nvp_driver.py

# Copyright 2013 Rackspace Hosting Inc.
# All Rights Reserved.
#
#    Licensed under the Apache License, Version 2.0 (the "License"); you may
#    not use this file except in compliance with the License. You may obtain
#    a copy of the License at
#
#         http://www.apache.org/licenses/LICENSE-2.0
#
#    Unless required by applicable law or agreed to in writing, software
#    distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
#    WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
#    License for the specific language governing permissions and limitations
#    under the License.

"""
NVP client driver for Quark
"""

import contextlib
import random

import aiclib
from neutron.extensions import securitygroup as sg_ext
from oslo_config import cfg
from oslo_log import log as logging

from quark.drivers import base
from quark.drivers import security_groups as sg_driver
from quark.environment import Capabilities
from quark import exceptions as q_exc
from quark import segment_allocations
from quark import utils

LOG = logging.getLogger(__name__)

CONF = cfg.CONF

nvp_opts = [
    cfg.IntOpt('max_ports_per_switch',
               default=0,
               help=_('Maximum amount of NVP ports on an NVP lswitch')),
    cfg.StrOpt('default_tz_type',
               help=_('The type of connector to use for the default tz'),
               default="stt"),
    cfg.ListOpt('additional_default_tz_types',
                default=[],
                help=_('List of additional default tz types to bind to the '
                       'default tz')),
    cfg.StrOpt('default_tz',
               help=_('The default transport zone UUID')),
    cfg.ListOpt('controller_connection',
                default=[],
                help=_('NVP Controller connection string')),
    cfg.IntOpt('max_rules_per_group',
               default=30,
               help=_('Maxiumum size of NVP SecurityRule list per group')),
    cfg.IntOpt('max_rules_per_port',
               default=30,
               help=_('Maximum rules per NVP lport across all groups')),
    cfg.IntOpt('backoff',
               default=0,
               help=_('Base seconds for exponential backoff')),
    cfg.BoolOpt("random_initial_controller",
                default=False,
                help=_("Whether or not to use a random controller or the "
                       "first controller when neutron starts up")),
    cfg.IntOpt("connection_switching_threshold",
               default=0,
               help=_("Number of times to use a connection before forcing "
                      "the next connection in the list to be used. A value "
                      "of 0 means only switch on Exceptions.")),
    cfg.BoolOpt("connection_switching_random",
                default=False,
                help=_("Determines whether connections are switched randomly "
                       "or using the default round-robin.")),
    cfg.IntOpt("operation_retries",
               default=3,
               help=_("Number of times to attempt to perform operations in "
                      "NVP.")),
]

physical_net_type_map = {
    "stt": "stt",
    "gre": "gre",
    "flat": "bridge",
    "bridge": "bridge",
    "vlan": "bridge",
    "local": "local",
}

CONF.register_opts(nvp_opts, "NVP")
SA_REGISTRY = segment_allocations.REGISTRY


class TransportZoneBinding(object):

    net_type = None

    def add(self, context, switch, tz_id, network_id):
        raise NotImplementedError()

    def remove(self, context, tz_id, network_id):
        raise NotImplementedError()


class VXLanTransportZoneBinding(TransportZoneBinding):

    net_type = 'vxlan'

    def add(self, context, switch, tz_id, network_id):
        driver = SA_REGISTRY.get_strategy(self.net_type)
        alloc = driver.allocate(context, tz_id, network_id)
        switch.transport_zone(
            tz_id, self.net_type, vxlan_id=alloc["id"])

    def remove(self, context, tz_id, network_id):
        driver = SA_REGISTRY.get_strategy(self.net_type)
        driver.deallocate(context, tz_id, network_id)


# A map of net_type (vlan, vxlan) to TransportZoneBinding impl.
TZ_BINDINGS = {
    VXLanTransportZoneBinding.net_type: VXLanTransportZoneBinding()
}


def _tag_roll(tags):
    return [{'scope': k, 'tag': v} for k, v in tags]


def _tag_unroll(tags):
    return dict((t['scope'], t['tag']) for t in tags)


class NVPDriver(base.BaseDriver):
    def __init__(self):
        self.nvp_connections = []
        self.conn_index = 0
        self.limits = {'max_ports_per_switch': 0,
                       'max_rules_per_group': 0,
                       'max_rules_per_port': 0}
        self.sg_driver = None
        if Capabilities.SECURITY_GROUPS in CONF.QUARK.environment_capabilities:
            self.sg_driver = sg_driver.SecurityGroupDriver()

        super(NVPDriver, self).__init__()

