[OpenStack 하반기 스터디] Interoperability with ML2: LinuxBridge, OVS and SDN

Interoperability with ML2:
LinuxBridge, OVS and SDN
Sungman Jang (LinuxBridge& OVS)
Guohua Lee (SDN)

ML2 with Linux Bridge & OVS
Part 1

Environment
• Host: HP Elitebook 820 G1 (Intel
i5-4300U, 2.50GHz / 12GB RAM)
/ Windows 7 Enterprise
• VirtualBox: 5.0.6 with Extension
Pack installed
• Guest: Ubuntu x64 14.04
• OpenStack: Liberty
(Provider Network)
• 설치목차
• 1. 기본 프로그램 설치(python-
openstackclient chrony) 및 네트
워크 인터페이스 설정
2. 컨트롤러 전용 프로그램 설치
(mariasql, rabbitmq)
3-1. Keystone 설치
3-2. 기본 계정 설정하기
(Admin/Demo)
3-3. Keystone 설치 후 동작 확인
4-1. Glance 설치
4-2. Glance 설치 후 동작 확인
• -----------
• 5-1. Nova 설치
• 5-2. Nova 설치 후 동작 확인
• 6-1. Neutron 설치
• 6-2. Neutron 설치 후 동작 확인
• 7. Horizon 설치

Provider Networks
• Provider networks generally offer
simplicity, performance, and
reliability at the cost of flexibility.
• Operators can build on this
minimal deployment to enable
more cloud networking features.
공유기=
<OpenStack Networking Guide> <VirtualBox로 구현한 네트워크 구조>

Why LinuxBridge and OVS?
• LinuxBridge (since 1999/01/25):
– A Linux bridge is more powerful than a pure
hardware bridge because it can also filter and
shape traffic. The combination of bridging and
firewalling is done with the companion project
ebtables.
• Open vSwitch (since 2009/07/29)
– It is designed to enable massive network
automation through programmatic extension, while
still supporting standard management interfaces
and protocols (e.g. NetFlow, sFlow, IPFIX,
RSPAN, CLI, LACP, 802.1ag).

Why OVS so complicated?
https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux_OpenStack_Platform/5/html/Cloud_Administrator_Guide/section_networking-scenarios.html
• Ideally, the TAP device vnet0 would be connected directly to the integration bridge, br-int. Unfortunately, this isn't possible because of how OpenStack security
groups are currently implemented. OpenStack uses iptables rules on the TAP devices such as vnet0 to implement security groups, and Open vSwitch is not
compatible with iptables rules that are applied directly on TAP devices that are connected to an Open vSwitch port.
• Networking uses an extra Linux bridge and a veth pair as a workaround for this issue. Instead of connecting vnet0 to an Open vSwitch bridge, it is connected to
a Linux bridge, qbrXXX. This bridge is connected to the integration bridge, br-int, through the (qvbXXX, qvoXXX) veth pair.
LinuxBridgeOpenvSwitch

What is ML2
• The Modular Layer 2 (ML2) Plugin is a framework allowing OpenStack
Networking to simultaneously utilize the variety of layer 2 networking
technologies found in complex real-world datacenters.
POST /v2.0/ports.json HTTP/1.1
content-type: application/json
{
"port": {
"admin_state_up": true,
"device_id": "d6b4d3a5-c700-476f-b609-1493dd9dadc0",
"name": "port1",
"network_id": "6aeaf34a-c482-4bd3-9dc3-7faf36412f12"
}
}
LB Agent

Expected Scenario
/etc/neutron/plugins/ml2/ml2_conf.ini
[ml2]
type_drivers = flat,vlan,gre,vxlan
tenant_network_types =
mechanism_drivers = linuxbridge,openvswitch
/etc/neutron/dhcp_agent.ini
[DEFAULT]
verbose = True
interface_driver = neutron.agent.linux.interface.BridgeInterfaceDriver
dhcp_driver = neutron.agent.linux.dhcp.Dnsmasq
dhcp_delete_namespaces = True
Mech Driver:
Linuxbridge,
Openvswitch
+
Type Driver:
flat,vlan,
gre,vxlan
Mech Driver:
Linuxbridge,
+
Type Driver:
vxlan
Mech Driver:
Openvswitch
+
Type Driver:
vxlan
neutron.agent.linux.interface.OVSInterfaceDriver

Oh… No…
• /etc/neutron/neutron.conf
• /etc/neutron/plugins/ml2/ml2_conf.ini
• /etc/neutron/dhcp_agent.ini

