason
group_id = policy_data.get('servergroup_id', None)
inherited_group = policy_data.get('inherited_group', False)
if group_id and not inherited_group:
try:
nc = self.nova(cluster.user, cluster.project)
nc.server_group_delete(group_id)
except Exception as ex:
msg = _('Failed in deleting servergroup.')
LOG.exception('%(msg)s: %(ex)s', {'msg': msg, 'ex': ex})
return False, msg
return True, reason
def pre_op(self, cluster_id, action):
"""Routine to be called before target action is executed.
This policy annotates the node with a server group ID before the
node is actually created. For vSphere DRS, it is equivalent to the
selection of vSphere host (cluster).
:param cluster_id: ID of the cluster on which the relevant action
is to be executed.
:param action: The action object that triggered this operation.
:returns: Nothing.
"""
zone_name = self.properties.get(self.AVAILABILITY_ZONE)
if not zone_name and self.enable_drs:
# we make a reasonable guess of the zone name for vSphere
# support because the zone name is required in that case.
zone_name = 'nova'
# we respect other policies decisions (if any) and fall back to the
# action inputs if no hints found.
pd = action.data.get('creation', None)
if pd is not None:
count = pd.get('count', 1)
elif action.action == consts.CLUSTER_SCALE_OUT:
count = action.inputs.get('count', 1)
elif action.action == consts.NODE_CREATE:
count = 1
else: # CLUSTER_RESIZE
cluster = action.entity
current = len(cluster.nodes)
su.parse_resize_params(action, cluster, current)
if 'creation' not in action.data:
return
count = action.data['creation']['count']
cp = cpo.ClusterPolicy.get(action.context, cluster_id, self.id)
policy_data = self._extract_policy_data(cp.data)
pd_entry = {'servergroup': policy_data['servergroup_id']}
# special handling for vSphere DRS case where we need to find out
# the name of the vSphere host which has DRS enabled.
if self.enable_drs:
obj = action.entity
nc = self.nova(obj.user, obj.project)
hypervisors = nc.hypervisor_list()
hv_id = ''
pattern = re.compile(r'.*drs*', re.I)
for hypervisor in hypervisors:
match = pattern.match(hypervisor.hypervisor_hostname)
if match:
hv_id = hypervisor.id
break
if not hv_id:
action.data['status'] = base.CHECK_ERROR
action.data['status_reason'] = _('No suitable vSphere host '
'is available.')
action.store(action.context)
return
hv_info = nc.hypervisor_get(hv_id)
hostname = hv_info['service']['host']
pd_entry['zone'] = ":".join([zone_name, hostname])
elif zone_name:
pd_entry['zone'] = zone_name
pd = {
'count': count,
'placements': [pd_entry] * count,
}
action.data.update({'placement': pd})
action.store(action.context)
return
class ScalingPolicy(base.Policy):
"""Policy for changing the size of a cluster.
This policy is expected to be enforced before the node count of a cluster
is changed.
"""
VERSION = '1.0'
VERSIONS = {'1.0': [{'status': consts.SUPPORTED, 'since': '2016.04'}]}
PRIORITY = 100
TARGET = [
('BEFORE', consts.CLUSTER_SCALE_IN),
('BEFORE', consts.CLUSTER_SCALE_OUT),
('AFTER', consts.CLUSTER_SCALE_IN),
('AFTER', consts.CLUSTER_SCALE_OUT),
]
PROFILE_TYPE = [
'ANY',
]
KEYS = (
EVENT,
ADJUSTMENT,
) = (
'event',
'adjustment',
)
_SUPPORTED_EVENTS = (
CLUSTER_SCALE_IN,
CLUSTER_SCALE_OUT,
) = (
consts.CLUSTER_SCALE_IN,
consts.CLUSTER_SCALE_OUT,
)
_ADJUSTMENT_KEYS = (
ADJUSTMENT_TYPE,
ADJUSTMENT_NUMBER,
MIN_STEP,
BEST_EFFORT,
COOLDOWN,
) = (
'type',
'number',
'min_step',
'best_effort',
'cooldown',
)
properties_schema = {
EVENT:
schema.String(
_('Event that will trigger this policy. Must be one of '
'CLUSTER_SCALE_IN and CLUSTER_SCALE_OUT.'),
constraints=[
constraints.AllowedValues(_SUPPORTED_EVENTS),
],
required=True,
),
ADJUSTMENT:
schema.Map(
_('Detailed specification for scaling adjustments.'),
schema={
ADJUSTMENT_TYPE:
schema.String(
_('Type of adjustment when scaling is triggered.'),
constraints=[
constraints.AllowedValues(consts.ADJUSTMENT_TYPES),
],
default=consts.CHANGE_IN_CAPACITY,
),
ADJUSTMENT_NUMBER:
schema.Number(
_('A number specifying the amount of adjustment.'),
default=1,
),
MIN_STEP:
schema.Integer(
_('When adjustment type is set to "CHANGE_IN_PERCENTAGE",'
' this specifies the cluster size will be decreased by '
'at least this number of nodes.'),
default=1,
),
BEST_EFFORT:
schema.Boolean(
_('Whether do best effort scaling when new size of '
'cluster will break the size limitation'),
default=False,
),
COOLDOWN:
schema.Integer(
_('Number of seconds to hold the cluster for cool-down '
'before allowing cluster to be resized again.'),
default=0,
),
}),
}
def __init__(self, name, spec, **kwargs):
"""Initialize a scaling policy object.
:param name: Name for the policy object.
:param spec: A dictionary containing the detailed specification for
the policy.
:param dict kwargs: Other optional parameters for policy object
creation.
:return: An object of `ScalingPolicy`.
