1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
|
.. _using_with_openstack:
************************************************************************
Using Ryu Network Operating System with OpenStack as OpenFlow controller
************************************************************************
This section describes how to setup openstack (nova, quantum) and
ryu-manager.
It is assumed that kvm with libvirt is used and each host machines that run
nova-compute/nova-network has two physical NICs.
It would be possible to deploy it with single NIC machines as described at
the last section.
NOTE: How to use nova isn't described in this document.
Overview
========
Ryu is designed/implemented with for production use in mind, so it cooperates
very well with `OpenStack <http://www.openstack.org/>`_ .
With nova and quantum OVS plugin,
Ryu provides L2 segregation of Multi-tenants without any switch feature/settings
like VLAN. So it's very easy to use/experiment/deploy this segregation as
the below figure.
.. image:: /images/logical-view.png
Physical machine setup
----------------------
The following figure depicts how physical hosts are connected and each daemons
are deployed.
.. image:: /images/physical-view.png
Although the nova-api, nova-scheduler, nova-network and related openstack
daemons are installed in each own physical machines in the above picture,
they can be installed on a physical machine which also runs nova-compute.
Each host machine has two nics and one is connected to management LAN
and other is connected to deployment LAN.
How to install/setup
====================
If you are not familiar with installing/setting up nova/quantum/openvswitch
from the source, please refer to OpenStack document and get back here again.
[
`OpenStack docs <http://docs.openstack.org/>`_ ,
`Nova <http://www.openstack.org/projects/compute/>`_ ,
`Quantum <http://docs.openstack.org/incubation/>`_ ,
`OpenvSwtich and Quantum Part 1 <http://openvswitch.org/openstack/2011/07/25/openstack-quantum-and-open-vswitch-part-1/>`_ ,
`OpenvSwtich and Quantum Part 2 <http://openvswitch.org/openstack/2011/07/25/openstack-quantum-and-open-vswitch-part-1/>`_ ,
`OVS Quantum Plugin Documentation <http://openvswitch.org/openstack/documentation/>`_
]
* Install ryu and run ryu-manager
* install ryu from the source code on the hosts on which you run
* nova-compute,
* quantum-server and
* ryu-manager.
This is because quantum-server and ova quantum agent which runs on
nova-compute node needs ryu-client library to communicate ryu-manager.
Type in ryu source directory::
% python ./setup.py install
* edit /etc/ryu/ryu.conf on the host on which you run ryu-manager
if necessary
No configuration is needed on hosts that runs quantum and ovs quantum
agent.
* run ryu network os::
% ryu-manager [----flagfile /etc/ryu/ryu.conf]
* get nova source and quantum source from github
* They are a bit modified from openstack master tree. They are available
at github for convinience
* https://github.com/osrg/nova/tree/ryu
* https://github.com/osrg/quantum/tree/ryu
clone them by typing the followings in an appropriate directory::
% git clone --branch ryu git://github.com/osrg/nova.git
% git clone --branch ryu git://github.com/osrg/quantum.git
If you prefer https, try those::
% git clone --branch ryu https://github.com/osrg/nova.git
% git clone --branch ryu https://github.com/osrg/quantum.git
* Install nova and quantum as usual.
(And other Openstack related component if necessary. e.g. glance)
Each daemons can be installed in a single machine or in different machines.
Please refer to Openstack documentation for details.
You may want to set up multiple nova-compute nodes for interesting use case.
* Setup nova daemons. (Edit nova.conf)
Specifically configure nova-network and nova-compute
* configure nova-network
* --fixed_ranges=<setup here>
* --network_size=<setup here>
* --network_manager=nova.network.quantum.manager.QuantumManager
* --quantum_connection_host=<quantume server ip address>
* --firewall_driver=nova.virt.libvirt.firewall.NopFirewallDriver
NOP firewall driver is newly introduced for demonstrating Ryu
capability.
If you want, other existing firewall driver can be specified.
But such specification don't have any effect in fact
because ryu directly controls packets to VM instance via OVS bypassing
netfilter/iptables.