Bridge, Switch and Router
• Bridge: A network bridge is network hardware that connects multiple network segments.
• Switch: A network switch is a multiport network bridge that uses hardware addresses to
process and forward data at the data link layer (layer 2) of the OSI model.
• Router: A router is connected to two or more data lines from different networks. When
multiple routers are used in interconnected networks, the routers exchange information
about destination addresses using a dynamic routing protocol.
주요 특징 Classical Router Layer 3 Switch
주요 OSI Layer Layer 3 Layer 3
Routing 방법 Software (CPU + Software) Hardware (ASIC chip)
L2 지원 Ethernet, TokerRing, FDDI, ATM, WAN FastEthernet, Gigabit Ethernet
forwarding 성능 Slow Fast
Latency 약 200 ms < 10 ms (100 Mbps)
관리/program가능 매우 높음 적음
지원 Protocol All IP (일부 IPX)
Routing Protocol All RIP1,2 OSPF 등
WAN 지원 지원함 지원하지 않음
비용 높음 낮음
표 출처: http://koreahojin.egloos.com/v/2341317

Booting a VM
• 0. _build_instance() & _allocate_network()
• 1. REST (C->N): Create Port
• 2. vif_driver.plug() & Add a port tap with external_ids set
– ‘Neutron OVS agent’ loop to detect port update
– Get the Neutron port id from the external_ids
– Call get_device_details(port_id) to Neutron Server
• 3. Bind_port() & Set up the ovs port
• 4. update_device_up()

ML2 in Code
neutron.ml2.type_drivers =
flat = neutron.plugins.ml2.drivers.type_flat
local = neutron.plugins.ml2.drivers.type_local
vlan = neutron.plugins.ml2.drivers.type_vlan
geneve = neutron.plugins.ml2.drivers.type_geneve
gre = neutron.plugins.ml2.drivers.type_gre
vxlan = neutron.plugins.ml2.drivers.type_vxlan
neutron.ml2.mechanism_drivers =
ovsvapp = neutron.plugins.ml2.drivers.ovsvapp.mech_driver
opendaylight = neutron.plugins.ml2.drivers.opendaylight.driver
logger = neutron.tests.unit.plugins.ml2.drivers.mechanism_logger
test = neutron.tests.unit.plugins.ml2.drivers.mechanism_test
linuxbridge = neutron.plugins.ml2.drivers.linuxbridge.mech_driver.mech_linuxbridge
openvswitch = neutron.plugins.ml2.drivers.openvswitch.mech_driver.mech_openvswitch
hyperv = neutron.plugins.ml2.drivers.hyperv.mech_hyperv
l2population = neutron.plugins.ml2.drivers.l2pop.mech_driver
ofagent = neutron.plugins.ml2.drivers.ofagent.driver
mlnx = neutron.plugins.ml2.drivers.mlnx.mech_mlnx
brocade = networking_brocade.vdx.ml2driver.mechanism_brocade
brocade_fi_ni = neutron.plugins.ml2.drivers.brocade.fi_ni.mechanism_brocade_fi_ni
fslsdn = neutron.plugins.ml2.drivers.freescale.mechanism_fslsdn
sriovnicswitch = neutron.plugins.ml2.drivers.mech_sriov.mech_driver.mech_driver
fake_agent = neutron.tests.unit.plugins.ml2.drivers.mech_fake_agent
sdnve = neutron.plugins.ml2.drivers.ibm.mechanism_sdnve
neutron.ml2.extension_drivers =
test = neutron.tests.unit.plugins.ml2.drivers.ext_test
testdb = neutron.tests.unit.plugins.ml2.drivers.ext_test
port_security = neutron.plugins.ml2.extensions.port_security
qos = neutron.plugins.ml2.extensions.qos
neutronpluginsml2driver_api.pysetup.cfg

ML2 in Code
neutronpluginsml2driver_api.pyquantum-2013.1.5.tardistquantum-2013.1.5quantumpluginslinuxbridgelb_quantum_plugin.py
RPC: Agent->Plugin
RPC: Plugin->Agent

class OpenDaylightMechanismDriver(driver_api.MechanismDriver):
"""Mechanism Driver for OpenDaylight.
This driver was a port from the NCS MechanismDriver. The API
exposed by ODL is slightly different from the API exposed by NCS,
but the general concepts are the same.
"""
def initialize(self):
self.url = cfg.CONF.ml2_odl.url
self.timeout = cfg.CONF.ml2_odl.timeout
self.username = cfg.CONF.ml2_odl.username
self.password = cfg.CONF.ml2_odl.password
required_opts = ('url', 'username', 'password')
for opt in required_opts:
if not getattr(self, opt):
raise cfg.RequiredOptError(opt, 'ml2_odl')
self.odl_drv = OpenDaylightDriver()
# Postcommit hooks are used to trigger synchronization.
def create_network_postcommit(self, context):
self.odl_drv.synchronize('create', odl_const.ODL_NETWORKS, context)
ML2 for ODL
from networking_odl.common import constants as odl_const
from networking_odl.ml2 import mech_driver
from oslo_config import cfg
from oslo_log import log
from neutron.plugins.ml2 import driver_api as api
https://github.com/openstack/networking-odl
(ODL backend implementation)