"""
super(ScalingPolicy, self).__init__(name, spec, **kwargs)
self.singleton = False
self.event = self.properties[self.EVENT]
adjustment = self.properties[self.ADJUSTMENT]
self.adjustment_type = adjustment[self.ADJUSTMENT_TYPE]
self.adjustment_number = adjustment[self.ADJUSTMENT_NUMBER]
self.adjustment_min_step = adjustment[self.MIN_STEP]
self.best_effort = adjustment[self.BEST_EFFORT]
self.cooldown = adjustment[self.COOLDOWN]
def validate(self, context, validate_props=False):
super(ScalingPolicy, self).validate(context, validate_props)
if self.adjustment_number <= 0:
msg = _("the 'number' for 'adjustment' must be > 0")
raise exc.InvalidSpec(message=msg)
if self.adjustment_min_step < 0:
msg = _("the 'min_step' for 'adjustment' must be >= 0")
raise exc.InvalidSpec(message=msg)
if self.cooldown < 0:
msg = _("the 'cooldown' for 'adjustment' must be >= 0")
raise exc.InvalidSpec(message=msg)
def _calculate_adjustment_count(self, current_size):
"""Calculate adjustment count based on current_size.
:param current_size: The current size of the target cluster.
:return: The number of nodes to add or to remove.
"""
if self.adjustment_type == consts.EXACT_CAPACITY:
if self.event == consts.CLUSTER_SCALE_IN:
count = current_size - self.adjustment_number
else:
count = self.adjustment_number - current_size
elif self.adjustment_type == consts.CHANGE_IN_CAPACITY:
count = self.adjustment_number
else: # consts.CHANGE_IN_PERCENTAGE:
count = int((self.adjustment_number * current_size) / 100.0)
if count < self.adjustment_min_step:
count = self.adjustment_min_step
return count
def pre_op(self, cluster_id, action):
"""The hook function that is executed before the action.
The checking result is stored in the ``data`` property of the action
object rather than returned directly from the function.
:param cluster_id: The ID of the target cluster.
:param action: Action instance against which the policy is being
checked.
:return: None.
"""
# check cooldown
last_op = action.inputs.get('last_op', None)
if last_op and not timeutils.is_older_than(last_op, self.cooldown):
action.data.update({
'status':
base.CHECK_ERROR,
'reason':
_('Policy %s cooldown is still '
'in progress.') % self.id
})
action.store(action.context)
return
# Use action input if count is provided
count_value = action.inputs.get('count', None)
cluster = action.entity
current = len(cluster.nodes)
if count_value is None:
# count not specified, calculate it
count_value = self._calculate_adjustment_count(current)
# Count must be positive value
success, count = utils.get_positive_int(count_value)
if not success:
action.data.update({
'status': base.CHECK_ERROR,
'reason': _("Invalid count (%(c)s) for action '%(a)s'.") % {
'c': count_value,
'a': action.action
}
})
action.store(action.context)
return
# Check size constraints
max_size = cluster.max_size
if max_size == -1:
max_size = cfg.CONF.max_nodes_per_cluster
if action.action == consts.CLUSTER_SCALE_IN:
if self.best_effort:
count = min(count, current - cluster.min_size)
result = su.check_size_params(cluster,
current - count,
strict=not self.best_effort)
else:
if self.best_effort:
count = min(count, max_size - current)
result = su.check_size_params(cluster,
current + count,
strict=not self.best_effort)
if result:
# failed validation
pd = {'status': base.CHECK_ERROR, 'reason': result}
else:
# passed validation
pd = {
'status': base.CHECK_OK,
'reason': _('Scaling request validated.'),
}
if action.action == consts.CLUSTER_SCALE_IN:
pd['deletion'] = {'count': count}
else:
pd['creation'] = {'count': count}
action.data.update(pd)
action.store(action.context)
return
def post_op(self, cluster_id, action):
# update last_op for next cooldown check
ts = timeutils.utcnow(True)
cpo.ClusterPolicy.update(action.context, cluster_id, self.id,
{'last_op': ts})
def need_check(self, target, action):
# check if target + action matches policy targets
if not super(ScalingPolicy, self).need_check(target, action):
return False
if target == 'BEFORE':
# Scaling policy BEFORE check should only be triggered if the
# incoming action matches the specific policy event.
# E.g. for scale-out policy the BEFORE check to select nodes for
# termination should only run for scale-out actions.
return self.event == action.action
else:
# Scaling policy AFTER check to reset cooldown timer should be
# triggered for all supported policy events (both scale-in and
# scale-out). E.g. a scale-out policy should reset cooldown timer
# whenever scale-out or scale-in action completes.
return action.action in list(self._SUPPORTED_EVENTS)
class LoadBalancingPolicy(base.Policy):
"""Policy for load balancing among members of a cluster.
This policy is expected to be enforced before or after the membership of a
cluster is changed. We need to refresh the load-balancer associated with
the cluster (which could be created by the policy) when these actions are
performed.