* --linuxnet_interface_driver=nova.network.linux_net.LinuxOVSOFInterfaceDriver
* set up OVS on each nova-compute node
If Ubuntu is used, you can install it from packages as
openvswitch-datapath-dkms, openvswitch-common, openvswitch-switch
If you already use bridge, you may need to edit /etc/modules to load
openvswitch kernel module, openvswitch_mod and brcompat_mod, before
bridge module and reboot to unload bridge module.
And then create ovs bridge::
# ovs-vsctl add-br <bridge name: usually br-int>
And if you connect NIC to OVS bridge.::
# ovs-vsctl add-port <bridge name:br-int> <ether interface: e.g. eth<N>>
* configure each nova-compute
* --libvirt_type=kvm
* --libvirt_ovs_integration_bridge=<OVS bridge:br-int>
* --libvirt_vif_type=ethernet
* --libvirt_vif_driver=nova.virt.libvirt.vif.LibvirtOpenVswitchDriver
* install quantum server and have quantum to use OVS pluging
* Edit [PLUGIN] section of /etc/quantum/plugins.ini
* provider = quantum.plugins.openvswitch.ovs_quantum_plugin.OVSQuantumPlugin
* Edit [OVS] section of
/etc/quantum/plugins/openvswitch/ovs_quantum_plugin.ini
In addition to normal quantum OVS settings, add the followings.
* integration-bridge = <OVS bridge name: br-int>
* plugin_driver = quantum.plugins.openvswitch.ovs_quantum_plugin.OFPRyuDriver
* agent_driver = OVSQuantumOFPRyuAgent
* openflow-controller = <ryu-manager IP address>:<ryu openflow port: default 6633>
* openflow-rest-api = <ryu-manager IP address>:<RYU reset API port: default 8080>
* Run quantum server
* install quantum OVS agent on each nova-compute node
* Edit /etc/quantum/plugins/openvswitch/ovs_quantum_plugin.ini
* Run ovs agent::
# ovs_quantum_agent.py -v ./etc/quantum/plugins/openvswitch/ovs_quantum_plugin.ini
* Then as usual openstack nova operation, create user, project, network and
run instances.
* Enjoy!
Testing
=======
Yay, now you have ryu network Operating System set up.
You would want to really they are L2-segregated.
* create multi projects and run instances.
* ping/traceroute between them.
* tcpdump in the instances
The routing between gateway(gw-xxx) of each tenants are disabled
by nova.network.linux_net.LinuxOVSOFInterfaceDriver by installing iptables
rule on nova-network host::
# iptable -t filter -A nova-network-FORWARD --in-interface gw-+ --out-interface gw-+
Thus pinging/tracerouting between VMs in distinct tenants doesn't work.
If you drop the above rule by::
# iptable -t filter -D nova-network-FORWARD --in-interface gw-+ --out-interface gw-+
You will see ping/tracerout works. Please notice that the packets go through
gw-xxx and gw-yyy, not directly.
.. image:: /images/trace-route.png
Caveats
=======
* Run the following daemons in this order
#. Run Ryu network Operating System
#. Run quantum with OVS plugin
#. Run quantum OVS agent
#. run your guest instance
For now, ryu-manager doesn't have persistent store, so if it's rebooted,
all the necessary information must be told again from quantum server/agent.
* nova-manage network delete doesn't work
At this moment, quantum doesn't implement network delete fully yet.
If you issue the command, it fails. And you need to fix nova/quantum DB
by hand using SQL.
Appendix
========
In the above, two physical NIC deployment is described.
Some people may want to use those settings with single NIC machine or even
with single machine.
It would be possible as the following pictures, but we haven't tested those
setting. If you success it, please report it.
single NIC setup
----------------
If your host machines have only single NIC, it would be possible to use
Ryu network Operating System with Linux bridge. However we haven't tested such
setups.
.. image:: /images/compute-node.png
All-in-One Setup
----------------
You can also setup in single physical host as the following picture.
.. image:: /images/minimul-setup.png
You can setup the above environment quickly using DevStack.
#. Install Ubuntu 11.10 (Oneiric)
#. Download Ryu enabled DevStack from github
::
% git clone --branch ryu git://github.com/osrg/devstack.git
#. Start the install
::
% cd devstack; ./stack.sh
It will take a few minutes.
|