    @classmethod
    def get_name(klass):
        return "NVP"

    def load_config(self):
        # NOTE(mdietz): What does default_tz actually mean?
        #               We don't have one default.
        # NOTE(jkoelker): Transport Zone
        # NOTE(mdietz): :-/ tz isn't the issue. default is
        default_tz = CONF.NVP.default_tz
        LOG.info("Loading NVP settings " + str(default_tz))
        connections = CONF.NVP.controller_connection
        backoff = CONF.NVP.backoff

        self.limits.update({
            'max_ports_per_switch': CONF.NVP.max_ports_per_switch,
            'max_rules_per_group': CONF.NVP.max_rules_per_group,
            'max_rules_per_port': CONF.NVP.max_rules_per_port})
        LOG.info("Loading NVP settings " + str(connections))

        for conn in connections:
            (ip, port, user, pw, req_timeout,
             http_timeout, retries, redirects) = conn.split(":")

            self.nvp_connections.append(dict(ip_address=ip,
                                        port=port,
                                        username=user,
                                        password=pw,
                                        req_timeout=req_timeout,
                                        http_timeout=int(http_timeout),
                                        retries=int(retries),
                                        redirects=redirects,
                                        default_tz=default_tz,
                                        backoff=backoff,
                                        usages=0))

        if connections:
            if CONF.NVP.random_initial_controller:
                self.conn_index = self._new_random_index(-1, len(connections))

            LOG.info("NVP Driver config loaded. Starting with controller %s" %
                     self.nvp_connections[self.conn_index]["ip_address"])
        else:
            LOG.critical("No NVP connection configurations found!")

    def _new_random_index(self, current, index_range):
        new_index = current
        while new_index == current:
            new_index = random.randint(0, index_range - 1)
        return new_index

    def _connection(self):
        if len(self.nvp_connections) == 0:
            raise q_exc.NoBackendConnectionsDefined(
                msg="No NVP connections defined cannot continue")

        conn = self.nvp_connections[self.conn_index]

        if CONF.NVP.connection_switching_threshold > 0:
            # NOTE(mdietz): This is racy. See get_connection below.
            conn["usages"] += 1
            if conn["usages"] >= CONF.NVP.connection_switching_threshold:
                conn["usages"] = 0
                self._next_connection()
                conn = self.nvp_connections[self.conn_index]

        if "connection" not in conn:
            scheme = conn["port"] == "443" and "https" or "http"
            uri = "%s://%s:%s" % (scheme, conn["ip_address"], conn["port"])
            user = conn['username']
            passwd = conn['password']
            timeout = conn['http_timeout']
            retries = conn['retries']
            backoff = conn['backoff']
            conn["connection"] = aiclib.nvp.Connection(uri,
                                                       username=user,
                                                       password=passwd,
                                                       timeout=timeout,
                                                       retries=retries,
                                                       backoff=backoff)
        return conn["connection"]

    def _next_connection(self):
        # TODO(anyone): Do we want to drop and create new connections at some
        #               point? What about recycling them after a certain
        #               number of usages or time, proactively?
        LOG.info("Switching NVP connections...")
        conn_len = len(self.nvp_connections)
        if conn_len and conn_len > 1:
            if CONF.NVP.connection_switching_random:
                self.conn_index = self._new_random_index(self.conn_index,
                                                         conn_len)
                LOG.info("New connection chosen at random is %s" %
                         self.nvp_connections[self.conn_index]["ip_address"])
            else:
                self.conn_index = (self.conn_index + 1) % conn_len
                LOG.info("New connection chosen round-robin is %s" %
                         self.nvp_connections[self.conn_index]["ip_address"])
        else:
            LOG.info("No other connections to choose from")

