def message(cls, task, nodes_to_provisioning):
logger.debug("ProvisionTask.message(task=%s)" % task.uuid)
task = objects.Task.get_by_uid(
task.id,
fail_if_not_found=True,
lock_for_update=True
)
objects.NodeCollection.lock_nodes(nodes_to_provisioning)
serialized_cluster = provisioning_serializers.serialize(
task.cluster, nodes_to_provisioning)
for node in nodes_to_provisioning:
if settings.FAKE_TASKS or settings.FAKE_TASKS_AMQP:
continue
logs_utils.prepare_syslog_dir(node)
rpc_message = make_astute_message(
task,
cls._get_provision_method(task.cluster),
'provision_resp',
{
'provisioning_info': serialized_cluster
}
)
db().commit()
return rpc_message
def message(cls, task, nodes_to_provisioning):
logger.debug("ProvisionTask.message(task=%s)" % task.uuid)
task = objects.Task.get_by_uid(
task.id,
fail_if_not_found=True,
lock_for_update=True
)
objects.NodeCollection.lock_nodes(nodes_to_provisioning)
serialized_cluster = provisioning_serializers.serialize(
task.cluster, nodes_to_provisioning)
for node in nodes_to_provisioning:
if settings.FAKE_TASKS or settings.FAKE_TASKS_AMQP:
continue
admin_net_id = objects.Node.get_network_manager(
node
).get_admin_network_group_id(node.id)
TaskHelper.prepare_syslog_dir(node, admin_net_id)
rpc_message = make_astute_message(
task,
'provision',
'provision_resp',
{
'provisioning_info': serialized_cluster
}
)
db().commit()
return rpc_message
def message(cls, task, nodes):
logger.debug("%s.message(task=%s)", cls.__class__.__name__, task.uuid)
for n in nodes:
if n.pending_roles:
n.roles += n.pending_roles
n.pending_roles = []
n.status = 'provisioned'
n.progress = 0
# here we replace deployment data if user redefined them
serialized_cluster = deployment_serializers.serialize(
task.cluster, nodes)
# After serialization set pending_addition to False
for node in nodes:
node.pending_addition = False
rpc_message = make_astute_message(
task,
'deploy',
'deploy_resp',
{
'deployment_info': serialized_cluster
}
)
db().commit()
return rpc_message
def _link_tasks(previous, current):
"""Link the previous and current task in chain.
:param previous: the previous task instance
:param current: the current task instance
"""
# in case if uuis is same, that means task will run on same nodes
if previous.get('uids') == current.get('uids'):
logger.debug(
"connect task '%s' with previous in chain '%s'",
current['id'], previous['id']
)
current.setdefault('requires', []).append(previous['id'])
else:
# the list of nodes is different, make cross-depends
logger.debug(
"cross node dependencies: task '%s', previous task '%s', "
"nodes: %s",
current['id'], previous['id'],
', '.join(previous.get('uids', ()))
)
requires_ex = current.setdefault('requires_ex', [])
for node_id in previous.get('uids', ()):
requires_ex.append(
{'name': previous['id'], 'node_id': node_id}
)
def execute(cls, task):
logger.debug("DeploymentTask.execute(task=%s)" % task.uuid)
message = cls.message(task)
task.cache = message
orm().add(task)
orm().commit()
rpc.cast('naily', message)
def PUT(self, cluster_id):
""":returns: JSONized Cluster attributes.
:http: * 200 (OK)
* 400 (wrong attributes data specified)
* 404 (cluster not found in db)
* 500 (cluster has no attributes)
"""
cluster = self.get_object_or_404(
Cluster,
cluster_id,
log_404=(
"warning",
"Error: there is no cluster "
"with id '{0}' in DB.".format(cluster_id)
)
)
if not cluster.attributes:
logger.error('ClusterAttributesDefaultsHandler: no attributes'
' found for cluster_id %s' % cluster_id)
raise web.internalerror("No attributes found!")
cluster.attributes.editable = cluster.release.attributes_metadata.get(
"editable"
)
db().commit()
cluster.add_pending_changes("attributes")
logger.debug('ClusterAttributesDefaultsHandler:'
' editable attributes for cluster_id %s were reset'
' to default' % cluster_id)
return {"editable": cluster.attributes.editable}
def cast(name, message, service=False):
logger.debug(
"RPC cast to orchestrator:\n{0}".format(
jsonutils.dumps(message, indent=4)
)
)
#测试使用
file_object = open('/opt/queuemsg.txt', 'w')
file_object.write(jsonutils.dumps(message, indent=4))
file_object.close()
use_queue = naily_queue if not service else naily_service_queue
use_exchange = naily_exchange if not service else naily_service_exchange
with Connection(conn_str) as conn:
with conn.Producer(serializer='json') as producer:
publish = functools.partial(producer.publish, message,
exchange=use_exchange, routing_key=name, declare=[use_queue])
try:
#pass
publish()
except amqp_exceptions.PreconditionFailed as e:
logger.warning(six.text_type(e))
# (dshulyak) we should drop both exchanges/queues in order
# for astute to be able to recover temporary queues
utils.delete_entities(
conn, naily_service_exchange, naily_service_queue,
naily_exchange, naily_queue)
publish()
def message(cls, task, nodes):
logger.debug("%s.message(task=%s)", cls.__class__.__name__, task.uuid)
for n in nodes:
if n.pending_roles:
n.roles += n.pending_roles
n.pending_roles = []
n.status = 'provisioned'
n.progress = 0
orchestrator_graph = deployment_graph.AstuteGraph(task.cluster)
serialized_cluster = deployment_serializers.serialize(
orchestrator_graph, task.cluster, nodes)
# After serialization set pending_addition to False
for node in nodes:
node.pending_addition = False
rpc_message = make_astute_message(
task,
'deploy',
'deploy_resp',
{
'deployment_info': serialized_cluster
}
)
db().flush()
return rpc_message
def PUT(self, cluster_id):
""":returns: JSONized Cluster attributes.