def sendjson(self, method, urlpath, obj):
"""Send json to the OpenDaylight controller."""
headers = {'Content-Type': 'application/json'}
data = jsonutils.dumps(obj, indent=2) if obj else None
url = '/'.join([self.url, urlpath])
LOG.debug("Sending METHOD (%(method)s) URL (%(url)s) JSON (%(obj)s)",
{'method': method, 'url': url, 'obj': obj})
r = requests.request(method, url=url,
headers=headers, data=data,
auth=self.auth, timeout=self.timeout)
r.raise_for_status()
def sync_single_resource(self, operation, object_type, context):
"""Sync over a single resource from Neutron to OpenDaylight.
Handle syncing a single operation over to OpenDaylight, and correctly
filter attributes out which are not required for the requisite
operation (create or update) being handled.
"""
# Convert underscores to dashes in the URL for ODL
object_type_url = object_type.replace('_', '-')
try:
obj_id = context.current['id']
if operation == odl_const.ODL_DELETE:
self.out_of_sync |= not self.client.try_delete(
object_type_url + '/' + obj_id)
else:
filter_cls = self.FILTER_MAP[object_type]
if operation == odl_const.ODL_CREATE:
urlpath = object_type_url
method = 'post'
attr_filter = filter_cls.filter_create_attributes
elif operation == odl_const.ODL_UPDATE:
urlpath = object_type_url + '/' + obj_id
method = 'put'
attr_filter = filter_cls.filter_update_attributes
resource = context.current.copy()
attr_filter(resource, context)
self.client.sendjson(method, urlpath,
{object_type_url[:-1]: resource})
def synchronize(self, operation, object_type, context):
"""Synchronize ODL with Neutron following a configuration change."""
if self.out_of_sync:
self.sync_full(context._plugin)
else:
self.sync_single_resource(operation, object_type, context)
ML2 for ODL
self.odl_drv.synchronize('create', odl_const.ODL_NETWORKS, context)
self.context = context.get_admin_context()
curl -X POST -H "Content-Type: application/json" -u admin:admin http://localhost:8080/controller/nb/v2/neutron/networks -d '{ "networks": [ { "id": "d17696ea-1d2f-42ff-
9a8f-895eca97b5ed", "name": "Network-1", "tenant_id": "93397de8-7b3c-4450-a700-40d5d0120e89","router:external": true,"shared": true} ]}'
https://github.com/openstack/networking-odl

ML2 for ONOS
entity_path = 'networks'
resource = context.current.copy()
onos_utils.send_msg(self.onos_path, self.onos_auth, 'post',
entity_path, {'network': resource})
def send_msg(onos_path, onos_auth, msg_type, entity_path, entity=None):
"""Send message to the ONOS controller."""
path = '/'.join([onos_path, entity_path])
hdr = {'Content-Type': 'application/json'}
body = jsonutils.dumps(entity, indent=2) if entity else None
LOG.debug("Sending MSG_TYPE (%(msg)s) URL (%(path)s) "
"OBJECT (%(entity)s) BODY (%(body)s)",
{'msg': msg_type, 'path': path,
'entity': entity, 'body': body})
req = requests.request(method=msg_type, url=path,
headers=hdr, data=body,
auth=onos_auth)
# Let's raise voice for an error
req.raise_for_status()
def __init__(self):
self.onos_path = cfg.CONF.onos.url_path
self.onos_auth = (cfg.CONF.onos.username, cfg.CONF.onos.password)
self.vif_type = portbindings.VIF_TYPE_OVS
self.vif_details = {portbindings.CAP_PORT_FILTER: True}
class Network(model_base.BASEV2, HasId, HasTenant):
"""Represents a v2 neutron network."""
name = sa.Column(sa.String(attr.NAME_MAX_LEN))
ports = orm.relationship(Port, backref='networks')
subnets = orm.relationship(
Subnet, backref=orm.backref('networks', lazy='joined'),
lazy="joined")
status = sa.Column(sa.String(16))
admin_state_up = sa.Column(sa.Boolean)
mtu = sa.Column(sa.Integer, nullable=True)
vlan_transparent = sa.Column(sa.Boolean, nullable=True)
rbac_entries = orm.relationship(rbac_db_models.NetworkRBAC,
backref='network', lazy='joined',
cascade='all, delete, delete-orphan')
https://github.com/openstack/networking-onos
ONOS JSON REST API