"""
VERSION = '1.1'
VERSIONS = {'1.0': [{'status': consts.SUPPORTED, 'since': '2016.04'}]}
PRIORITY = 500
TARGET = [
('AFTER', consts.CLUSTER_ADD_NODES),
('AFTER', consts.CLUSTER_SCALE_OUT),
('AFTER', consts.CLUSTER_RESIZE),
('AFTER', consts.NODE_RECOVER),
('AFTER', consts.NODE_CREATE),
('BEFORE', consts.CLUSTER_DEL_NODES),
('BEFORE', consts.CLUSTER_SCALE_IN),
('BEFORE', consts.CLUSTER_RESIZE),
('BEFORE', consts.NODE_DELETE),
]
PROFILE_TYPE = [
'os.nova.server-1.0',
]
KEYS = (POOL, VIP, HEALTH_MONITOR,
LB_STATUS_TIMEOUT) = ('pool', 'vip', 'health_monitor',
'lb_status_timeout')
_POOL_KEYS = (
POOL_PROTOCOL,
POOL_PROTOCOL_PORT,
POOL_SUBNET,
POOL_LB_METHOD,
POOL_ADMIN_STATE_UP,
POOL_SESSION_PERSISTENCE,
) = (
'protocol',
'protocol_port',
'subnet',
'lb_method',
'admin_state_up',
'session_persistence',
)
PROTOCOLS = (
HTTP,
HTTPS,
TCP,
) = (
'HTTP',
'HTTPS',
'TCP',
)
LB_METHODS = (
ROUND_ROBIN,
LEAST_CONNECTIONS,
SOURCE_IP,
) = (
'ROUND_ROBIN',
'LEAST_CONNECTIONS',
'SOURCE_IP',
)
HEALTH_MONITOR_TYPES = (
PING,
TCP,
HTTP,
HTTPS,
) = (
'PING',
'TCP',
'HTTP',
'HTTPS',
)
HTTP_METHODS = (
GET,
POST,
PUT,
DELETE,
) = (
'GET',
'POST',
'PUT',
'DELETE',
)
_VIP_KEYS = (
VIP_SUBNET,
VIP_ADDRESS,
VIP_CONNECTION_LIMIT,
VIP_PROTOCOL,
VIP_PROTOCOL_PORT,
VIP_ADMIN_STATE_UP,
) = (
'subnet',
'address',
'connection_limit',
'protocol',
'protocol_port',
'admin_state_up',
)
HEALTH_MONITOR_KEYS = (
HM_TYPE,
HM_DELAY,
HM_TIMEOUT,
HM_MAX_RETRIES,
HM_ADMIN_STATE_UP,
HM_HTTP_METHOD,
HM_URL_PATH,
HM_EXPECTED_CODES,
) = (
'type',
'delay',
'timeout',
'max_retries',
'admin_state_up',
'http_method',
'url_path',
'expected_codes',
)
_SESSION_PERSISTENCE_KEYS = (
PERSISTENCE_TYPE,
COOKIE_NAME,
) = (
'type',
'cookie_name',
)
PERSISTENCE_TYPES = (
PERSIST_SOURCE_IP,
PERSIST_HTTP_COOKIE,
PERSIST_APP_COOKIE,
) = (
'SOURCE_IP',
'HTTP_COOKIE',
'APP_COOKIE',
)
properties_schema = {
POOL:
schema.Map(
_('LB pool properties.'),
schema={
POOL_PROTOCOL:
schema.String(
_('Protocol used for load balancing.'),
constraints=[
constraints.AllowedValues(PROTOCOLS),
],
default=HTTP,
),
POOL_PROTOCOL_PORT:
schema.Integer(
_('Port on which servers are running on the nodes.'),
default=80,
),
POOL_SUBNET:
schema.String(
_('Name or ID of subnet for the port on which nodes can '
'be connected.'),
required=True,
),
POOL_LB_METHOD:
schema.String(
_('Load balancing algorithm.'),
constraints=[
constraints.AllowedValues(LB_METHODS),
],
default=ROUND_ROBIN,
),
POOL_ADMIN_STATE_UP:
schema.Boolean(
_('Administrative state of the pool.'),
default=True,
),
POOL_SESSION_PERSISTENCE:
schema.Map(
_('Session persistence configuration.'),
schema={
PERSISTENCE_TYPE:
schema.String(
_('Type of session persistence implementation.'),
constraints=[
constraints.AllowedValues(PERSISTENCE_TYPES),
],
),
COOKIE_NAME:
schema.String(
_('Name of cookie if type set to APP_COOKIE.'), ),
},
default={},
),
},
),
VIP:
schema.Map(
_('VIP address and port of the pool.'),
schema={
VIP_SUBNET:
schema.String(
_('Name or ID of Subnet on which the VIP address will be '
'allocated.'),
required=True,
),
VIP_ADDRESS:
schema.String(
_('IP address of the VIP.'),
default=None,
),
VIP_CONNECTION_LIMIT:
schema.Integer(
_('Maximum number of connections per second allowed for '
'this VIP'),
default=-1,
),
VIP_PROTOCOL:
schema.String(
_('Protocol used for VIP.'),
constraints=[
constraints.AllowedValues(PROTOCOLS),
],
default=HTTP,
),
VIP_PROTOCOL_PORT:
schema.Integer(
_('TCP port to listen on.'),
default=80,
),
VIP_ADMIN_STATE_UP:
schema.Boolean(
_('Administrative state of the VIP.'),
default=True,
),
},
),
HEALTH_MONITOR:
schema.Map(
_('Health monitor for loadbalancer.'),
schema={
HM_TYPE:
schema.String(
_('The type of probe sent by the loadbalancer to verify '
'the member state.'),
constraints=[
constraints.AllowedValues(HEALTH_MONITOR_TYPES),
],
default=PING,
),
HM_DELAY:
schema.Integer(
_('The amount of time in milliseconds between sending '
'probes to members.'),
default=10,
),
HM_TIMEOUT:
schema.Integer(
_('The maximum time in milliseconds that a monitor waits '
'to connect before it times out.'),
default=5,
),
HM_MAX_RETRIES:
schema.Integer(
_('The number of allowed connection failures before '
'changing the status of the member to INACTIVE.'),
default=3,
),
HM_ADMIN_STATE_UP:
schema.Boolean(
_('Administrative state of the health monitor.'),
default=True,
),
HM_HTTP_METHOD:
schema.String(
_('The HTTP method that the monitor uses for requests.'),
constraints=[
constraints.AllowedValues(HTTP_METHODS),
],
),
HM_URL_PATH:
schema.String(
_('The HTTP path of the request sent by the monitor to '
'test the health of a member.'), ),
HM_EXPECTED_CODES:
schema.String(
_('Expected HTTP codes for a passing HTTP(S) monitor.'), ),
},
),
LB_STATUS_TIMEOUT:
schema.Integer(
_('Time in second to wait for loadbalancer to become ready '
'after senlin requests LBaaS V2 service for operations.'),
default=300,
)
}
def __init__(self, name, spec, **kwargs):
super(LoadBalancingPolicy, self).__init__(name, spec, **kwargs)
self.pool_spec = self.properties.get(self.POOL, {})
self.vip_spec = self.properties.get(self.VIP, {})
self.hm_spec = self.properties.get(self.HEALTH_MONITOR, None)
self.lb_status_timeout = self.properties.get(self.LB_STATUS_TIMEOUT)
self.lb = None
def validate(self, context, validate_props=False):
super(LoadBalancingPolicy, self).validate(context, validate_props)
if not validate_props:
return True
nc = self.network(context.user, context.project)
# validate pool subnet
name_or_id = self.pool_spec.get(self.POOL_SUBNET)
try:
nc.subnet_get(name_or_id)
except exc.InternalError:
msg = _(
"The specified %(key)s '%(value)s' could not be found.") % {
'key': self.POOL_SUBNET,
'value': name_or_id
}
raise exc.InvalidSpec(message=msg)
# validate VIP subnet
name_or_id = self.vip_spec.get(self.VIP_SUBNET)
try:
nc.subnet_get(name_or_id)
except exc.InternalError:
msg = _(
"The specified %(key)s '%(value)s' could not be found.") % {
'key': self.VIP_SUBNET,
'value': name_or_id
}
raise exc.InvalidSpec(message=msg)
def attach(self, cluster, enabled=True):
"""Routine to be invoked when policy is to be attached to a cluster.