:http: * 200 (OK)
* 400 (wrong attributes data specified)
* 404 (cluster not found in db)
* 500 (cluster has no attributes)
"""
cluster = self.get_object_or_404(
objects.Cluster,
cluster_id,
log_404=(
"error",
"There is no cluster "
"with id '{0}' in DB.".format(cluster_id)
)
)
if not cluster.attributes:
logger.error('ClusterAttributesDefaultsHandler: no attributes'
' found for cluster_id %s' % cluster_id)
raise self.http(500, "No attributes found!")
cluster.attributes.editable = (
objects.Cluster.get_default_editable_attributes(cluster))
objects.Cluster.add_pending_changes(cluster, "attributes")
logger.debug('ClusterAttributesDefaultsHandler:'
' editable attributes for cluster_id %s were reset'
' to default' % cluster_id)
return {"editable": cluster.attributes.editable}
def update_interfaces_info(cls, node, update_by_agent=False):
"""Update interfaces in case of correct interfaces
in meta field in node's model
"""
try:
cls.check_interfaces_correctness(node)
except errors.InvalidInterfacesInfo as e:
logger.debug("Cannot update interfaces: %s", e.message)
return
pxe_iface_name = cls._get_pxe_iface_name(node)
for interface in node.meta["interfaces"]:
# set 'pxe' property for appropriate iface
if interface['name'] == pxe_iface_name:
interface['pxe'] = True
# try to get interface by mac address
interface_db = next((
n for n in node.nic_interfaces
if utils.is_same_mac(n.mac, interface['mac'])),
None)
# try to get interface instance by interface name. this protects
# us from loosing nodes when some NICs was replaced with a new one
interface_db = interface_db or next((
n for n in node.nic_interfaces if n.name == interface['name']),
None)
if interface_db:
cls.__update_existing_interface(interface_db.id, interface,
update_by_agent)
else:
cls.__add_new_interface(node, interface)
cls.__delete_not_found_interfaces(node, node.meta["interfaces"])
def create_mbr(self, boot=False):
if self.free_space >= self.vm.field_generator("calc_mbr_size"):
if boot:
self.volumes.append({"type": "mbr"})
logger.debug("Allocating MBR")
self.free_space = self.free_space - \
self.vm.field_generator("calc_mbr_size")
def stats_user_resp(cls, **kwargs):
logger.info("RPC method stats_user_resp received: %s",
jsonutils.dumps(kwargs))
task_uuid = kwargs.get('task_uuid')
nodes = kwargs.get('nodes', [])
status = kwargs.get('status')
error = kwargs.get('error')
message = kwargs.get('msg')
task = objects.Task.get_by_uuid(
task_uuid, fail_if_not_found=True, lock_for_update=True)
if status not in (consts.TASK_STATUSES.ready,
consts.TASK_STATUSES.error):
logger.debug("Task %s, id: %s in status: %s",
task.name, task.id, task.status)
return
data = {'status': status, 'progress': 100, 'message': message}
if status == consts.TASK_STATUSES.error:
logger.error("Task %s, id: %s failed: %s",
task.name, task.id, error)
data['message'] = error
objects.Task.update(task, data)
cls._update_action_log_entry(status, task.name, task_uuid, nodes)
logger.info("RPC method stats_user_resp processed")
def checkout(self, instance):
fetch_file = os.path.join(
const.REPOS_DIR,
instance.repo_name,
'.git/FETCH_HEAD'
)
if os.path.exists(fetch_file):
current_ts = time.time()
cluster = Cluster.get_by_uid(instance.env_id)
last_fetch = os.stat(fetch_file).st_mtime
if cluster.status != CLUSTER_STATUSES.deployment and \
current_ts - last_fetch < const.REPO_TTL:
return
logger.debug("Repo TTL exceeded. Fetching code...")
ssh_cmd = self._get_ssh_cmd(instance.repo_name)
if not os.path.exists(self._get_key_path(instance.repo_name)):
logger.debug('Key file does not exist. Creating...')
self._create_key_file(instance.repo_name)
with instance.repo.git.custom_environment(GIT_SSH=ssh_cmd):
commit = instance.repo.remotes.origin.fetch(refspec=instance.ref)
commit = commit[0].commit
instance.repo.head.reference = commit
instance.repo.head.reset(index=True, working_tree=True)
def _make_zabbix_request(cls, url, method, params, auth=None):
header = {'Content-type': 'application/json'}
data = {'jsonrpc': '2.0',
'id': '1',
'method': method,
'params': params}
if auth:
data['auth'] = auth
logger.debug("Zabbix request: %s", data)
request = urllib2.Request(url, json.dumps(data), header)
try:
response = urllib2.urlopen(request)
except urllib2.URLError as e:
raise errors.CannotMakeZabbixRequest(
"Can't make a request to Zabbix: {0}".format(e)
)
result = json.loads(response.read())
logger.debug("Zabbix response: %s", result)
if 'error' in result:
code = result['error']['code']
msg = result['error']['message']
data = result['error'].get('data', '')
raise errors.ZabbixRequestError(
"Zabbix returned error code {0}, {1}: {2}".format(
code, msg, data
)
)
return result['result']
def update_slave_nodes_fqdn(cls, nodes):
for n in nodes:
fqdn = cls.make_slave_fqdn(n.id)
if n.fqdn != fqdn:
n.fqdn = fqdn
logger.debug("Updating node fqdn: %s %s", n.id, n.fqdn)
db().commit()
def execute(self, task, data):
task_uuid = task.uuid
nodes = []
for n in task.cluster.nodes:
node_json = {'uid': n.id, 'networks': []}
for nic in n.interfaces:
vlans = []
for ng in nic.assigned_networks:
# Handle FuelWeb admin network first.
if not ng.cluster_id:
vlans.append(0)
continue
data_ng = filter(
lambda i: i['name'] == ng.name,
data
)[0]
vlans.extend(data_ng['vlans'])
if not vlans:
continue
node_json['networks'].append(
{'iface': nic.name, 'vlans': vlans}
)
nodes.append(node_json)
message = {'method': 'verify_networks',
'respond_to': 'verify_networks_resp',
'args': {'task_uuid': task.uuid,
'nodes': nodes}}
logger.debug("Network verification is called with: %s", message)
task.cache = message
orm().add(task)
orm().commit()
rpc.cast('naily', message)
def execute(self, nodes_to_deployment):
TaskHelper.update_slave_nodes_fqdn(nodes_to_deployment)
logger.debug('Nodes to deploy: {0}'.format(
' '.join([n.fqdn for n in nodes_to_deployment])))
task_deployment = Task(name='deployment', cluster=self.cluster)
db().add(task_deployment)
db().commit()
deployment_message = self._call_silently(
task_deployment,
tasks.DeploymentTask,
nodes_to_deployment,
method_name='message')
db().refresh(task_deployment)
task_deployment.cache = deployment_message
for node in nodes_to_deployment:
node.status = 'deploying'
node.progress = 0
db().commit()
rpc.cast('naily', deployment_message)
return task_deployment
def _traverse(self, cdict):
new_dict = {}
if isinstance(cdict, dict):
for i, val in cdict.iteritems():
if type(val) in (str, unicode, int, float):
new_dict[i] = val
elif isinstance(val, dict):
if "generator" in val:
logger.debug("Generating value: generator: %s "
"generator_args: %s", val["generator"],
val.get("generator_args", []))
genval = self.field_generator(
val["generator"],
*(val.get("generator_args", []))
)
logger.debug("Generated value: %s", str(genval))
new_dict[i] = genval
else:
new_dict[i] = self._traverse(val)
elif isinstance(val, list):
new_dict[i] = []
for d in val:
new_dict[i].append(self._traverse(d))
elif isinstance(cdict, list):
new_dict = []
for d in cdict:
new_dict.append(self._traverse(d))
return new_dict
def execute(self, nodes_to_provision):
"""Run provisioning task on specified nodes
"""
# locking nodes
nodes_ids = [node.id for node in nodes_to_provision]
nodes = objects.NodeCollection.filter_by_list(None, "id", nodes_ids, order_by="id")
objects.NodeCollection.lock_for_update(nodes).all()
objects.NodeCollection.update_slave_nodes_fqdn(nodes_to_provision)
logger.debug("Nodes to provision: {0}".format(" ".join([n.fqdn for n in nodes_to_provision])))
task_provision = Task(name="provision")
task_provision.node_ids = nodes_ids
# node_ids参数在安装成功时候无用,但在安装失败的时候需要用到
db().add(task_provision)
db().commit()
provision_message = self._call_silently(
task_provision, tasks.InstallosTask, nodes_to_provision, method_name="message"
)
task_provision = objects.Task.get_by_uid(task_provision.id, fail_if_not_found=True, lock_for_update=True)
task_provision.cache = provision_message
objects.NodeCollection.lock_for_update(nodes).all()
for node in nodes_to_provision:
node.pending_addition = False
node.status = NODE_STATUSES.provisioning
node.progress = 0
db().commit()
rpc.cast("naily", provision_message)
logger.info(u"消息发送完毕")
return task_provision
def update_primary_roles(cls, instance, new_primary_roles):
"""Update primary_roles for Node instance.