    @contextlib.contextmanager
    def get_connection(self):
        try:
            yield self._connection()
        except Exception:
            # This is racy. A pile-up of failures could occur on one
            # controller, causing them to rapidly switch and fail back
            # to the original failing controller. However, we can't be in
            # the business of implementing a load balancer inside the code.
            # TODO(anyone): Investigate whether NVP is now giving us sequence
            #               IDs. If so, rapid round-robining becomes plausible,
            #               though we don't have an easy path to dropping bad
            #               controllers.
            self._next_connection()
            raise

    def create_network(self, context, network_name, tags=None,
                       network_id=None, **kwargs):
        return self._lswitch_create(context, network_name, tags,
                                    network_id, **kwargs)

    def delete_network(self, context, network_id):
        lswitches = self._lswitches_for_network(context, network_id).results()
        for switch in lswitches["results"]:
            try:
                self._lswitch_delete(context, switch["uuid"])
                # NOTE(morgabra) If we haven't thrown here, we can be sure the
                # resource was deleted from NSX. So we give a chance for any
                # previously allocated segment ids to deallocate.
                self._remove_default_tz_bindings(
                    context, network_id)
            except aiclib.core.AICException as ae:
                if ae.code == 404:
                    LOG.info("LSwitch/Network %s not found in NVP."
                             " Ignoring explicitly. Code: %s, Message: %s"
                             % (network_id, ae.code, ae.message))
                else:
                    LOG.info("AICException deleting LSwitch/Network %s in NVP."
                             " Ignoring explicitly. Code: %s, Message: %s"
                             % (network_id, ae.code, ae.message))
            except Exception as e:
                LOG.info("Failed to delete LSwitch/Network %s in NVP."
                         " Ignoring explicitly. Message: %s"
                         % (network_id, e.args[0]))

    def _collect_lswitch_info(self, lswitch, get_status):
        info = {
            'port_isolation_enabled': lswitch['port_isolation_enabled'],
            'display_name': lswitch['display_name'],
            'uuid': lswitch['uuid'],
            'transport_zones': lswitch['transport_zones'],
        }
        info.update(_tag_unroll(lswitch['tags']))
        if get_status:
            status = lswitch.pop('_relations')['LogicalSwitchStatus']
            info.update({
                'lport_stats': {
                    'fabric_up': status['lport_fabric_up_count'],
                    'admin_up': status['lport_admin_up_count'],
                    'link_up': status['lport_link_up_count'],
                    'count': status['lport_count'],
                }, 'fabric_status': status['fabric_status'],
            })
        return info

    def diag_network(self, context, network_id, get_status, **kwargs):
        switches = self._lswitch_status_query(context, network_id)['results']
        return {'logical_switches': [self._collect_lswitch_info(s, get_status)
                for s in switches]}

    def create_port(self, context, network_id, port_id, status=True,
                    security_groups=None, device_id="", **kwargs):
        security_groups = security_groups or []
        tenant_id = context.tenant_id
        lswitch = self._create_or_choose_lswitch(context, network_id)

        @utils.retry_loop(CONF.NVP.operation_retries)
        def _create_lswitch_port():
            with self.get_connection() as connection:
                port = connection.lswitch_port(lswitch)
                port.admin_status_enabled(status)
                if not self.sg_driver:
                    nvp_group_ids = self._get_security_groups_for_port(
                        context, security_groups)
                    port.security_profiles(nvp_group_ids)
                tags = [dict(tag=network_id, scope="neutron_net_id"),
                        dict(tag=port_id, scope="neutron_port_id"),
                        dict(tag=tenant_id, scope="os_tid"),
                        dict(tag=device_id, scope="vm_id")]
                LOG.debug("Creating port on switch %s" % lswitch)
                port.tags(tags)
                res = port.create()
                try:
                    """Catching odd NVP returns here will make it safe to
                    assume that NVP returned something correct."""
                    res["lswitch"] = lswitch
                except TypeError:
                    LOG.exception("Unexpected return from NVP: %s" % res)
                    raise
                port = connection.lswitch_port(lswitch)
                port.uuid = res["uuid"]
                port.attachment_vif(port_id)
                return res
        return _create_lswitch_port()

    @utils.retry_loop(CONF.NVP.operation_retries)
    def update_port(self, context, port_id, mac_address=None, device_id=None,
                    status=True, security_groups=None, **kwargs):
        if not self.sg_driver:
            security_groups = security_groups or []
        else:
            kwargs.update({'security_groups': security_groups})
        with self.get_connection() as connection:
            if self.sg_driver:
                kwargs.update({'mac_address': mac_address,
                               'device_id': device_id})
                self.sg_driver.update_port(**kwargs)
            lswitch_id = self._lswitch_from_port(context, port_id)
            port = connection.lswitch_port(lswitch_id, port_id)
            if not self.sg_driver:
                nvp_group_ids = self._get_security_groups_for_port(
                    context, security_groups)
                if nvp_group_ids:
                    port.security_profiles(nvp_group_ids)
            port.admin_status_enabled(status)
            return port.update()

