path: root/heat/engine/resources/instance.py

#
#    Licensed under the Apache License, Version 2.0 (the "License"); you may
#    not use this file except in compliance with the License. You may obtain
#    a copy of the License at
#
#         http://www.apache.org/licenses/LICENSE-2.0
#
#    Unless required by applicable law or agreed to in writing, software
#    distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
#    WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
#    License for the specific language governing permissions and limitations
#    under the License.

from heat.engine import signal_responder
from heat.engine import clients
from heat.engine import constraints
from heat.engine import properties
from heat.engine import resource
from heat.engine import scheduler
from heat.engine.resources.neutron import neutron
from heat.engine.resources import nova_utils
from heat.engine.resources import volume

from heat.common import exception
from heat.engine.resources.network_interface import NetworkInterface

from heat.openstack.common.gettextutils import _
from heat.openstack.common import log as logging

logger = logging.getLogger(__name__)


class Restarter(signal_responder.SignalResponder):
    PROPERTIES = (
        INSTANCE_ID,
    ) = (
        'InstanceId',
    )

    properties_schema = {
        INSTANCE_ID: properties.Schema(
            properties.Schema.STRING,
            _('Instance ID to be restarted.'),
            required=True
        ),
    }

    attributes_schema = {
        "AlarmUrl": _("A signed url to handle the alarm "
                      "(Heat extension).")
    }

    def _find_resource(self, resource_id):
        '''
        Return the resource with the specified instance ID, or None if it
        cannot be found.
        '''
        for resource in self.stack.itervalues():
            if resource.resource_id == resource_id:
                return resource
        return None

    def handle_create(self):
        super(Restarter, self).handle_create()
        self.resource_id_set(self._get_user_id())

    def handle_signal(self, details=None):
        if details is None:
            alarm_state = 'alarm'
        else:
            alarm_state = details.get('state', 'alarm').lower()

        logger.info(_('%(name)s Alarm, new state %(state)s') % {
                    'name': self.name, 'state': alarm_state})

        if alarm_state != 'alarm':
            return

        victim = self._find_resource(self.properties[self.INSTANCE_ID])
        if victim is None:
            logger.info(_('%(name)s Alarm, can not find instance '
                        '%(instance)s') % {
                            'name': self.name,
                            'instance': self.properties[self.INSTANCE_ID]})
            return

        logger.info(_('%(name)s Alarm, restarting resource: %(victim)s') % {
                    'name': self.name, 'victim': victim.name})
        self.stack.restart_resource(victim.name)

    def _resolve_attribute(self, name):
        '''
        heat extension: "AlarmUrl" returns the url to post to the policy
        when there is an alarm.
        '''
        if name == 'AlarmUrl' and self.resource_id is not None:
            return unicode(self._get_signed_url())


class Instance(resource.Resource):

    PROPERTIES = (
        IMAGE_ID, INSTANCE_TYPE, KEY_NAME, AVAILABILITY_ZONE,
        DISABLE_API_TERMINATION, KERNEL_ID, MONITORING,
        PLACEMENT_GROUP_NAME, PRIVATE_IP_ADDRESS, RAM_DISK_ID,
        SECURITY_GROUPS, SECURITY_GROUP_IDS, NETWORK_INTERFACES,
        SOURCE_DEST_CHECK, SUBNET_ID, TAGS, NOVA_SCHEDULER_HINTS, TENANCY,
        USER_DATA, VOLUMES,
    ) = (
        'ImageId', 'InstanceType', 'KeyName', 'AvailabilityZone',
        'DisableApiTermination', 'KernelId', 'Monitoring',
        'PlacementGroupName', 'PrivateIpAddress', 'RamDiskId',
        'SecurityGroups', 'SecurityGroupIds', 'NetworkInterfaces',
        'SourceDestCheck', 'SubnetId', 'Tags', 'NovaSchedulerHints', 'Tenancy',
        'UserData', 'Volumes',
    )

    _TAG_KEYS = (
        TAG_KEY, TAG_VALUE,
    ) = (
        'Key', 'Value',
    )

    _NOVA_SCHEDULER_HINT_KEYS = (
        NOVA_SCHEDULER_HINT_KEY, NOVA_SCHEDULER_HINT_VALUE,
    ) = (
        'Key', 'Value',
    )