SDN, OpenFlow, SDN Controller
How to integrate SDN controller with
OpenStack?
Demo
Contents

SDN(Software Defined Networking) 개념
소프트웨어 정의 네트워킹으로 불리 우며 개방형 API(예를 들어 OpenFlow)를
통해 네트워크 장비에서 하드웨어 기능과 소프트웨어 기능을 분리하여
소프트웨어 기반 컨트롤러에 의해 제거/관리하는 접근 방식임
SDN Architecture
SDN, OpenFlow, SDN Controller
Infrastructure
Layer
Network
Control Layer
Application
Layer
Network Aware
Application
SDN Controller
(e.g., onos, opendaylight, floodlight …)
Physical
Network Device
Virtual
Network Device
Control Data Plane Interface
(e.g., OpenFlow)
Network Information
API API
Network Control
*출처: “OpenFlow를 이용한 SDN입문”
Control Plane
Data Plane

Neutron + SDN Controller
Neutron Components
192.168.0.0/24
192.168.1.0/24
117.16.149.0/24
https://github.com/openstack/networking-onos
Mechanism driver: ONOSMechanismDriver (instead of
OpenvswitchMechanismDriver)
Service plugin: ONOSL3Plugin(instead of L3RouterPlugin)
Detail Configuration: http://heavenkong.blogspot.kr/2015/12/integrate-
onosemu-with-openstack-kilo.html
x
x
x
수정

Five VMs or host machines are needed
Prerequisite
No usage CPU Memory Disk
1 OpenStack
Controller Node
2 Core 4 GB 20 GB
2 OpenStack Network
Node
2 Core 4 GB 20 GB
3 ONOS Controller
Node
2 Core 4 GB 20 GB
4 Compute Node1 1 Core 2 GB 20 GB
5 Compute Node2 1 Core 2 GB 20 GB
*Reference: SONA User Guide
https://wiki.onosproject.org/display/ONOS/How+to+set+up+environment

System Configuration
Controller/Network/ONOS Node Compute Node
Supporting Services
Database
MySQL
Message Broker
rabbitmq
Basic Services
Identity: Keystone
Image Service: Glance
Compute: Nova mgmt
Networking: Neutron
Server
Dashboard: Horizon
ONOS Controller
Java 8
Karaf & mvn
onos (emu)
Basic Services
Compute:
Nova Hypervisor
KVM/libvirt
Networking:
Ml2 Plug-in
OVS Agent
Network Interface
Management
(192.168.0.3/24)
External (117.16.149.115/24)
Network Interface
Management
(192.168.0.4/24)Instance Tunnels
(192.168.1.3/24)
Instance Tunnels
(192.168.1.4/24)
Networking:
ONOS Ml2 Plug-in
OVS Agent
Layer 3 Agent
DHCP Agent
Metadata Agent

Openstackswitching from SONA architecture
https://wiki.onosproject.org/display/ONOS/SONA+Architectur
e
They only support switching part
They didn’t support gre tunnel
Networking-onos는 nova로 부터 ONOS사이
REST API형식으로 전달해주는 작용을 함
Openstackswitching은 REST API를 받아
instance들을 위한 가상 네트워크를 만들어 줌
Openstackswithcing app uses the REST API call from onos-
networking plugin only when pushing the DHCP
mapping(VM IP-MAC pair), and it pushes flow rules when
VMs are created and PORT_ADDED/UPDATED events occurs
We need both networking-onos and
openstackswitching app

Demo
https://youtu.be/S-ANMA5EC24

Instance can not receive fixed IP
It means ONOS can not receive fixed IP
Reference:
http://uni2u.tistory.com/58
http://blog.night9.cc/onos-on-centos7/
https://www.youtube.com/watch?v=qP8nPYhz_Mo
https://wiki.onosproject.org/display/ONOS/Simple+workthrough
http://forum.onosfw.com/t/how-to-integrate-onos-master-1-4-0-
with-openstack-kilo/80
https://github.com/hyunsun/documentations/wiki/Neutron-
ONOS-Integration-for-CORD-VTN
Work to be continued …

[OpenStack 하반기 스터디] Interoperability with ML2: LinuxBridge, OVS and SDN

Recommended

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to [OpenStack 하반기 스터디] Interoperability with ML2: LinuxBridge, OVS and SDN (20)

More from OpenStack Korea Community (20)

Recently uploaded (20)

[OpenStack 하반기 스터디] Interoperability with ML2: LinuxBridge, OVS and SDN

Editor's Notes