    @utils.retry_loop(CONF.NVP.operation_retries)
    def delete_port(self, context, port_id, **kwargs):
        with self.get_connection() as connection:
            lswitch_uuid = kwargs.get('lswitch_uuid', None)
            try:
                if not lswitch_uuid:
                    lswitch_uuid = self._lswitch_from_port(context, port_id)
                LOG.debug("Deleting port %s from lswitch %s"
                          % (port_id, lswitch_uuid))
                connection.lswitch_port(lswitch_uuid, port_id).delete()
                if self.sg_driver:
                    self.sg_driver.delete_port(**kwargs)
            except aiclib.core.AICException as ae:
                if ae.code == 404:
                    LOG.info("LSwitchPort/Port %s not found in NVP."
                             " Ignoring explicitly. Code: %s, Message: %s"
                             % (port_id, ae.code, ae.message))
                else:
                    LOG.info("AICException deleting LSwitchPort/Port %s in "
                             "NVP. Ignoring explicitly. Code: %s, Message: %s"
                             % (port_id, ae.code, ae.message))

            except Exception as e:
                LOG.info("Failed to delete LSwitchPort/Port %s in NVP."
                         " Ignoring explicitly. Message: %s"
                         % (port_id, e.args[0]))

    def _collect_lport_info(self, lport, get_status):
        info = {
            'mirror_targets': lport['mirror_targets'],
            'display_name': lport['display_name'],
            'portno': lport['portno'],
            'allowed_address_pairs': lport['allowed_address_pairs'],
            'nvp_security_groups': lport['security_profiles'],
            'uuid': lport['uuid'],
            'admin_status_enabled': lport['admin_status_enabled'],
            'queue_uuid': lport['queue_uuid'],
        }
        if get_status:
            stats = lport['statistics']
            status = lport['status']
            lswitch = {
                'uuid': status['lswitch']['uuid'],
                'display_name': status['lswitch']['display_name'],
            }
            lswitch.update(_tag_unroll(status['lswitch']['tags']))
            info.update({
                'statistics': {
                    'recieved': {
                        'packets': stats['rx_packets'],
                        'bytes': stats['rx_bytes'],
                        'errors': stats['rx_errors']},
                    'transmitted': {
                        'packets': stats['tx_packets'],
                        'bytes': stats['tx_bytes'],
                        'errors': stats['tx_errors']
                    },
                },
                'status': {
                    'link_status_up': status['link_status_up'],
                    'admin_status_up': status['admin_status_up'],
                    'fabric_status_up': status['fabric_status_up'],
                },
                'lswitch': lswitch,
            })
        info.update(_tag_unroll(lport['tags']))
        return info

    def diag_port(self, context, port_id, get_status=False, **kwargs):
        with self.get_connection() as connection:
            lswitch_uuid = self._lswitch_from_port(context, port_id)
            lswitch_port = connection.lswitch_port(lswitch_uuid, port_id)

            query = lswitch_port.query()
            query.relations("LogicalPortAttachment")
            results = query.results()
            if results['result_count'] == 0:
                return {'lport': "Logical port not found."}

            config = results['results'][0]
            relations = config.pop('_relations')
            config['attachment'] = relations['LogicalPortAttachment']['type']
            if get_status:
                config['status'] = lswitch_port.status()
                config['statistics'] = lswitch_port.statistics()
            return {'lport': self._collect_lport_info(config, get_status)}

    def _get_network_details(self, context, network_id, switches):
        name, phys_net, phys_type, segment_id = None, None, None, None
        for res in switches["results"]:
            name = res["display_name"]
            for zone in res["transport_zones"]:
                phys_net = zone["zone_uuid"]
                phys_type = zone["transport_type"]
                if "binding_config" in zone:
                    binding = zone["binding_config"]
                    segment_id = binding["vlan_translation"][0]["transport"]
                break
            return dict(network_name=name, phys_net=phys_net,
                        phys_type=phys_type, segment_id=segment_id)
        return {}

    def create_security_group(self, context, group_name, **group):
        tenant_id = context.tenant_id
        with self.get_connection() as connection:
            group_id = group.get('group_id')
            profile = connection.securityprofile()
            if group_name:
                profile.display_name(group_name)
            ingress_rules = group.get('port_ingress_rules', [])
            egress_rules = group.get('port_egress_rules', [])