    _VOLUME_KEYS = (
        VOLUME_DEVICE, VOLUME_ID,
    ) = (
        'Device', 'VolumeId',
    )

    properties_schema = {
        IMAGE_ID: properties.Schema(
            properties.Schema.STRING,
            _('Glance image ID or name.'),
            constraints=[
                constraints.CustomConstraint('glance.image')
            ],
            required=True
        ),
        # AWS does not require InstanceType but Heat does because the nova
        # create api call requires a flavor
        INSTANCE_TYPE: properties.Schema(
            properties.Schema.STRING,
            _('Nova instance type (flavor).'),
            required=True,
            update_allowed=True
        ),
        KEY_NAME: properties.Schema(
            properties.Schema.STRING,
            _('Optional Nova keypair name.'),
            constraints=[
                constraints.CustomConstraint("nova.keypair")
            ]
        ),
        AVAILABILITY_ZONE: properties.Schema(
            properties.Schema.STRING,
            _('Availability zone to launch the instance in.')
        ),
        DISABLE_API_TERMINATION: properties.Schema(
            properties.Schema.STRING,
            _('Not Implemented.'),
            implemented=False
        ),
        KERNEL_ID: properties.Schema(
            properties.Schema.STRING,
            _('Not Implemented.'),
            implemented=False
        ),
        MONITORING: properties.Schema(
            properties.Schema.BOOLEAN,
            _('Not Implemented.'),
            implemented=False
        ),
        PLACEMENT_GROUP_NAME: properties.Schema(
            properties.Schema.STRING,
            _('Not Implemented.'),
            implemented=False
        ),
        PRIVATE_IP_ADDRESS: properties.Schema(
            properties.Schema.STRING,
            _('Not Implemented.'),
            implemented=False
        ),
        RAM_DISK_ID: properties.Schema(
            properties.Schema.STRING,
            _('Not Implemented.'),
            implemented=False
        ),
        SECURITY_GROUPS: properties.Schema(
            properties.Schema.LIST,
            _('Security group names to assign.')
        ),
        SECURITY_GROUP_IDS: properties.Schema(
            properties.Schema.LIST,
            _('Security group IDs to assign.')
        ),
        NETWORK_INTERFACES: properties.Schema(
            properties.Schema.LIST,
            _('Network interfaces to associate with instance.')
        ),
        SOURCE_DEST_CHECK: properties.Schema(
            properties.Schema.BOOLEAN,
            _('Not Implemented.'),
            implemented=False
        ),
        SUBNET_ID: properties.Schema(
            properties.Schema.STRING,
            _('Subnet ID to launch instance in.')
        ),
        TAGS: properties.Schema(
            properties.Schema.LIST,
            _('Tags to attach to instance.'),
            schema=properties.Schema(
                properties.Schema.MAP,
                schema={
                    TAG_KEY: properties.Schema(
                        properties.Schema.STRING,
                        required=True
                    ),
                    TAG_VALUE: properties.Schema(
                        properties.Schema.STRING,
                        required=True
                    ),
                },
            ),
            update_allowed=True
        ),
        NOVA_SCHEDULER_HINTS: properties.Schema(
            properties.Schema.LIST,
            _('Scheduler hints to pass to Nova (Heat extension).'),
            schema=properties.Schema(
                properties.Schema.MAP,
                schema={
                    NOVA_SCHEDULER_HINT_KEY: properties.Schema(
                        properties.Schema.STRING,
                        required=True
                    ),
                    NOVA_SCHEDULER_HINT_VALUE: properties.Schema(
                        properties.Schema.STRING,
                        required=True
                    ),
                },
            )
        ),
        TENANCY: properties.Schema(
            properties.Schema.STRING,
            _('Not Implemented.'),
            constraints=[
                constraints.AllowedValues(['dedicated', 'default']),
            ],
            implemented=False
        ),
        USER_DATA: properties.Schema(
            properties.Schema.STRING,
            _('User data to pass to instance.')
        ),
        VOLUMES: properties.Schema(
            properties.Schema.LIST,
            _('Volumes to attach to instance.'),
            default=[],
            schema=properties.Schema(
                properties.Schema.MAP,
                schema={
                    VOLUME_DEVICE: properties.Schema(
                        properties.Schema.STRING,
                        _('The device where the volume is exposed on the '
                          'instance. This assignment may not be honored and '
                          'it is advised that the path '
                          '/dev/disk/by-id/virtio-<VolumeId> be used '
                          'instead.'),
                        required=True
                    ),
                    VOLUME_ID: properties.Schema(
                        properties.Schema.STRING,
                        _('The ID of the volume to be attached.'),
                        required=True
                    ),
                }
            )
        ),
    }

    attributes_schema = {'AvailabilityZone': _('The Availability Zone where '
                                               'the specified instance is '
                                               'launched.'),
                         'PrivateDnsName': _('Private DNS name of the'
                                             ' specified instance.'),
                         'PublicDnsName': _('Public DNS name of the specified '
                                            'instance.'),
                         'PrivateIp': _('Private IP address of the specified '
                                        'instance.'),
                         'PublicIp': _('Public IP address of the specified '
                                       'instance.')}

    update_allowed_keys = ('Metadata', 'Properties')