Logs an error if node doesn't belong to Cluster
:param instance: Node instance
:param new_primary_roles: list of new pending role names
:returns: None
"""
if not instance.cluster_id:
logger.warning(
u"Attempting to assign pending roles to node "
u"'{0}' which isn't added to cluster".format(
instance.full_name))
return
assigned_roles = set(instance.roles + instance.pending_roles)
for role in new_primary_roles:
if role not in assigned_roles:
logger.warning(
u"Could not mark node {0} as primary for {1} role, "
u"because there's no assigned {1} role.".format(
instance.full_name, role)
)
return
logger.debug(
u"Updating primary roles for node {0}: {1}".format(
instance.full_name,
new_primary_roles))
instance.primary_roles = new_primary_roles
db().flush()
def execute(cls, task):
logger.debug("ProvisionTask.execute(task=%s)" % task.uuid)
message = cls.message(task)
task.cache = message
db().add(task)
db().commit()
rpc.cast('naily', message)
def message(cls, task):
logger.debug("DeploymentTask.message(task=%s)" % task.uuid)
task.cluster.prepare_for_deployment()
nodes = TaskHelper.nodes_to_deploy(task.cluster)
nodes_ids = [n.id for n in nodes]
for n in db().query(Node).filter_by(
cluster=task.cluster).order_by(Node.id):
# However, we must not pass nodes which are set to be deleted.
if n.pending_deletion:
continue
if n.id in nodes_ids: # It's node which we need to redeploy
n.pending_addition = False
if n.pending_roles:
n.roles += n.pending_roles
n.pending_roles = []
if n.status in ('deploying'):
n.status = 'provisioned'
n.progress = 0
db().add(n)
db().commit()
# here we replace provisioning data if user redefined them
serialized_cluster = task.cluster.replaced_deployment_info or \
deployment_serializers.serialize(task.cluster)
return {
'method': 'deploy',
'respond_to': 'deploy_resp',
'args': {
'task_uuid': task.uuid,
'deployment_info': serialized_cluster}}
def update_verify_networks(cls, uuid, status,
progress, msg, result):
#TODO(dshulyak) move network tests into ostf
task = db().query(Task).filter_by(uuid=uuid).first()
if not task:
logger.error("Can't set status='%s', message='%s': No task \
with UUID %s found!", status, msg, uuid)
return
previous_status = task.status
statuses = [sub.status for sub in task.subtasks]
messages = [sub.message for sub in task.subtasks]
messages.append(msg)
statuses.append(status)
if any(st == 'error' for st in statuses):
task.status = 'error'
else:
task.status = status or task.status
task.progress = progress or task.progress
task.result = result or task.result
# join messages if not None or ""
task.message = '\n'.join([m for m in messages if m])
db().commit()
if previous_status != task.status and task.cluster_id:
logger.debug("Updating cluster status: "
"cluster_id: %s status: %s",
task.cluster_id, status)
cls.update_cluster_status(uuid)
def update_pending_roles(cls, instance, new_pending_roles):
"""Update pending_roles for Node instance.
Logs an error if node doesn't belong to Cluster
:param instance: Node instance
:param new_pending_roles: list of new pending role names
:returns: None
"""
if not instance.cluster_id:
logger.warning(
u"Attempting to assign pending roles to node "
u"'{0}' which isn't added to cluster".format(
instance.full_name))
return
logger.debug(
u"Updating pending roles for node {0}: {1}".format(
instance.full_name,
new_pending_roles))
if new_pending_roles == []:
# TODO(enchantner): research why the hell we need this
Cluster.clear_pending_changes(
instance.cluster,
node_id=instance.id
)
instance.pending_roles = new_pending_roles
db().flush()
def make_bootable(self):
logger.debug("Allocating /boot partition")
self.create_partition(
"/boot",
self.vm.field_generator("calc_boot_size")
)
self.create_mbr(True)
def execute(self, nodes_to_provision, **kwargs):
"""Run provisioning task on specified nodes."""
logger.debug('Nodes to provision: {0}'.format(
' '.join([objects.Node.get_node_fqdn(n)
for n in nodes_to_provision])))
self.check_running_task()
task_provision = Task(name=consts.TASK_NAMES.provision,
status=consts.TASK_STATUSES.pending,
cluster=self.cluster)
db().add(task_provision)
# update cluster status
self.cluster.status = consts.CLUSTER_STATUSES.deployment
db().commit()
nodes_ids_to_provision = [node.id for node in nodes_to_provision]
# perform async call of _execute_async
mule.call_task_manager_async(
self.__class__,
'_execute_async',
self.cluster.id,
task_provision.id,
nodes_ids_to_provision=nodes_ids_to_provision,
**kwargs
)
return task_provision
def update_by_agent(cls, instance, data):
"""Update Node instance with some specific cases for agent.