            if (len(ingress_rules) + len(egress_rules) >
                    self.limits['max_rules_per_group']):
                raise q_exc.DriverLimitReached(limit="rules per group")

            if egress_rules:
                profile.port_egress_rules(egress_rules)
            if ingress_rules:
                profile.port_ingress_rules(ingress_rules)
            tags = [dict(tag=group_id, scope="neutron_group_id"),
                    dict(tag=tenant_id, scope="os_tid")]
            LOG.debug("Creating security profile %s" % group_name)
            profile.tags(tags)
            return profile.create()

    def delete_security_group(self, context, group_id, **kwargs):
        guuid = self._get_security_group_id(context, group_id)
        with self.get_connection() as connection:
            LOG.debug("Deleting security profile %s" % group_id)
            connection.securityprofile(guuid).delete()

    def update_security_group(self, context, group_id, **group):
        query = self._get_security_group(context, group_id)
        with self.get_connection() as connection:
            profile = connection.securityprofile(query.get('uuid'))

            ingress_rules = group.get('port_ingress_rules',
                                      query.get('logical_port_ingress_rules'))
            egress_rules = group.get('port_egress_rules',
                                     query.get('logical_port_egress_rules'))

            if (len(ingress_rules) + len(egress_rules) >
                    self.limits['max_rules_per_group']):
                raise q_exc.DriverLimitReached(limit="rules per group")

            if group.get('name', None):
                profile.display_name(group['name'])
            if group.get('port_ingress_rules', None) is not None:
                profile.port_ingress_rules(ingress_rules)
            if group.get('port_egress_rules', None) is not None:
                profile.port_egress_rules(egress_rules)
            return profile.update()

    def _update_security_group_rules(self, context, group_id, rule, operation,
                                     checks):
        groupd = self._get_security_group(context, group_id)
        direction, secrule = self._get_security_group_rule_object(context,
                                                                  rule)
        rulelist = groupd['logical_port_%s_rules' % direction]
        for check in checks:
            if not check(secrule, rulelist):
                raise checks[check]
        getattr(rulelist, operation)(secrule)

        LOG.debug("%s rule on security group %s" % (operation, groupd['uuid']))
        group = {'port_%s_rules' % direction: rulelist}
        return self.update_security_group(context, group_id, **group)

    def create_security_group_rule(self, context, group_id, rule):
        return self._update_security_group_rules(
            context, group_id, rule, 'append',
            {(lambda x, y: x not in y):
             sg_ext.SecurityGroupRuleExists(id=group_id),
             (lambda x, y:
                 self._check_rule_count_per_port(context, group_id) <
                 self.limits['max_rules_per_port']):
             q_exc.DriverLimitReached(limit="rules per port")})

    def delete_security_group_rule(self, context, group_id, rule):
        return self._update_security_group_rules(
            context, group_id, rule, 'remove',
            {(lambda x, y: x in y):
             sg_ext.SecurityGroupRuleNotFound(id="with group_id %s" %
                                              group_id)})

    def _create_or_choose_lswitch(self, context, network_id):
        switches = self._lswitch_status_query(context, network_id)
        switch = self._lswitch_select_open(context, network_id=network_id,
                                           switches=switches)
        if switch:
            LOG.debug("Found open switch %s" % switch)
            return switch

        switch_details = self._get_network_details(context, network_id,
                                                   switches)
        if not switch_details:
            raise q_exc.BadNVPState(net_id=network_id)

        return self._lswitch_create(context, network_id=network_id,
                                    **switch_details)

    def _lswitch_status_query(self, context, network_id):
        query = self._lswitches_for_network(context, network_id)
        query.relations("LogicalSwitchStatus")
        results = query.results()
        LOG.debug("Query results: %s" % results)
        return results

    def _lswitch_select_open(self, context, switches=None, **kwargs):
        """Selects an open lswitch for a network.