    # Server host name limit to 53 characters by due to typical default
    # linux HOST_NAME_MAX of 64, minus the .novalocal appended to the name
    physical_resource_name_limit = 53

    def __init__(self, name, json_snippet, stack):
        super(Instance, self).__init__(name, json_snippet, stack)
        self.ipaddress = None

    def _set_ipaddress(self, networks):
        '''
        Read the server's IP address from a list of networks provided by Nova
        '''
        # Just record the first ipaddress
        for n in networks:
            if len(networks[n]) > 0:
                self.ipaddress = networks[n][0]
                break

    def _ipaddress(self):
        '''
        Return the server's IP address, fetching it from Nova if necessary
        '''
        if self.ipaddress is None:
            self.ipaddress = nova_utils.server_to_ipaddress(
                self.nova(), self.resource_id)

        return self.ipaddress or '0.0.0.0'

    def _resolve_attribute(self, name):
        res = None
        if name == 'AvailabilityZone':
            res = self.properties[self.AVAILABILITY_ZONE]
        elif name in ['PublicIp', 'PrivateIp', 'PublicDnsName',
                      'PrivateDnsName']:
            res = self._ipaddress()

        logger.info(_('%(name)s._resolve_attribute(%(attname)s) == %(res)s'),
                    {'name': self.name,
                     'attname': name,
                     'res': res})
        return unicode(res) if res else None

    def _build_nics(self, network_interfaces,
                    security_groups=None, subnet_id=None):

        nics = None

        if network_interfaces:
            unsorted_nics = []
            for entry in network_interfaces:
                nic = (entry
                       if not isinstance(entry, basestring)
                       else {'NetworkInterfaceId': entry,
                             'DeviceIndex': len(unsorted_nics)})
                unsorted_nics.append(nic)
            sorted_nics = sorted(unsorted_nics,
                                 key=lambda nic: int(nic['DeviceIndex']))
            nics = [{'port-id': nic['NetworkInterfaceId']}
                    for nic in sorted_nics]
        else:
            # if SubnetId property in Instance, ensure subnet exists
            if subnet_id:
                neutronclient = self.neutron()
                network_id = NetworkInterface.network_id_from_subnet_id(
                    neutronclient, subnet_id)
                # if subnet verified, create a port to use this subnet
                # if port is not created explicitly, nova will choose
                # the first subnet in the given network.
                if network_id:
                    fixed_ip = {'subnet_id': subnet_id}
                    props = {
                        'admin_state_up': True,
                        'network_id': network_id,
                        'fixed_ips': [fixed_ip]
                    }

                    if security_groups:
                        props['security_groups'] = \
                            neutron.NeutronResource.get_secgroup_uuids(
                                security_groups, self.neutron())

                    port = neutronclient.create_port({'port': props})['port']
                    nics = [{'port-id': port['id']}]

        return nics

    def _get_security_groups(self):
        security_groups = []
        for key in (self.SECURITY_GROUPS, self.SECURITY_GROUP_IDS):
            if self.properties.get(key) is not None:
                for sg in self.properties.get(key):
                    security_groups.append(sg)
        if not security_groups:
            security_groups = None
        return security_groups

    def _get_nova_metadata(self, properties):
        if properties is None or properties.get(self.TAGS) is None:
            return None

        return dict((tm[self.TAG_KEY], tm[self.TAG_VALUE])
                    for tm in properties[self.TAGS])

    def handle_create(self):
        security_groups = self._get_security_groups()

        userdata = self.properties[self.USER_DATA] or ''
        flavor = self.properties[self.INSTANCE_TYPE]
        availability_zone = self.properties[self.AVAILABILITY_ZONE]

        image_name = self.properties[self.IMAGE_ID]

        image_id = nova_utils.get_image_id(self.nova(), image_name)

        flavor_id = nova_utils.get_flavor_id(self.nova(), flavor)