* don't update provisioning or error state back to discover
* don't update volume information if disks arrays is empty
:param data: dictionary of key-value pairs as object fields
:returns: Node instance
"""
# don't update provisioning and error back to discover
if instance.status in ('provisioning', 'error'):
if data.get('status', 'discover') == 'discover':
logger.debug(
u"Node {0} has provisioning or error status - "
u"status not updated by agent".format(
instance.human_readable_name
)
)
data['status'] = instance.status
# don't update volume information, if agent has sent an empty array
meta = data.get('meta', {})
if meta and len(meta.get('disks', [])) == 0 \
and instance.meta.get('disks'):
logger.warning(
u'Node {0} has received an empty disks array - '
u'volume information will not be updated'.format(
instance.human_readable_name
)
)
meta['disks'] = instance.meta['disks']
return cls.update(instance, data)
def execute(self):
logger.debug("Creating release dowload task")
task = Task(name="download_release")
db().add(task)
db().commit()
self._call_silently(task, tasks.DownloadReleaseTask, self.release_data)
return task
def execute(self, nodes_to_provision):
"""Run provisioning task on specified nodes
"""
objects.NodeCollection.update_slave_nodes_fqdn(nodes_to_provision)
logger.debug('Nodes to provision: {0}'.format(
' '.join([n.fqdn for n in nodes_to_provision])))
task_provision = Task(name='provision', cluster=self.cluster)
db().add(task_provision)
db().commit()
provision_message = self._call_silently(
task_provision,
tasks.ProvisionTask,
nodes_to_provision,
method_name='message'
)
db().refresh(task_provision)
task_provision.cache = provision_message
for node in nodes_to_provision:
node.pending_addition = False
node.status = 'provisioning'
node.progress = 0
db().commit()
rpc.cast('naily', provision_message)
return task_provision
def assign_admin_ips(cls, node_id, num=1):
"""Method for assigning admin IP addresses to nodes.
:param node_id: Node database ID.
:type node_id: int
:param num: Number of IP addresses for node.
:type num: int
:returns: None
"""
admin_net_id = cls.get_admin_network_group_id()
node_admin_ips = db().query(IPAddr).filter_by(
node=node_id,
network=admin_net_id
).all()
if not node_admin_ips or len(node_admin_ips) < num:
admin_net = db().query(NetworkGroup).get(admin_net_id)
logger.debug(
u"Trying to assign admin ips: node=%s count=%s",
node_id,
num - len(node_admin_ips)
)
free_ips = cls.get_free_ips(
admin_net.id,
num=num - len(node_admin_ips)
)
logger.info(len(free_ips))
for ip in free_ips:
ip_db = IPAddr(
node=node_id,
ip_addr=ip,
network=admin_net_id
)
db().add(ip_db)
db().commit()
def prepare_for_deployment(cls, instances):
"""Prepare environment for deployment. Assign IPs for all networks."""
if not instances:
logger.debug("prepare_for_deployment was called with no instances")
return
cluster = instances[0].cluster
netmanager = Cluster.get_network_manager(cluster)
if cluster.network_config.configuration_template:
return netmanager.assign_ips_for_nodes_w_template(
cluster, instances)
nodes_by_id = dict((n.id, n) for n in instances)
query = (db().query(
models.Node.id, models.NetworkGroup.id, models.NetworkGroup.name,
models.NetworkGroup.meta).join(models.NodeGroup.nodes).join(
models.NodeGroup.networks).filter(
models.NodeGroup.cluster_id == cluster.id,
models.NetworkGroup.name !=
consts.NETWORKS.fuelweb_admin).order_by(
models.NetworkGroup.id))
# Group by NetworkGroup.id
for key, items in itertools.groupby(query, operator.itemgetter(1)):
items = list(items)
network_name = items[0][2]
network_metadata = items[0][3]
if not network_metadata.get('notation'):
continue
nodes = [
nodes_by_id[item[0]] for item in items
if item[0] in nodes_by_id
]
netmanager.assign_ips(nodes, network_name)
netmanager.assign_admin_ips(instances)
def create_networks(self, nw_group):
'''
Method for creation of networks for network group.
:param nw_group: NetworkGroup object.
:type nw_group: NetworkGroup
:returns: None
'''
fixnet = IPNetwork(nw_group.cidr)
subnet_bits = int(math.ceil(math.log(nw_group.network_size, 2)))
logger.debug("Specified network size requires %s bits", subnet_bits)
subnets = list(fixnet.subnet(32 - subnet_bits, count=nw_group.amount))
logger.debug("Base CIDR sliced on subnets: %s", subnets)
for net in nw_group.networks:
logger.debug("Deleting old network with id=%s, cidr=%s", net.id,
net.cidr)
ips = db().query(IPAddr).filter(IPAddr.network == net.id).all()
map(db().delete, ips)
db().delete(net)
db().commit()
# Dmitry's hack for clearing VLANs without networks
self.clear_vlans()
db().commit()
nw_group.networks = []
for n in xrange(nw_group.amount):
vlan_id = None
if nw_group.vlan_start is not None:
vlan_db = db().query(Vlan).get(nw_group.vlan_start + n)
if vlan_db:
logger.warning("Intersection with existing vlan_id: %s",
vlan_db.id)
else:
vlan_db = Vlan(id=nw_group.vlan_start + n)
db().add(vlan_db)
vlan_id = vlan_db.id
logger.debug("Created VLAN object, vlan_id=%s", vlan_id)
gateway = None
if nw_group.gateway:
gateway = nw_group.gateway
net_db = Network(release=nw_group.release,
name=nw_group.name,
access=nw_group.access,
cidr=str(subnets[n]),
vlan_id=vlan_id,
gateway=gateway,
network_group_id=nw_group.id)
db().add(net_db)
db().commit()
def __init__(self):
settings_files = []
logger.debug("Looking for settings.yaml package config "
"using old style __file__")
project_path = os.path.dirname(__file__)
project_settings_file = os.path.join(project_path, 'settings.yaml')
settings_files.append(project_settings_file)
settings_files.append('/etc/nailgun/settings.yaml')
settings_files.append('/etc/nailgun/version.yaml')
self.config = {}
for sf in settings_files:
try:
logger.debug("Trying to read config file %s" % sf)
self.update_from_file(sf)
except Exception as e:
logger.debug("Error while reading config file %s: %s" %
(sf, str(e)))
if int(self.config.get("DEVELOPMENT")):
logger.info("DEVELOPMENT MODE ON:")
here = os.path.abspath(
os.path.join(os.path.dirname(__file__), '..'))