:param cluster: The cluster to which the policy is being attached to.
:param enabled: The attached cluster policy is enabled or disabled.
:returns: When the operation was successful, returns a tuple (True,
message); otherwise, return a tuple (False, error).
"""
res, data = super(LoadBalancingPolicy, self).attach(cluster)
if res is False:
return False, data
nodes = nm.Node.load_all(oslo_context.get_current(),
cluster_id=cluster.id)
lb_driver = self.lbaas(cluster.user, cluster.project)
lb_driver.lb_status_timeout = self.lb_status_timeout
# TODO(Anyone): Check if existing nodes has conflicts regarding the
# subnets. Each VM addresses detail has a key named to the network
# which can be used for validation.
res, data = lb_driver.lb_create(self.vip_spec, self.pool_spec,
self.hm_spec)
if res is False:
return False, data
port = self.pool_spec.get(self.POOL_PROTOCOL_PORT)
subnet = self.pool_spec.get(self.POOL_SUBNET)
for node in nodes:
member_id = lb_driver.member_add(node, data['loadbalancer'],
data['pool'], port, subnet)
if member_id is None:
# When failed in adding member, remove all lb resources that
# were created and return the failure reason.
# TODO(anyone): May need to "roll-back" changes caused by any
# successful member_add() calls.
lb_driver.lb_delete(**data)
return False, 'Failed in adding node into lb pool'
node.data.update({'lb_member': member_id})
node.store(oslo_context.get_current())
cluster_data_lb = cluster.data.get('loadbalancers', {})
cluster_data_lb[self.id] = {'vip_address': data.pop('vip_address')}
cluster.data['loadbalancers'] = cluster_data_lb
policy_data = self._build_policy_data(data)
return True, policy_data
def detach(self, cluster):
"""Routine to be called when the policy is detached from a cluster.
:param cluster: The cluster from which the policy is to be detached.
:returns: When the operation was successful, returns a tuple of
(True, data) where the data contains references to the resources
created; otherwise returns a tuple of (False, err) where the err
contains a error message.
"""
reason = _('LB resources deletion succeeded.')
lb_driver = self.lbaas(cluster.user, cluster.project)
lb_driver.lb_status_timeout = self.lb_status_timeout
cp = cluster_policy.ClusterPolicy.load(oslo_context.get_current(),
cluster.id, self.id)
policy_data = self._extract_policy_data(cp.data)
if policy_data is None:
return True, reason
res, reason = lb_driver.lb_delete(**policy_data)
if res is False:
return False, reason
nodes = nm.Node.load_all(oslo_context.get_current(),
cluster_id=cluster.id,
project_safe=False)
for node in nodes:
if 'lb_member' in node.data:
node.data.pop('lb_member')
node.store(oslo_context.get_current())
lb_data = cluster.data.get('loadbalancers', {})
if lb_data and isinstance(lb_data, dict):
lb_data.pop(self.id, None)
if lb_data:
cluster.data['loadbalancers'] = lb_data
else:
cluster.data.pop('loadbalancers')
return True, reason
def _get_delete_candidates(self, cluster_id, action):
deletion = action.data.get('deletion', None)
# No deletion field in action.data which means no scaling
# policy or deletion policy is attached.
candidates = None
if deletion is None:
if action.action == consts.NODE_DELETE:
candidates = [action.node.id]
count = 1
elif action.action == consts.CLUSTER_DEL_NODES:
# Get candidates from action.input
candidates = action.inputs.get('candidates', [])
count = len(candidates)
elif action.action == consts.CLUSTER_RESIZE:
# Calculate deletion count based on action input
db_cluster = co.Cluster.get(action.context, cluster_id)
current = no.Node.count_by_cluster(action.context, cluster_id)
scaleutils.parse_resize_params(action, db_cluster, current)
if 'deletion' not in action.data:
return []
else:
count = action.data['deletion']['count']
else: # action.action == consts.CLUSTER_SCALE_IN
count = 1
else:
count = deletion.get('count', 0)
candidates = deletion.get('candidates', None)
# Still no candidates available, pick count of nodes randomly
if candidates is None:
if count == 0:
return []
nodes = no.Node.get_all_by_cluster(action.context, cluster_id)
if count > len(nodes):
count = len(nodes)
candidates = scaleutils.nodes_by_random(nodes, count)
deletion_data = action.data.get('deletion', {})
deletion_data.update({
'count': len(candidates),
'candidates': candidates
})
action.data.update({'deletion': deletion_data})
return candidates
def _remove_member(self,
candidates,
policy,
action,
driver,
handle_err=True):
# Load policy data
policy_data = self._extract_policy_data(policy.data)
lb_id = policy_data['loadbalancer']
pool_id = policy_data['pool']
failed_nodes = []
for node_id in candidates:
node = no.Node.get(action.context, node_id=node_id)
node_data = node.data or {}
member_id = node_data.get('lb_member', None)
if member_id is None:
LOG.warning('Node %(n)s not found in lb pool %(p)s.', {
'n': node_id,
'p': pool_id
})
continue
res = driver.member_remove(lb_id, pool_id, member_id)
values = {}
if res is not True and handle_err is True:
failed_nodes.append(node.id)
values['status'] = consts.NS_WARNING
values['status_reason'] = _(
'Failed in removing node from lb pool.')