        scheduler_hints = {}
        if self.properties[self.NOVA_SCHEDULER_HINTS]:
            for tm in self.properties[self.NOVA_SCHEDULER_HINTS]:
                scheduler_hints[tm[self.TAG_KEY]] = tm[self.TAG_VALUE]
        else:
            scheduler_hints = None

        nics = self._build_nics(self.properties[self.NETWORK_INTERFACES],
                                security_groups=security_groups,
                                subnet_id=self.properties[self.SUBNET_ID])
        server = None

        try:
            server = self.nova().servers.create(
                name=self.physical_resource_name(),
                image=image_id,
                flavor=flavor_id,
                key_name=self.properties[self.KEY_NAME],
                security_groups=security_groups,
                userdata=nova_utils.build_userdata(self, userdata),
                meta=self._get_nova_metadata(self.properties),
                scheduler_hints=scheduler_hints,
                nics=nics,
                availability_zone=availability_zone)
        finally:
            # Avoid a race condition where the thread could be cancelled
            # before the ID is stored
            if server is not None:
                self.resource_id_set(server.id)

        return server, scheduler.TaskRunner(self._attach_volumes_task())

    def _attach_volumes_task(self):
        attach_tasks = (volume.VolumeAttachTask(self.stack,
                                                self.resource_id,
                                                volume_id,
                                                device)
                        for volume_id, device in self.volumes())
        return scheduler.PollingTaskGroup(attach_tasks)

    def check_create_complete(self, cookie):
        server, volume_attach_task = cookie
        return (self._check_active(server) and
                self._check_volume_attached(server, volume_attach_task))

    def _check_volume_attached(self, server, volume_attach_task):
        if not volume_attach_task.started():
            self._set_ipaddress(server.networks)
            volume_attach_task.start()
            return volume_attach_task.done()
        else:
            return volume_attach_task.step()

    def _check_active(self, server):
        if server.status != 'ACTIVE':
            server.get()

        if server.status == 'ACTIVE':
            return True

        # Some clouds append extra (STATUS) strings to the status
        short_server_status = server.status.split('(')[0]
        if short_server_status in nova_utils.deferred_server_statuses:
            return False

        if server.status == 'ERROR':
            fault = getattr(server, 'fault', {})
            message = fault.get('message', 'Unknown')
            code = fault.get('code', 500)
            exc = exception.Error(_("Creation of server %(server)s "
                                    "failed: %(message)s (%(code)s)") %
                                  dict(server=server.name,
                                       message=message,
                                       code=code))
            raise exc

        exc = exception.Error(_("Creation of server %(server)s failed "
                                "with unknown status: %(status)s") %
                              dict(server=server.name,
                                   status=server.status))
        raise exc

    def volumes(self):
        """
        Return an iterator over (volume_id, device) tuples for all volumes
        that should be attached to this instance.
        """
        volumes = self.properties[self.VOLUMES]

        return ((vol[self.VOLUME_ID],
                 vol[self.VOLUME_DEVICE]) for vol in volumes)

    def handle_update(self, json_snippet, tmpl_diff, prop_diff):
        if 'Metadata' in tmpl_diff:
            self.metadata = tmpl_diff['Metadata']

        server = None
        if self.TAGS in prop_diff:
            server = self.nova().servers.get(self.resource_id)
            nova_utils.meta_update(self.nova(),
                                   server,
                                   self._get_nova_metadata(prop_diff))

        if self.INSTANCE_TYPE in prop_diff:
            flavor = prop_diff[self.INSTANCE_TYPE]
            flavor_id = nova_utils.get_flavor_id(self.nova(), flavor)
            if not server:
                server = self.nova().servers.get(self.resource_id)
            checker = scheduler.TaskRunner(nova_utils.resize, server, flavor,
                                           flavor_id)
            checker.start()
            return checker

    def check_update_complete(self, checker):
        return checker.step() if checker is not None else True

    def metadata_update(self, new_metadata=None):
        '''
        Refresh the metadata if new_metadata is None
        '''
        if new_metadata is None:
            self.metadata = self.parsed_template('Metadata')

    def validate(self):
        '''
        Validate any of the provided params
        '''
        res = super(Instance, self).validate()
        if res:
            return res

        # check validity of security groups vs. network interfaces
        security_groups = self._get_security_groups()
        if security_groups and self.properties.get(self.NETWORK_INTERFACES):
            raise exception.ResourcePropertyConflict(
                '/'.join([self.SECURITY_GROUPS, self.SECURITY_GROUP_IDS]),
                self.NETWORK_INTERFACES)