self.config.update({
'STATIC_DIR': os.path.join(here, 'static'),
'TEMPLATE_DIR': os.path.join(here, 'static')
})
logger.info("Static dir is %s" % self.config.get("STATIC_DIR"))
logger.info("Template dir is %s" % self.config.get("TEMPLATE_DIR"))
def execute(self, nodes_to_deployment, deployment_tasks=None):
deployment_tasks = deployment_tasks or []
# locking nodes for update
objects.NodeCollection.lock_nodes(nodes_to_deployment)
objects.NodeCollection.update_slave_nodes_fqdn(nodes_to_deployment)
logger.debug('Nodes to deploy: {0}'.format(' '.join(
[n.fqdn for n in nodes_to_deployment])))
task_deployment = Task(name=consts.TASK_NAMES.deployment,
cluster=self.cluster)
db().add(task_deployment)
deployment_message = self._call_silently(
task_deployment,
tasks.DeploymentTask,
nodes_to_deployment,
deployment_tasks=deployment_tasks,
method_name='message')
db().refresh(task_deployment)
# locking task
task_deployment = objects.Task.get_by_uid(task_deployment.id,
fail_if_not_found=True,
lock_for_update=True)
# locking nodes
objects.NodeCollection.lock_nodes(nodes_to_deployment)
task_deployment.cache = deployment_message
for node in nodes_to_deployment:
node.status = 'deploying'
node.progress = 0
db().commit()
rpc.cast('naily', deployment_message)
return task_deployment
def check_before_deployment(self, supertask):
# checking admin intersection with untagged
if self.cluster.net_provider == 'nova_network':
net_serializer = NovaNetworkConfigurationSerializer
elif self.cluster.net_provider == 'neutron':
net_serializer = NeutronNetworkConfigurationSerializer
network_info = net_serializer.serialize_for_cluster(self.cluster)
network_info["networks"] = [
n for n in network_info["networks"] if n["name"] != "fuelweb_admin"
]
check_networks = supertask.create_subtask('check_networks')
self._call_silently(check_networks,
tasks.CheckNetworksTask,
data=network_info,
check_admin_untagged=True)
db().refresh(check_networks)
if check_networks.status == 'error':
logger.warning("Checking networks failed: %s",
check_networks.message)
raise errors.CheckBeforeDeploymentError(check_networks.message)
db().delete(check_networks)
db().commit()
# checking prerequisites
check_before = supertask.create_subtask('check_before_deployment')
logger.debug("Checking prerequisites task: %s", check_before.uuid)
self._call_silently(check_before, tasks.CheckBeforeDeploymentTask)
db().refresh(check_before)
# if failed to check prerequisites
# then task is already set to error
if check_before.status == 'error':
logger.warning("Checking prerequisites failed: %s",
check_before.message)
raise errors.CheckBeforeDeploymentError(check_before.message)
logger.debug(
"Checking prerequisites is successful, starting deployment...")
db().delete(check_before)
db().commit()
def update_task_status(cls, uuid, status, progress, msg="", result=None):
logger.debug("Updating task: %s", uuid)
task = db().query(Task).filter_by(uuid=uuid).first()
if not task:
logger.error(
"Can't set status='%s', message='%s':no task \
with UUID %s found!", status, msg, uuid)
return
previous_status = task.status
data = {
'status': status,
'progress': progress,
'message': msg,
'result': result
}
for key, value in data.iteritems():
if value is not None:
setattr(task, key, value)
logger.info(u"Task {0} ({1}) {2} is set to {3}".format(
task.uuid, task.name, key, value))
db().add(task)
db().commit()
if previous_status != status and task.cluster_id:
logger.debug(
"Updating cluster status: "
"cluster_id: %s status: %s", task.cluster_id, status)
cls.update_cluster_status(uuid)
if task.parent:
logger.debug("Updating parent task: %s.", task.parent.uuid)
cls.update_parent_task(task.parent.uuid)
def create_pv(self, volume_info, size=None):
"""Allocates all available space if size is None.
Size in parameter should include size of lvm meta
"""
name = volume_info['id']
logger.debug('Creating PV: disk=%s vg=%s, size=%s', self.id, name,
str(size))
if size is None:
logger.debug(
'Size is not defined. Will use all free space on this disk.')
size = self.free_space
self.free_space -= size
# Don't allocate lvm if size equal 0
lvm_meta_size = self.get_lvm_meta_from_pool() if size else 0
logger.debug('Appending PV to volumes.')
self.volumes.append({
'type': 'pv',
'vg': name,
'size': size + lvm_meta_size,
'lvm_meta_size': lvm_meta_size
})
def execute(self, nodes_to_provision):
"""Run provisioning task on specified nodes
"""
# locking nodes
nodes_ids = [node.id for node in nodes_to_provision]
nodes = objects.NodeCollection.filter_by_list(None,
'id',
nodes_ids,
order_by='id')
objects.NodeCollection.lock_for_update(nodes).all()
objects.NodeCollection.update_slave_nodes_fqdn(nodes_to_provision)
logger.debug('Nodes to provision: {0}'.format(' '.join(
[n.fqdn for n in nodes_to_provision])))
task_provision = Task(name='provision', cluster=self.cluster)
db().add(task_provision)
db().commit()
provision_message = self._call_silently(task_provision,
tasks.ProvisionTask,
nodes_to_provision,
method_name='message')
task_provision = objects.Task.get_by_uid(task_provision.id,
fail_if_not_found=True,
lock_for_update=True)
task_provision.cache = provision_message
objects.NodeCollection.lock_for_update(nodes).all()
for node in nodes_to_provision:
node.pending_addition = False
node.status = NODE_STATUSES.provisioning
node.progress = 0
db().commit()
rpc.cast('naily', provision_message)
return task_provision
def update_pending_roles(cls, instance, new_pending_roles):
"""Update pending_roles for Node instance.