        Note that it does not select the most full switch, but merely one with
        ports available.
        """

        if switches is not None:
            for res in switches["results"]:
                count = res["_relations"]["LogicalSwitchStatus"]["lport_count"]
                if (self.limits['max_ports_per_switch'] == 0 or
                        count < self.limits['max_ports_per_switch']):
                    return res["uuid"]
        return None

    @utils.retry_loop(CONF.NVP.operation_retries)
    def _lswitch_delete(self, context, lswitch_uuid):
        with self.get_connection() as connection:
            LOG.debug("Deleting lswitch %s" % lswitch_uuid)
            connection.lswitch(lswitch_uuid).delete()

    def _config_provider_attrs(self, connection, switch, phys_net,
                               net_type, segment_id):
        if not (phys_net or net_type):
            return
        if not phys_net and net_type:
            raise q_exc.ProvidernetParamError(
                msg="provider:physical_network parameter required")
        if phys_net and not net_type:
            raise q_exc.ProvidernetParamError(
                msg="provider:network_type parameter required")
        if net_type not in ("bridge", "vlan", "vxlan") and segment_id:
            raise q_exc.SegmentIdUnsupported(net_type=net_type)
        if net_type == "vlan" and not segment_id:
            raise q_exc.SegmentIdRequired(net_type=net_type)

        phys_type = physical_net_type_map.get(net_type.lower())
        if not phys_type:
            raise q_exc.InvalidPhysicalNetworkType(net_type=net_type)

        tz_query = connection.transportzone(phys_net).query()
        transport_zone = tz_query.results()

        if transport_zone["result_count"] == 0:
            raise q_exc.PhysicalNetworkNotFound(phys_net=phys_net)
        switch.transport_zone(zone_uuid=phys_net,
                              transport_type=phys_type,
                              vlan_id=segment_id)

    def _add_default_tz_bindings(self, context, switch, network_id):
        """Configure any additional default transport zone bindings."""
        default_tz = CONF.NVP.default_tz

        # If there is no default tz specified it's pointless to try
        # and add any additional default tz bindings.
        if not default_tz:
            LOG.warn("additional_default_tz_types specified, "
                     "but no default_tz. Skipping "
                     "_add_default_tz_bindings().")
            return

        # This should never be called without a neutron network uuid,
        # we require it to bind some segment allocations.
        if not network_id:
            LOG.warn("neutron network_id not specified, skipping "
                     "_add_default_tz_bindings()")
            return

        for net_type in CONF.NVP.additional_default_tz_types:
            if net_type in TZ_BINDINGS:
                binding = TZ_BINDINGS[net_type]
                binding.add(context, switch, default_tz, network_id)
            else:
                LOG.warn("Unknown default tz type %s" % (net_type))

    def _remove_default_tz_bindings(self, context, network_id):
        """Deconfigure any additional default transport zone bindings."""
        default_tz = CONF.NVP.default_tz

        if not default_tz:
            LOG.warn("additional_default_tz_types specified, "
                     "but no default_tz. Skipping "
                     "_remove_default_tz_bindings().")
            return

        if not network_id:
            LOG.warn("neutron network_id not specified, skipping "
                     "_remove_default_tz_bindings()")
            return

        for net_type in CONF.NVP.additional_default_tz_types:
            if net_type in TZ_BINDINGS:
                binding = TZ_BINDINGS[net_type]
                binding.remove(context, default_tz, network_id)
            else:
                LOG.warn("Unknown default tz type %s" % (net_type))

    @utils.retry_loop(CONF.NVP.operation_retries)
    def _lswitch_create(self, context, network_name=None, tags=None,
                        network_id=None, phys_net=None,
                        phys_type=None, segment_id=None,
                        **kwargs):
        # NOTE(mdietz): physical net uuid maps to the transport zone uuid
        # physical net type maps to the transport/connector type
        # if type maps to 'bridge', then segment_id, which maps
        # to vlan_id, is conditionally provided
        LOG.debug("Creating new lswitch for %s network %s" %
                  (context.tenant_id, network_name))

        tenant_id = context.tenant_id
        with self.get_connection() as connection:
            switch = connection.lswitch()
            if network_name is None:
                network_name = network_id
            switch.display_name(network_name[:40])
            tags = tags or []
            tags.append({"tag": tenant_id, "scope": "os_tid"})
            if network_id:
                tags.append({"tag": network_id, "scope": "neutron_net_id"})
            switch.tags(tags)
            LOG.debug("Creating lswitch for network %s" % network_id)

            # TODO(morgabra) It seems like this whole interaction here is
            # broken. We force-add either the id/type from the network, or
            # the config default, *then* we still call _config_provider_attrs?
            # It seems like we should listen to the network then fall back
            # to the config defaults, but I'm leaving this as-is for now.
            pnet = phys_net or CONF.NVP.default_tz
            ptype = phys_type or CONF.NVP.default_tz_type
            switch.transport_zone(pnet, ptype)