    @scheduler.wrappertask
    def _delete_server(self, server):
        '''
        Return a co-routine that deletes the server and waits for it to
        disappear from Nova.
        '''
        yield self._detach_volumes_task()()
        server.delete()

        while True:
            yield

            try:
                server.get()
                if server.status == "DELETED":
                    self.resource_id_set(None)
                    break
                elif server.status == "ERROR":
                    raise exception.Error(_("Deletion of server %s failed.") %
                                          server.id)
            except clients.novaclient.exceptions.NotFound:
                self.resource_id_set(None)
                break

    def _detach_volumes_task(self):
        '''
        Detach volumes from the instance
        '''
        detach_tasks = (volume.VolumeDetachTask(self.stack,
                                                self.resource_id,
                                                volume_id)
                        for volume_id, device in self.volumes())
        return scheduler.PollingTaskGroup(detach_tasks)

    def handle_delete(self):
        '''
        Delete an instance, blocking until it is disposed by OpenStack
        '''
        if self.resource_id is None:
            return

        try:
            server = self.nova().servers.get(self.resource_id)
        except clients.novaclient.exceptions.NotFound:
            self.resource_id_set(None)
            return

        server_delete_task = scheduler.TaskRunner(self._delete_server,
                                                  server=server)
        server_delete_task.start()
        return server_delete_task

    def check_delete_complete(self, server_delete_task):
        # if the resource was already deleted, server_delete_task will be None
        if server_delete_task is None:
            return True
        else:
            return server_delete_task.step()

    def handle_suspend(self):
        '''
        Suspend an instance - note we do not wait for the SUSPENDED state,
        this is polled for by check_suspend_complete in a similar way to the
        create logic so we can take advantage of coroutines
        '''
        if self.resource_id is None:
            raise exception.Error(_('Cannot suspend %s, resource_id not set') %
                                  self.name)

        try:
            server = self.nova().servers.get(self.resource_id)
        except clients.novaclient.exceptions.NotFound:
            raise exception.NotFound(_('Failed to find instance %s') %
                                     self.resource_id)
        else:
            logger.debug(_("suspending instance %s") % self.resource_id)
            # We want the server.suspend to happen after the volume
            # detachement has finished, so pass both tasks and the server
            suspend_runner = scheduler.TaskRunner(server.suspend)
            volumes_runner = scheduler.TaskRunner(self._detach_volumes_task())
            return server, suspend_runner, volumes_runner

    def check_suspend_complete(self, cookie):
        server, suspend_runner, volumes_runner = cookie

        if not volumes_runner.started():
            volumes_runner.start()

        if volumes_runner.done():
            if not suspend_runner.started():
                suspend_runner.start()

            if suspend_runner.done():
                if server.status == 'SUSPENDED':
                    return True

                server.get()
                logger.debug(_("%(name)s check_suspend_complete "
                               "status = %(status)s"),
                             {'name': self.name,
                              'status': server.status})
                if server.status in list(nova_utils.deferred_server_statuses +
                                         ['ACTIVE']):
                    return server.status == 'SUSPENDED'
                else:
                    raise exception.Error(_(' nova reported unexpected '
                                            'instance[%(instance)s] '
                                            'status[%(status)s]') %
                                          {'instance': self.name,
                                           'status': server.status})
            else:
                suspend_runner.step()
        else:
            return volumes_runner.step()

    def handle_resume(self):
        '''
        Resume an instance - note we do not wait for the ACTIVE state,
        this is polled for by check_resume_complete in a similar way to the
        create logic so we can take advantage of coroutines
        '''
        if self.resource_id is None:
            raise exception.Error(_('Cannot resume %s, resource_id not set') %
                                  self.name)

        try:
            server = self.nova().servers.get(self.resource_id)
        except clients.novaclient.exceptions.NotFound:
            raise exception.NotFound(_('Failed to find instance %s') %
                                     self.resource_id)
        else:
            logger.debug(_("resuming instance %s") % self.resource_id)
            server.resume()
            return server, scheduler.TaskRunner(self._attach_volumes_task())

    def check_resume_complete(self, cookie):
        server, volume_attach_task = cookie
        return (self._check_active(server) and
                self._check_volume_attached(server, volume_attach_task))


def resource_mapping():
    return {
        'AWS::EC2::Instance': Instance,
        'OS::Heat::HARestarter': Restarter,
    }