Logs an error if node doesn't belong to Cluster
:param instance: Node instance
:param new_pending_roles: list of new pending role names
:returns: None
"""
if not instance.cluster_id:
logger.warning(
u"Attempting to assign pending roles to node "
u"'{0}' which isn't added to cluster".format(
instance.name or instance.id
)
)
return
logger.debug(
u"Updating pending roles for node {0}: {1}".format(
instance.id,
new_pending_roles
)
)
if new_pending_roles == []:
instance.pending_role_list = []
#TODO(enchantner): research why the hell we need this
Cluster.clear_pending_changes(
instance.cluster,
node_id=instance.id
)
else:
instance.pending_role_list = db().query(models.Role).filter_by(
release_id=instance.cluster.release_id,
).filter(
models.Role.name.in_(new_pending_roles)
).all()
db().flush()
db().refresh(instance)
def task_in_orchestrator(cls, **kwargs):
logger.info("RPC method task_in_orchestrator received: %s",
jsonutils.dumps(kwargs))
task_uuid = kwargs.get('task_uuid')
try:
task = objects.Task.get_by_uuid(task_uuid,
fail_if_not_found=True,
lock_for_update=True)
if task.status == consts.TASK_STATUSES.pending:
objects.Task.update(task,
{'status': consts.TASK_STATUSES.running})
logger.debug("Task '%s' is acknowledged as running", task_uuid)
else:
logger.debug(
"Task '%s' in status '%s' can not "
"be acknowledged as running", task_uuid, task.status)
except nailgun_errors.ObjectNotFound:
logger.warning(
"Task '%s' acknowledgement as running failed "
"due to task doesn't exist in DB", task_uuid)
def update_verify_networks(cls, instance, status,
progress, msg, result):
#TODO(dshulyak) move network tests into ostf
previous_status = instance.status
statuses = [sub.status for sub in instance.subtasks]
messages = [sub.message for sub in instance.subtasks]
messages.append(msg)
statuses.append(status)
if any(st == 'error' for st in statuses):
instance.status = 'error'
else:
instance.status = status or instance.status
instance.progress = progress or instance.progress
instance.result = result or instance.result
# join messages if not None or ""
instance.message = '\n'.join([m for m in messages if m])
if previous_status != instance.status and instance.cluster_id:
logger.debug("Updating cluster status: "
"cluster_id: %s status: %s",
instance.cluster_id, status)
cls._update_cluster_data(instance)
def execute(self):
# locking required tasks
locked_tasks = objects.TaskCollection.lock_cluster_tasks(
self.cluster.id)
# locking cluster
objects.Cluster.get_by_uid(self.cluster.id,
fail_if_not_found=True,
lock_for_update=True)
# locking nodes
nodes = objects.NodeCollection.filter_by(None,
cluster_id=self.cluster.id)
nodes = objects.NodeCollection.order_by(nodes, 'id')
objects.NodeCollection.lock_for_update(nodes).all()
current_cluster_tasks = objects.TaskCollection.filter_by_list(
locked_tasks, 'name', (consts.TASK_NAMES.cluster_deletion, ))
deploy_running = objects.TaskCollection.filter_by(
None,
cluster_id=self.cluster.id,
name=consts.TASK_NAMES.deploy,
status=consts.TASK_STATUSES.running)
deploy_running = objects.TaskCollection.order_by(deploy_running,
'id').first()
if deploy_running:
logger.error(u"Deleting cluster '{0}' "
"while deployment is still running".format(
self.cluster.name))
# Updating action logs for deploy task
TaskHelper.set_ready_if_not_finished(deploy_running)
logger.debug("Removing cluster tasks")
for task in current_cluster_tasks:
if task.status == consts.TASK_STATUSES.running:
db().rollback()
raise errors.DeletionAlreadyStarted()
elif task.status in (consts.TASK_STATUSES.ready,
consts.TASK_STATUSES.error):
for subtask in task.subtasks:
db().delete(subtask)
db().delete(task)
db().flush()
logger.debug("Labeling cluster nodes to delete")
for node in self.cluster.nodes:
node.pending_deletion = True
db().add(node)
db().flush()
self.cluster.status = consts.CLUSTER_STATUSES.remove
db().add(self.cluster)
logger.debug("Creating cluster deletion task")
task = Task(name=consts.TASK_NAMES.cluster_deletion,
cluster=self.cluster)
db().add(task)
db().commit()
self._call_silently(task, tasks.ClusterDeletionTask)
return task
def process_task(self, task, node_ids):
"""Processes one task one nodes of cluster.
:param task: the task instance
:param node_ids: the list of nodes, where this tasks should run
"""
logger.debug("applying task '%s' for nodes: %s", task['id'], node_ids)
task_serializer = self.factory.create_serializer(task)
for node_id in node_ids:
try:
task = task_serializer.serialize(node_id)
except Exception:
logger.exception("Failed to serialize task %s", task['id'])
raise
node_tasks = self.tasks_graph.setdefault(node_id, {})
# de-duplication the tasks on node
# since task can be added after expand group need to
# overwrite if existed task is skipped and new is not skipped.
if self.need_update_task(node_tasks, task):
node_tasks[task['id']] = task
def remove_cluster_resp(cls, **kwargs):
logger.info("RPC method remove_cluster_resp received: %s" % kwargs)
task_uuid = kwargs.get('task_uuid')
cls.remove_nodes_resp(**kwargs)
task = cls.db.query(Task).filter_by(uuid=task_uuid).first()
cluster = task.cluster
if task.status in ('ready', ):
logger.debug("Removing environment itself")
cluster_name = cluster.name
nws = itertools.chain(
*[n.networks for n in cluster.network_groups])
ips = cls.db.query(IPAddr).filter(
IPAddr.network.in_([n.id for n in nws]))
map(cls.db.delete, ips)
cls.db.commit()
cls.db.delete(cluster)
cls.db.commit()
# Dmitry's hack for clearing VLANs without networks
cls.network_manager.clear_vlans()
notifier.notify(
"done", u"Environment '%s' and all its nodes are deleted" %
(cluster_name))
elif task.status in ('error', ):
cluster.status = 'error'
cls.db.add(cluster)
cls.db.commit()
if not task.message:
task.message = "Failed to delete nodes:\n{0}".format(
cls._generate_error_message(task,
error_types=('deletion', )))
notifier.notify("error", task.message, cluster.id)
def clear_pending_changes(cls, instance, node_id=None):
"""Clear pending changes for current Cluster.
If node_id is specified then only clears changes connected
to this node.
:param instance: Cluster instance
:param node_id: node id for changes
:returns: None
"""
logger.debug(
u"Removing pending changes in environment {0}{1}".format(
instance.id,
u" where node_id={0}".format(node_id) if node_id else u""
)
)
chs = db().query(models.ClusterChanges).filter_by(
cluster_id=instance.id
)
if node_id:
chs = chs.filter_by(node_id=node_id)
map(db().delete, chs.all())
db().flush()
def process_deployment(cls, data, cluster, nodes, **kwargs):
"""Find and replace cpu pinning parameters in deployment data,
this includes changing nova hash, nodes hash on every node and
contrail plugin parameters.
"""
nodes_data = [
node_data for node_data in data if node_data['uid'] != 'master'
]
pinning_nodes = [
node_data['uid'] for node_data in nodes_data
if CpuPinOverride.get_by_uid(node_data['uid'])
]
logger.debug(pinning_nodes)
for node_data in nodes_data:
pins_data = CpuPinOverride.get_by_uid(node_data['uid'])
if pins_data:
# Setting nova cores and kernel params
node_data['nova']['cpu_pinning'] = pins_data.nova_cores
kparams = node_data['kernel_params']['kernel']
newkparams = PinningOverridePipeline._generate_kernel_params(
kparams.split(), pins_data)
node_data['kernel_params']['kernel'] = newkparams
node_data['release']['attributes_metadata']['editable'][
'kernel_params']['kernel']['value'] = newkparams
# Setting contrail vrouter coremask
if pins_data.vrouter_cores and 'dpdk' in node_data['roles']:
pins_str = ','.join(pins_data.vrouter_cores)
# vn637v. Concatenate pins + empty string to convert value from
# FixNum to String
node_data['contrail']['vrouter_core_mask'] = ' ' + pins_str
# Overriding network_metadata['nodes'] hash on all nodes
for nm_val in node_data['network_metadata']['nodes'].values():
if nm_val['uid'] in pinning_nodes:
nm_val['nova_cpu_pinning_enabled'] = True
logger.debug('Overriding CPU pinning values in deployment data')
return data
def serialize(self, node_id):
logger.debug(
"serialize task %s for node %s",
self.task_template['id'], node_id
)
task = utils.traverse(
self.task_template,
utils.text_format_safe,
self.context.get_formatter_context(node_id),
{
'yaql_exp': self.context.get_yaql_interpreter(node_id)
}
)
if not self.should_execute(task, node_id):
logger.debug(
"Task %s is skipped by condition.", self.task_template['id']
)
return super(DefaultTaskSerializer, self).serialize(node_id)
task.setdefault('parameters', {}).setdefault('cwd', '/')
task.setdefault('fail_on_error', True)
return self.get_required_fields(task)
def _get_allowed_nodes_statuses(self, context):
"""Extracts node statuses that allows distributed serialization"""
common = context.new.get('common', {})
cluster = common.get('cluster', {})
logger.debug(
"Getting allowed nodes statuses to use as serialization "
"workers for cluster %s", cluster.get('id'))
check_fields = {
'ds_use_ready': consts.NODE_STATUSES.ready,
'ds_use_provisioned': consts.NODE_STATUSES.provisioned,
'ds_use_discover': consts.NODE_STATUSES.discover,
'ds_use_error': consts.NODE_STATUSES.error
}
statuses = set()
for field, node_status in check_fields.items():
if common.get(field):
statuses.add(node_status)
logger.debug(
"Allowed nodes statuses to use as serialization workers "
"for cluster %s are: %s", cluster.get('id'), statuses)
return statuses
def add_pending_changes(cls, instance, changes_type, node_id=None):
"""Add pending changes for current Cluster.