else:
node.data.pop('lb_member', None)
values['data'] = node.data
no.Node.update(action.context, node_id, values)
return failed_nodes
def _add_member(self, candidates, policy, action, driver):
# Load policy data
policy_data = self._extract_policy_data(policy.data)
lb_id = policy_data['loadbalancer']
pool_id = policy_data['pool']
port = self.pool_spec.get(self.POOL_PROTOCOL_PORT)
subnet = self.pool_spec.get(self.POOL_SUBNET)
failed_nodes = []
for node_id in candidates:
node = no.Node.get(action.context, node_id=node_id)
node_data = node.data or {}
member_id = node_data.get('lb_member', None)
if member_id:
LOG.warning('Node %(n)s already in lb pool %(p)s.', {
'n': node_id,
'p': pool_id
})
continue
member_id = driver.member_add(node, lb_id, pool_id, port, subnet)
values = {}
if member_id is None:
failed_nodes.append(node.id)
values['status'] = consts.NS_WARNING
values['status_reason'] = _(
'Failed in adding node into lb pool.')
else:
node.data.update({'lb_member': member_id})
values['data'] = node.data
no.Node.update(action.context, node_id, values)
return failed_nodes
def _get_post_candidates(self, action):
# This method will parse action data passed from action layer
candidates = []
if action.action == consts.NODE_CREATE:
candidates = [action.node.id]
elif action.action == consts.NODE_RECOVER:
recovery = action.outputs.get('recovery', None)
if recovery is not None and 'action' in recovery:
action_name = recovery['action']
if action_name.upper() == consts.RECOVER_RECREATE:
candidates = recovery.get('node', [])
else:
creation = action.data.get('creation', None)
candidates = creation.get('nodes', []) if creation else []
return candidates
def pre_op(self, cluster_id, action):
"""Routine to be called before an action has been executed.
For this particular policy, we take this chance to update the pool
maintained by the load-balancer.
:param cluster_id: The ID of the cluster on which a relevant action
has been executed.
:param action: The action object that triggered this operation.
:returns: Nothing.
"""
candidates = self._get_delete_candidates(cluster_id, action)
if len(candidates) == 0:
return
db_cluster = co.Cluster.get(action.context, cluster_id)
lb_driver = self.lbaas(db_cluster.user, db_cluster.project)
lb_driver.lb_status_timeout = self.lb_status_timeout
cp = cluster_policy.ClusterPolicy.load(action.context, cluster_id,
self.id)
# Remove nodes that will be deleted from lb pool
failed_nodes = self._remove_member(candidates, cp, action, lb_driver)
if failed_nodes:
error = _('Failed in removing deleted node(s) from lb pool: %s'
) % failed_nodes
action.data['status'] = base.CHECK_ERROR
action.data['reason'] = error
return
def post_op(self, cluster_id, action):
"""Routine to be called after an action has been executed.
For this particular policy, we take this chance to update the pool
maintained by the load-balancer.
:param cluster_id: The ID of the cluster on which a relevant action
has been executed.
:param action: The action object that triggered this operation.
:returns: Nothing.
"""
# TODO(Yanyanhu): Need special handling for cross-az scenario
# which is supported by Neutron lbaas.
candidates = self._get_post_candidates(action)
if len(candidates) == 0:
return
db_cluster = co.Cluster.get(action.context, cluster_id)
lb_driver = self.lbaas(db_cluster.user, db_cluster.project)
lb_driver.lb_status_timeout = self.lb_status_timeout
cp = cluster_policy.ClusterPolicy.load(action.context, cluster_id,
self.id)
if action.action == consts.NODE_RECOVER:
self._remove_member(candidates,
cp,
action,
lb_driver,
handle_err=False)
# Add new nodes to lb pool
failed_nodes = self._add_member(candidates, cp, action, lb_driver)
if failed_nodes:
error = _('Failed in adding nodes into lb pool: %s') % failed_nodes
action.data['status'] = base.CHECK_ERROR
action.data['reason'] = error
return
class ServerProfile(base.Profile):
'''Profile for an OpenStack Nova server.'''