            # When connecting to public or snet, we need switches that are
            # connected to their respective public/private transport zones
            # using a "bridge" connector. Public uses no VLAN, whereas private
            # uses VLAN 122 in netdev. Probably need this to be configurable
            self._config_provider_attrs(connection, switch, phys_net,
                                        phys_type, segment_id)

            # NOTE(morgabra) A hook for any statically-configured tz bindings
            # that should be added to the switch before create()
            # TODO(morgabra) I'm not sure the normal usage of provider net
            # attrs, and which should superscede which. This all needs a
            # refactor after some discovery probably.
            self._add_default_tz_bindings(context, switch, network_id)

            res = switch.create()
            try:
                uuid = res["uuid"]
                return uuid
            except TypeError:
                LOG.exception("Unexpected return from NVP: %s" % res)
                raise

    def get_lswitch_ids_for_network(self, context, network_id):
        """Public interface for fetching lswitch ids for a given network.

        NOTE(morgabra) This is here because calling private methods
        from outside the class feels wrong, and we need to be able to
        fetch lswitch ids for use in other drivers.
        """
        lswitches = self._lswitches_for_network(context, network_id).results()
        return [s['uuid'] for s in lswitches["results"]]

    def _lswitches_for_network(self, context, network_id):
        with self.get_connection() as connection:
            query = connection.lswitch().query()
            query.tagscopes(['os_tid', 'neutron_net_id'])
            query.tags([context.tenant_id, network_id])
            return query

    def _lswitch_from_port(self, context, port_id):
        with self.get_connection() as connection:
            query = connection.lswitch_port("*").query()
            query.relations("LogicalSwitchConfig")
            query.uuid(port_id)
            port = query.results()
            if port['result_count'] > 1:
                raise Exception("Could not identify lswitch for port %s" %
                                port_id)
            if port['result_count'] < 1:
                raise Exception("No lswitch found for port %s" % port_id)
            cfg = port['results'][0]["_relations"]["LogicalSwitchConfig"]
            return cfg["uuid"]

    def _get_security_group(self, context, group_id):
        with self.get_connection() as connection:
            query = connection.securityprofile().query()
            query.tagscopes(['os_tid', 'neutron_group_id'])
            query.tags([context.tenant_id, group_id])
            query = query.results()
            if query['result_count'] != 1:
                raise sg_ext.SecurityGroupNotFound(id=group_id)
            return query['results'][0]

    def _get_security_group_id(self, context, group_id):
        return self._get_security_group(context, group_id)['uuid']

    def _get_security_group_rule_object(self, context, rule):
        ethertype = rule.get('ethertype', None)
        rule_clone = {}

        ip_prefix = rule.get('remote_ip_prefix', None)
        if ip_prefix:
            rule_clone['ip_prefix'] = ip_prefix
        profile_uuid = rule.get('remote_group_id', None)
        if profile_uuid:
            rule_clone['profile_uuid'] = profile_uuid
        for key in ['protocol', 'port_range_min', 'port_range_max']:
            if rule.get(key):
                rule_clone[key] = rule[key]

        with self.get_connection() as connection:
            secrule = connection.securityrule(ethertype, **rule_clone)

            direction = rule.get('direction', '')
            if direction not in ['ingress', 'egress']:
                raise AttributeError(
                    "Direction not specified as 'ingress' or 'egress'.")
            return (direction, secrule)

    def _check_rule_count_per_port(self, context, group_id):
        with self.get_connection() as connection:
            ports = connection.lswitch_port("*").query().security_profile_uuid(
                '=', self._get_security_group_id(
                    context, group_id)).results().get('results', [])
            groups = (port.get('security_profiles', []) for port in ports)
            return max([self._check_rule_count_for_groups(
                context, (connection.securityprofile(gp).read()
                          for gp in group))
                        for group in groups] or [0])

    def _check_rule_count_for_groups(self, context, groups):
        return sum(len(group['logical_port_ingress_rules']) +
                   len(group['logical_port_egress_rules'])
                   for group in groups)

    def _get_security_groups_for_port(self, context, groups):
        if (self._check_rule_count_for_groups(
                context,
                (self._get_security_group(context, g) for g in groups))
                > self.limits['max_rules_per_port']):
            raise q_exc.DriverLimitReached(limit="rules per port")

        return [self._get_security_group(context, group)['uuid']
                for group in groups]