If node_id is specified then links created changes with node.
:param instance: Cluster instance
:param changes_type: name of changes to add
:param node_id: node id for changes
:returns: None
"""
logger.debug(
u"New pending changes in environment {0}: {1}{2}".format(
instance.id,
changes_type,
u" node_id={0}".format(node_id) if node_id else u""
)
)
# TODO(enchantner): check if node belongs to cluster
ex_chs = db().query(models.ClusterChanges).filter_by(
cluster=instance,
name=changes_type
)
if not node_id:
ex_chs = ex_chs.first()
else:
ex_chs = ex_chs.filter_by(node_id=node_id).first()
# do nothing if changes with the same name already pending
if ex_chs:
return
ch = models.ClusterChanges(
cluster_id=instance.id,
name=changes_type
)
if node_id:
ch.node_id = node_id
db().add(ch)
db().flush()
def detach_from_model(cls, instance, graph_type=None):
"""Detach existing deployment graph to given model if it exists.
:param instance: model that should have relation to graph
:type instance: models.Plugin|models.Cluster|models.Release|
:param graph_type: graph type
:type graph_type: basestring
:returns: if graph was detached
:rtype: bool
"""
if graph_type is None:
graph_type = consts.DEFAULT_DEPLOYMENT_GRAPH_TYPE
existing_graph = cls.get_for_model(instance, graph_type)
if existing_graph:
association = cls.get_association_for_model(instance)
instance.deployment_graphs_assoc.filter(
association.type == graph_type).delete()
db().flush()
logger.debug(
'Graph with ID={0} was detached from model {1} with ID={2}'.
format(existing_graph.id, instance, instance.id))
return existing_graph
def message(cls, task):
logger.debug("DeploymentTask.message(task=%s)" % task.uuid)
task.cluster.prepare_for_deployment()
nodes = TaskHelper.nodes_to_deploy(task.cluster)
nodes_ids = [n.id for n in nodes]
for n in db().query(Node).filter_by(cluster=task.cluster).order_by(
Node.id):
# However, we must not pass nodes which are set to be deleted.
if n.pending_deletion:
continue
if n.id in nodes_ids:
if n.pending_roles:
n.roles += n.pending_roles
n.pending_roles = []
if n.status in ('deploying'):
n.status = 'provisioned'
n.progress = 0
db().add(n)
db().commit()
# here we replace provisioning data if user redefined them
serialized_cluster = task.cluster.replaced_deployment_info or \
deployment_serializers.serialize(task.cluster)
# After searilization set pending_addition to False
for node in db().query(Node).filter(Node.id.in_(nodes_ids)):
node.pending_addition = False
db().commit()
return {
'method': 'deploy',
'respond_to': 'deploy_resp',
'args': {
'task_uuid': task.uuid,
'deployment_info': serialized_cluster
}
}
def serialize(self, node_id, formatter_context=None):
"""Serialize task in expected by orchestrator format
If serialization is performed on the remote worker
we should pass formatter_context parameter with values
from the master node settings
:param formatter_context: formatter context
:param node_id: the target node_id
"""
logger.debug("serialize task %s for node %s", self.task_template['id'],
node_id)
formatter_context = formatter_context \
or self.context.get_formatter_context(node_id)
task = utils.traverse(
self.task_template, utils.text_format_safe, formatter_context, {
'yaql_exp':
self.context.get_yaql_interpreter(node_id,
self.task_template['id'])
})
return self.normalize(self.finalize(task, node_id))
def execute(self, nodes_to_provision):
"""Run provisioning task on specified nodes
Constraints: currently this task cannot deploy RedHat.
For redhat here should be added additional
tasks e.i. check credentials, check licenses,
redhat downloading.
Status of this task you can track here:
https://blueprints.launchpad.net/fuel/+spec
/nailgun-separate-provisioning-for-redhat
"""
TaskHelper.update_slave_nodes_fqdn(nodes_to_provision)
logger.debug('Nodes to provision: {0}'.format(
' '.join([n.fqdn for n in nodes_to_provision])))
task_provision = Task(name='provision', cluster=self.cluster)
db().add(task_provision)
db().commit()
provision_message = self._call_silently(
task_provision,
tasks.ProvisionTask,
nodes_to_provision,
method_name='message'
)
db().refresh(task_provision)
task_provision.cache = provision_message
for node in nodes_to_provision:
node.pending_addition = False
node.status = 'provisioning'
node.progress = 0
db().commit()
rpc.cast('naily', provision_message)
return task_provision
def cast(name, message, service=False):
logger.debug("RPC cast to orchestrator:\n{0}".format(
jsonutils.dumps(message, indent=4)))
use_queue = naily_queue if not service else naily_service_queue
use_exchange = naily_exchange if not service else naily_service_exchange
with Connection(conn_str) as conn:
with conn.Producer(serializer='json') as producer:
publish = functools.partial(producer.publish,
message,
exchange=use_exchange,
routing_key=name,
declare=[use_queue])
try:
publish()
except amqp_exceptions.PreconditionFailed as e:
logger.warning(six.text_type(e))
# (dshulyak) we should drop both exchanges/queues in order
# for astute to be able to recover temporary queues
utils.delete_entities(conn, naily_service_exchange,
naily_service_queue, naily_exchange,
naily_queue)
publish()
def _execute_sync(self, sub_transaction):
cluster = sub_transaction.cluster
graph = objects.Cluster.get_deployment_graph(
cluster, sub_transaction.graph_type)
nodes = _get_nodes_to_run(cluster, graph.get('node_filter'),
sub_transaction.cache.get('nodes'))
logger.debug("execute graph %s on nodes %s",
sub_transaction.graph_type, [n.id for n in nodes])
# we should initialize primary roles for cluster before
# role resolve has been created
objects.Cluster.set_primary_tags(cluster, nodes)
resolver = resolvers.TagResolver(nodes)
_adjust_graph_tasks(graph, cluster, resolver,
sub_transaction.cache.get('tasks'))
context = lcm.TransactionContext(
_get_expected_state(cluster, nodes),
_get_current_state(cluster, nodes, graph['tasks'],
sub_transaction.cache.get('force')))
_prepare_nodes(nodes, sub_transaction.dry_run, context.new['nodes'])