KEYS = (
CONTEXT,
ADMIN_PASS,
AUTO_DISK_CONFIG,
AVAILABILITY_ZONE,
BLOCK_DEVICE_MAPPING, # BLOCK_DEVICE_MAPPING_V2,
CONFIG_DRIVE,
FLAVOR,
IMAGE,
KEY_NAME,
METADATA,
NAME,
NETWORKS,
PERSONALITY,
SECURITY_GROUPS,
TIMEOUT,
USER_DATA,
SCHEDULER_HINTS,
) = (
'context',
'adminPass',
'auto_disk_config',
'availability_zone',
'block_device_mapping',
# 'block_device_mapping_v2',
'config_drive',
'flavor',
'image',
'key_name',
'metadata',
'name',
'networks',
'personality',
'security_groups',
'timeout',
'user_data',
'scheduler_hints',
)
BDM_KEYS = (
BDM_DEVICE_NAME,
BDM_VOLUME_SIZE,
) = (
'device_name',
'volume_size',
)
NETWORK_KEYS = (
PORT,
FIXED_IP,
NETWORK,
) = (
'port',
'fixed-ip',
'network',
)
PERSONALITY_KEYS = (
PATH,
CONTENTS,
) = (
'path',
'contents',
)
SCHEDULER_HINTS_KEYS = (GROUP, ) = ('group', )
properties_schema = {
CONTEXT:
schema.Map(_('Customized security context for operating servers.'), ),
ADMIN_PASS:
schema.String(_('Password for the administrator account.'), ),
AUTO_DISK_CONFIG:
schema.Boolean(
_('Whether the disk partition is done automatically.'),
default=True,
),
AVAILABILITY_ZONE:
schema.String(
_('Name of availability zone for running the server.'), ),
BLOCK_DEVICE_MAPPING:
schema.List(
_('A list specifying the properties of block devices to be used '
'for this server.'),
schema=schema.Map(
_('A map specifying the properties of a block device to be '
'used by the server.'),
schema={
BDM_DEVICE_NAME:
schema.String(
_('Block device name, should be <=255 chars.'), ),
BDM_VOLUME_SIZE:
schema.Integer(_('Block device size in GB.'), ),
}),
),
CONFIG_DRIVE:
schema.Boolean(
_('Whether config drive should be enabled for the server.'), ),
FLAVOR:
schema.String(
_('ID of flavor used for the server.'),
required=True,
),
IMAGE:
schema.String(
# IMAGE is not required, because there could be BDM or BDMv2
# support and the corresponding settings effective
_('ID of image to be used for the new server.'), ),
KEY_NAME:
schema.String(_('Name of Nova keypair to be injected to server.'), ),
METADATA:
schema.Map(
_('A collection of key/value pairs to be associated with the '
'server created. Both key and value should be <=255 chars.'), ),
NAME:
schema.String(
_('Name of the server.'),
required=True,
),
NETWORKS:
schema.List(
_('List of networks for the server.'),
schema=schema.Map(
_('A map specifying the properties of a network for uses.'),
schema={
NETWORK:
schema.String(
_('Name or ID of network to create a port on.'), ),
PORT:
schema.String(_('Port ID to be used by the network.'), ),
FIXED_IP:
schema.String(_('Fixed IP to be used by the network.'), ),
},
),
),
PERSONALITY:
schema.List(
_('List of files to be injected into the server, where each.'),
schema=schema.Map(
_('A map specifying the path & contents for an injected '
'file.'),
schema={
PATH:
schema.String(
_('In-instance path for the file to be injected.'),
required=True,
),
CONTENTS:
schema.String(
_('Contents of the file to be injected.'),
required=True,
),
},
),
),
SCHEDULER_HINTS:
schema.Map(
_('A collection of key/value pairs to be associated with the '
'Scheduler hints. Both key and value should be <=255 chars.'), ),
SECURITY_GROUPS:
schema.List(
_('List of security groups.'),
schema=schema.String(
_('Name of a security group'),
required=True,
),
),
TIMEOUT:
schema.Integer(
_('Time out threshold for server operations.'),
default=120,
),
USER_DATA:
schema.String(_('User data to be exposed by the metadata server.'), ),
}
def __init__(self, type_name, name, **kwargs):
super(ServerProfile, self).__init__(type_name, name, **kwargs)
self._novaclient = None
self._neutronclient = None
self.server_id = None
def validate(self):
super(ServerProfile, self).validate()
if self.properties[self.TIMEOUT] > cfg.CONF.default_action_timeout:
suggest = cfg.CONF.default_action_timeout
err = _("Value of the 'timeout' property must be lower than the "
"upper limit (%s).") % suggest
raise exception.InvalidSpec(message=err)
def nova(self, obj):
'''Construct nova client based on object.
:param obj: Object for which the client is created. It is expected to
be None when retrieving an existing client. When creating
a client, it contains the user and project to be used.
'''
if self._novaclient is not None:
return self._novaclient
params = self._build_conn_params(obj.user, obj.project)
self._novaclient = driver_base.SenlinDriver().compute(params)
return self._novaclient
def neutron(self, obj):
'''Construct neutron client based on object.
:param obj: Object for which the client is created. It is expected to
be None when retrieving an existing client. When creating
a client, it contains the user and project to be used.
'''
if self._neutronclient is not None:
return self._neutronclient
params = self._build_conn_params(obj.user, obj.project)
self._neutronclient = driver_base.SenlinDriver().network(params)
return self._neutronclient
def do_validate(self, obj):
'''Validate if the spec has provided valid info for server creation.'''
return True
def do_create(self, obj):
'''Create a server using the given profile.'''
kwargs = {}
for key in self.KEYS:
# context is treated as connection parameters
if key == self.CONTEXT:
continue
if self.properties[key] is not None:
kwargs[key] = self.properties[key]
name_or_id = self.properties[self.IMAGE]
if name_or_id is not None:
image = self.nova(obj).image_get_by_name(name_or_id)
# wait for new version of openstacksdk to fix this
kwargs.pop(self.IMAGE)
kwargs['imageRef'] = image.id
flavor_id = self.properties[self.FLAVOR]
flavor = self.nova(obj).flavor_find(flavor_id, False)
# wait for new verson of openstacksdk to fix this
kwargs.pop(self.FLAVOR)
kwargs['flavorRef'] = flavor.id
if obj.name is not None:
kwargs[self.NAME] = obj.name + '-' + utils.random_name(8)
metadata = self.properties[self.METADATA] or {}
if obj.cluster_id is not None:
metadata['cluster'] = obj.cluster_id
kwargs['metadata'] = metadata
scheduler_hint = self.properties[self.SCHEDULER_HINTS]
if scheduler_hint is not None:
kwargs['scheduler_hints'] = scheduler_hint
user_data = self.properties[self.USER_DATA]
if user_data is not None:
ud = encodeutils.safe_encode(user_data)
kwargs['user_data'] = encodeutils.safe_decode(base64.b64encode(ud))
networks = self.properties[self.NETWORKS]
if networks is not None:
for network in networks:
net_name_id = network.get(self.NETWORK)
if net_name_id:
res = self.neutron(obj).network_get(net_name_id)
network['uuid'] = res.id
del network[self.NETWORK]
if network['port'] is None:
del network['port']
if network['fixed-ip'] is None:
del network['fixed-ip']
kwargs['networks'] = networks
LOG.info('Creating server: %s' % kwargs)
try:
server = self.nova(obj).server_create(**kwargs)
self.nova(obj).wait_for_server(server)
except Exception as ex:
LOG.exception(_('Failed in creating server: %s'),
six.text_type(ex))
return False
self.server_id = server.id
return server.id
def do_delete(self, obj):
self.server_id = obj.physical_id
if not obj.physical_id:
return True
try:
self.nova(obj).server_delete(self.server_id)
self.nova(obj).wait_for_server_delete(self.server_id)
except Exception as ex:
LOG.error('Error: %s' % six.text_type(ex))
return False
return True
def do_update(self, obj, new_profile, **params):
'''Perform update on the server.
:param obj: the server to operate on
:param new_profile: the new profile for the server.
:param params: a dictionary of optional parameters.
'''
self.server_id = obj.physical_id
if not self.server_id:
return True
# TODO(anyone): Validate the new profile
# TODO(anyone): Do update based on the fields provided.
# self.nova(obj).server_update(**fields)
return True
def do_check(self, obj):
# TODO(anyone): Check server status
return True
def do_get_details(self, obj):
if obj.physical_id is None or obj.physical_id == '':
return {}
server = self.nova(obj).server_get(obj.physical_id)
if server is None:
return {}
details = {
'id': server.id,
'name': server.name,
'access_ipv4': server.access_ipv4,
'access_ipv6': server.access_ipv6,
'addresses': server.addresses,
'created_at': server.created_at,
'flavor': server.flavor,
'host_id': server.host_id,
'image': server.image,
'links': server.links,
'metadata': server.metadata,
'progress': server.progress,
'project_id': server.project_id,
'status': server.status,
'updated_at': server.updated_at,
'user_id': server.user_id,
}
return details
def do_join(self, obj, cluster_id):
if not obj.physical_id:
return {}
metadata = self.nova(obj).server_metadata_get(
server_id=obj.physical_id) or {}
metadata['cluster'] = cluster_id
return self.nova(obj).server_metadata_update(**metadata)
def do_leave(self, obj):
if not obj.physical_id:
return
metadata = self.nova(obj).server_metadata_get(
server_id=obj.physical_id) or {}
if 'cluster' in metadata:
del metadata['cluster']
return self.nova(obj).server_metadata_update(**metadata)
class ScalingPolicy(base.Policy):
"""Policy for changing the size of a cluster.
This policy is expected to be enforced before the node count of a cluster
is changed.
"""
VERSION = '1.0'
PRIORITY = 100
TARGET = [
('BEFORE', consts.CLUSTER_SCALE_IN),
('BEFORE', consts.CLUSTER_SCALE_OUT),
]
PROFILE_TYPE = [
'ANY',
]
KEYS = (
EVENT,
ADJUSTMENT,
) = (
'event',
'adjustment',
)
_SUPPORTED_EVENTS = (
CLUSTER_SCALE_IN,
CLUSTER_SCALE_OUT,
) = (
consts.CLUSTER_SCALE_IN,
consts.CLUSTER_SCALE_OUT,
)
_ADJUSTMENT_KEYS = (
ADJUSTMENT_TYPE,
ADJUSTMENT_NUMBER,
MIN_STEP,
BEST_EFFORT,
COOLDOWN,
) = (
'type',
'number',
'min_step',
'best_effort',
'cooldown',
)
properties_schema = {
EVENT:
schema.String(
_('Event that will trigger this policy. Must be one of '
'CLUSTER_SCALE_IN and CLUSTER_SCALE_OUT.'),
constraints=[
constraints.AllowedValues(_SUPPORTED_EVENTS),
],
required=True,
),
ADJUSTMENT:
schema.Map(
_('Detailed specification for scaling adjustments.'),
schema={
ADJUSTMENT_TYPE:
schema.String(
_('Type of adjustment when scaling is triggered.'),
constraints=[
constraints.AllowedValues(consts.ADJUSTMENT_TYPES),
],
default=consts.CHANGE_IN_CAPACITY,
),
ADJUSTMENT_NUMBER:
schema.Number(
_('A number specifying the amount of adjustment.'),
default=1,
),
MIN_STEP:
schema.Integer(
_('When adjustment type is set to "CHANGE_IN_PERCENTAGE",'
' this specifies the cluster size will be decreased by '
'at least this number of nodes.'),
default=1,
),
BEST_EFFORT:
schema.Boolean(
_('Whether do best effort scaling when new size of '
'cluster will break the size limitation'),
default=False,
),
COOLDOWN:
schema.Integer(
_('Number of seconds to hold the cluster for cool-down '
'before allowing cluster to be resized again.'),
default=0,
),
}),
}
def __init__(self, name, spec, **kwargs):
"""Intialize a scaling policy object.
:param name: Name for the policy object.
:param spec: A dictionary containing the detailed specification for
the policy.
:param \*\*kwargs: Other optional parameters for policy object
creation.
:return: An object of `ScalingPolicy`.
"""
super(ScalingPolicy, self).__init__(name, spec, **kwargs)
self.singleton = False
self.event = self.properties[self.EVENT]
adjustment = self.properties[self.ADJUSTMENT]
self.adjustment_type = adjustment[self.ADJUSTMENT_TYPE]
self.adjustment_number = adjustment[self.ADJUSTMENT_NUMBER]
self.adjustment_min_step = adjustment[self.MIN_STEP]
self.best_effort = adjustment[self.BEST_EFFORT]
self.cooldown = adjustment[self.COOLDOWN]
def _calculate_adjustment_count(self, current_size):
"""Calculate adjustment count based on current_size.
:param current_size: The current size of the target cluster.
:return: The number of nodes to add or to remove.