# Attach desired state to the sub transaction, so when we continue
# our top-level transaction, the new state will be calculated on
# top of this.
_dump_expected_state(sub_transaction, context.new, graph['tasks'])
message = make_astute_message(sub_transaction, context, graph,
resolver)
objects.Transaction.on_start(sub_transaction)
helpers.TaskHelper.create_action_log(sub_transaction)
# Once rpc.cast() is called, the message is sent to Astute. By
# that moment all transaction instanced must exist in database,
# otherwise we may get wrong result due to RPC receiver won't
# found entry to update.
db().commit()
rpc.cast('naily', [message])
def message(cls, task, nodes):
logger.debug("DeploymentTask.message(task=%s)" % task.uuid)
nodes_ids = [n.id for n in nodes]
for n in db().query(Node).filter_by(cluster=task.cluster).order_by(
Node.id):
if n.id in nodes_ids:
if n.pending_roles:
n.roles += n.pending_roles
n.pending_roles = []
# If reciever for some reasons didn't update
# node's status to provisioned when deployment
# started, we should do it in nailgun
if n.status in ('deploying'):
n.status = 'provisioned'
n.progress = 0
db().add(n)
db().commit()
# here we replace provisioning data if user redefined them
serialized_cluster = task.cluster.replaced_deployment_info or \
deployment_serializers.serialize(task.cluster, nodes)
# After searilization set pending_addition to False
for node in nodes:
node.pending_addition = False
db().commit()
return {
'method': 'deploy',
'respond_to': 'deploy_resp',
'args': {
'task_uuid': task.uuid,
'deployment_info': serialized_cluster
}
}
def resolve(self, roles, policy=None):
result = set()
if roles == consts.TASK_ROLES.all:
# optimization
result = {
uid
for nodes in six.itervalues(self.__mapping) for uid in nodes
}
else:
if isinstance(roles, six.string_types):
roles = [roles]
if not isinstance(roles, (list, tuple, set)):
# TODO(bgaifullin) fix wrong format for roles in tasks.yaml
# After it will be allowed to raise exception here
logger.warn(
'Wrong roles format, `roles` should be a list or "*": %s',
roles)
return result
for role in roles:
if role in self.SPECIAL_ROLES:
result.update(self.SPECIAL_ROLES[role])
else:
pattern = NameMatchingPolicy.create(role)
for node_role, nodes_ids in six.iteritems(self.__mapping):
if pattern.match(node_role):
result.update(nodes_ids)
# in some cases need only one any node from pool
# for example if need only one any controller.
# to distribute load select first node from pool
if result and policy == consts.NODE_RESOLVE_POLICY.any:
result = {next(iter(result))}
logger.debug("Role '%s' and policy '%s' was resolved to: %s", roles,
policy, result)
return result
def message(cls, task):
logger.debug("ProvisionTask.message(task=%s)" % task.uuid)
nodes = TaskHelper.nodes_to_provision(task.cluster)
USE_FAKE = settings.FAKE_TASKS or settings.FAKE_TASKS_AMQP
# We need to assign admin ips
# and only after that prepare syslog
# directories
task.cluster.prepare_for_provisioning()
for node in nodes:
if USE_FAKE:
continue
if node.offline:
raise errors.NodeOffline(
u'Node "%s" is offline.'
' Remove it from environment and try again.' %
node.full_name)
TaskHelper.prepare_syslog_dir(node)
node.status = 'provisioning'
db().commit()
serialized_cluster = task.cluster.replaced_provisioning_info or \
provisioning_serializers.serialize(task.cluster)
message = {
'method': 'provision',
'respond_to': 'provision_resp',
'args': {
'task_uuid': task.uuid,
'provisioning_info': serialized_cluster
}
}
return message
def update_by_agent(cls, instance, data):
"""Update Node instance with some specific cases for agent.
* don't update provisioning or error state back to discover
* don't update volume information if disks arrays is empty
:param data: dictionary of key-value pairs as object fields
:returns: Node instance
"""
# don't update provisioning and error back to discover
if instance.status in ('provisioning', 'error'):
if data.get('status', 'discover') == 'discover':
logger.debug(u"Node {0} has provisioning or error status - "
u"status not updated by agent".format(
instance.human_readable_name))
data['status'] = instance.status
# don't update volume information, if agent has sent an empty array
meta = data.get('meta', {})
if meta and len(meta.get('disks', [])) == 0 \
and instance.meta.get('disks'):
logger.warning(u'Node {0} has received an empty disks array - '
u'volume information will not be updated'.format(
instance.human_readable_name))
meta['disks'] = instance.meta['disks']
#(dshulyak) change this verification to NODE_STATUSES.deploying
# after we will reuse ips from dhcp range
netmanager = Cluster.get_network_manager()
admin_ng = netmanager.get_admin_network_group(instance.id)
if data.get('ip') and not netmanager.is_same_network(
data['ip'], admin_ng.cidr):
logger.debug('Corrupted network data %s, skipping update',
instance.id)
return instance
return cls.update(instance, data)
def checkout(self, instance):
fetch_file = os.path.join(const.REPOS_DIR, instance.repo_name,
'.git/FETCH_HEAD')
if os.path.exists(fetch_file):
current_ts = time.time()
cluster = Cluster.get_by_uid(instance.env_id)
last_fetch = os.stat(fetch_file).st_mtime
if cluster.status != CLUSTER_STATUSES.deployment and \
current_ts - last_fetch < const.REPO_TTL:
return
logger.debug("Repo TTL exceeded. Fetching code...")
ssh_cmd = self._get_ssh_cmd(instance.repo_name)
if not os.path.exists(self._get_key_path(instance.repo_name)):
logger.debug('Key file does not exist. Creating...')
self._create_key_file(instance.repo_name)
with instance.repo.git.custom_environment(GIT_SSH=ssh_cmd):
commit = instance.repo.remotes.origin.fetch(refspec=instance.ref)
commit = commit[0].commit
instance.repo.head.reference = commit
instance.repo.head.reset(index=True, working_tree=True)