"""
if self.adjustment_type == consts.EXACT_CAPACITY:
if self.event == consts.CLUSTER_SCALE_IN:
count = current_size - self.adjustment_number
else:
count = self.adjustment_number - current_size
elif self.adjustment_type == consts.CHANGE_IN_CAPACITY:
count = self.adjustment_number
else: # consts.CHANGE_IN_PERCENTAGE:
count = int((self.adjustment_number * current_size) / 100.0)
if count < self.adjustment_min_step:
count = self.adjustment_min_step
return count
def pre_op(self, cluster_id, action):
"""The hook function that is executed before the action.
The checking result is stored in the ``data`` property of the action
object rather than returned directly from the function.
:param cluster_id: The ID of the target cluster.
:param action: Action instance against which the policy is being
checked.
:return: None.
"""
# Use action input if count is provided
count = action.inputs.get('count', None)
current = db_api.node_count_by_cluster(action.context, cluster_id)
if count is None:
# count not specified, calculate it
count = self._calculate_adjustment_count(current)
# Count must be positive value
try:
count = utils.parse_int_param('count', count, allow_zero=False)
except exception.InvalidParameter:
action.data.update({
'status': base.CHECK_ERROR,
'reason': _("Invalid count (%(c)s) for action '%(a)s'.") % {
'c': count,
'a': action.action
}
})
action.store(action.context)
return
# Check size constraints
cluster = db_api.cluster_get(action.context, cluster_id)
if action.action == consts.CLUSTER_SCALE_IN:
if self.best_effort:
count = min(count, current - cluster.min_size)
result = su.check_size_params(cluster,
current - count,
strict=not self.best_effort)
else:
if self.best_effort:
count = min(count, cluster.max_size - current)
result = su.check_size_params(cluster,
current + count,
strict=not self.best_effort)
if result:
# failed validation
pd = {'status': base.CHECK_ERROR, 'reason': result}
else:
# passed validation
pd = {
'status': base.CHECK_OK,
'reason': _('Scaling request validated.'),
}
if action.action == consts.CLUSTER_SCALE_IN:
pd['deletion'] = {'count': count}
else:
pd['creation'] = {'count': count}
action.data.update(pd)
action.store(action.context)
return
def need_check(self, target, action):
res = super(ScalingPolicy, self).need_check(target, action)
if res:
# Check if the action is expected by the policy
res = (self.event == action.action)
return res
class Alarm(base.Trigger):
# time constraints
alarm_schema = {
REPEAT:
schema.Boolean(
_('Whether the actions should be re-triggered on each evaluation '
'cycle. Default to False.'),
default=False,
),
TIME_CONSTRAINTS:
schema.List(schema=schema.Map(
_('A map of time constraint settings.'),
schema={
NAME:
schema.String(_('Name of the time constraint.'), ),
TC_DESCRIPTION:
schema.String(_('A description of the time constraint.'), ),
TC_START:
schema.String(
_('Start point of the time constraint, expressed as a '
'string in cron expression format.'),
required=True,
),
TC_DURATION:
schema.Integer(
_('How long the constraint should last, in seconds.'),
required=True,
),
TC_TIMEZONE:
schema.String(
_('Time zone of the constraint.'),
default='',
),
},
), )
}
def __init__(self, name, spec, **kwargs):
super(Alarm, self).__init__(name, spec, **kwargs)
self.alarm_properties = schema.Spec(self.alarm_schema, spec)
self.namespace = 'default'
self.rule = None
def validate(self):
# validate cron expression if specified
if TIME_CONSTRAINTS in self.spec:
tcs = self.alarm_properties[TIME_CONSTRAINTS]
for tc in tcs:
exp = tc.get(TC_START, '')
try:
croniter.croniter(exp)
except Exception as ex:
msg = _("Invalid cron expression specified for property "
"'%(property)s' (%(exp)s): %(ex)s") % {
'property': TC_START,
'exp': exp,
'ex': six.text_type(ex)
}
raise exc.InvalidSpec(message=msg)
tz = tc.get(TC_TIMEZONE, '')
try:
pytz.timezone(tz)
except Exception as ex:
msg = _("Invalid timezone value specified for property "
"'%(property)s' (%(tz)s): %(ex)s") % {
'property': TC_TIMEZONE,
'tz': tz,
'ex': six.text_type(ex)
}
raise exc.InvalidSpec(message=msg)
def create(self, ctx, **kwargs):
"""Create an alarm for a cluster.
:param name: The name for the alarm.
:param urls: A list of URLs for webhooks to be triggered.
:returns: A dict containing properties of the alarm.
"""
self.ok_actions = kwargs.get(OK_ACTIONS, [])
self.alarm_actions = kwargs.get(ALARM_ACTIONS, [])
self.insufficient_data_actions = kwargs.get(INSUFFICIENT_DATA_ACTIONS,
[])
rule_name = self.namespace + '_rule'
rule_data = dict((k, v) for k, v in self.rule.items())
params = {
NAME: self.name,
DESCRIPTION: self.desc,
TYPE: self.namespace,
STATE: self.state,
SEVERITY: self.severity,
ENABLED: self.enabled,
OK_ACTIONS: self.ok_actions,
ALARM_ACTIONS: self.alarm_actions,
INSUFFICIENT_DATA_ACTIONS: self.insufficient_data_actions,
TIME_CONSTRAINTS: self.alarm_properties[TIME_CONSTRAINTS],
REPEAT: self.alarm_properties[REPEAT],
rule_name: rule_data,
}
try:
cc = driver_base.SenlinDriver().telemetry(ctx.to_dict())
alarm = cc.alarm_create(**params)
self.physical_id = alarm.id
self.store(ctx)
return True, alarm.to_dict()
except exc.SenlinException as ex:
return False, six.text_type(ex)
def delete(self, ctx, identifier):
"""Delete an alarm.
:param identifier: This must be an alarm ID.
"""
try:
cc = driver_base.SenlinDriver().telemetry(ctx)
res = cc.alarm_delete(identifier, True)
return True, res
except exc.InternalError as ex:
return False, six.text_type(ex)
def update(self, identifier, values):
return NotImplemented