Fr

ee

OpenStack Networking Essentials kicks off by describing

the various components of Openstack Neutron and walks
you through installing OpenStack on CentOS using RDO
and Packstack. Further on, you will use various methods
to interface with Neutron in order to create and manage
network resources. You will also get to grips with the
relationships among ports, networks, and subnets
through diagrams and explanations and see how the
logical components are implemented via plugins
and agents. Moving forward, you will learn how virtual
switches are implemented and how to build Neutron
routers. You will also configure networks, subnets, and
routers to provide connectivity to instances using simple
examples. Towards the end, you will configure and
manage security groups, and will observe how these
rules translate to iptables rules on host machines.

Who this book is written for

Install the latest Liberty (2015.2) release

of OpenStack using RDO in VirtualBox
Discover the basics of the Neutron API,
including networks, subnets, and ports
Interact with Neutron using the CLI and
the Horizon dashboard
Create networks and subnets that provide
connectivity to instances
Implement software routers that connect
networks and provide network address
translation
Secure instances using Neutron's security
group functionality

$ 39.99 US
25.99 UK

community experience distilled

P U B L I S H I N G

E x p e r i e n c e

D i s t i l l e d

OpenStack Networking
Essentials
Build and manage networks in OpenStack using Neutron

Prices do not include

local sales tax or VAT
where applicable

Visit www.PacktPub.com for books, eBooks,

code, downloads, and PacktLib.

Sa
m

pl

C o m m u n i t y

James Denton

The book is for those who are new to OpenStack

and Neutron and want to learn cloud networking
fundamentals and get started with OpenStack networking.
Prior networking experience and a virtual or physical
server is recommended in order to follow along with the
concepts demonstrated in the book.

What you will learn from this book

OpenStack Networking Essentials

OpenStack Networking
Essentials

James Denton

In this package, you will find:

The author biography

A preview chapter from the book, Chapter 6 'Routing'
A synopsis of the books content
More information on OpenStack Networking Essentials

About the Author

James Denton has more than 15 years of experience in system administration and

networking and has been deploying, operating, and maintaining OpenStack clouds
since late 2012. He is a Principal Architect at Rackspace, and prior to joining the
Rackspace Private Cloud team, he spent 5 years as an enterprise network security
engineer. James has a bachelor's degree in business management, with a focus on
computer information systems, from Texas State University in San Marcos, Texas. In
his spare time, James enjoys spending time with his wife and son and camping in the
Texas hill country.

Preface
OpenStack is an open source cloud operating system designed to control pools of
compute, storage, and networking resources. This powerful system fosters rapid
innovation while decreasing operational and capital costs. OpenStack has exploded
in popularity in recent years, thanks to its features, flexibility, and overall maturity.
In this book, we will explore the networking component of OpenStack, known as
Neutron. Neutron provides an API for users to build virtual network resources
such as switches, routers, load balancers, and firewalls. We will walk through the
installation of OpenStack using RDO and will look at the core components of the
API, made up of networks, subnets, and ports. By the end of the book, you will have
harnessed the power of OpenStack and Neutron to create and access virtual network
resources of your own.

What this book covers

Chapter 1, OpenStack Networking Components an Overview, provides an introduction
to OpenStack Networking features, components, and the basic physical architectures
required to support an OpenStack cloud.
Chapter 2, Installing OpenStack Using RDO, provides instructions for installing the
Liberty release of OpenStack using RDO on the CentOS 7.1 operating system.
Chapter 3, Neutron API Basics, looks at the core components of the Neutron API,
made up of networks, subnets, and ports, and how they're used to construct
virtual networks.
Chapter 4, Interfacing with Neutron, explores the use of the Horizon dashboard and the
Neutron command-line client to interface with the Neutron API.
Chapter 5, Switching, looks at how Neutron constructs and implements the virtual
network infrastructure to enable the flow of traffic across the cloud.

Preface

Chapter 6, Routing, discusses how Neutron implements virtual routers that provide
routing between Neutron networks and the outside world using source network
address translation and floating IPs.
Chapter 7, Building Networks and Routers, covers basic virtual network architectures
and showcases the traffic flow from client workstations to virtual machine instances
via fixed and floating IPs.
Chapter 8, Security Group Fundamentals, examines the use of Neutron security groups
to secure instance traffic at the virtual switch port and walks you through creating
and managing security groups and associated rules.
Appendix, Configuring VirtualBox, is meant to assist with the setup of a virtual
environment using VirtualBox so that many of the examples throughout the
book can be followed.

What you need for this book

For this book, the following is required:

Operating system:

CentOS Linux 7.1

Software:

VirtualBox 5.0 or higher

RDO (Liberty release)

This book assumes a beginner-to-moderate level of networking experience and

experience with Linux operating systems. While this book will walk you through a
basic installation of OpenStack using RDO, little time will be spent on services other
than Neutron as well as any configuration of OpenStack outside of what's available
via the API. It will be helpful for you have a basic understanding of OpenStack and
its components prior to reading this book.
Internet connectivity is required to install OpenStack packages. An all-in-one
OpenStack deployment will be performed on a single virtual machine within
VirtualBox. CentOS must be installed prior to installing RDO. Alternative
virtualization platforms such as VMware, or physical hardware, are optional.

Routing
In the previous chapter, we discovered how Neutron builds out logical networks
using two of the most popular open source virtual switching platforms for
OpenStack: Open vSwitch and LinuxBridge. In this chapter, we will take a look at
how Neutron implements virtual routers that provide routing between Neutron
networks and the outside world. Concepts that will be covered include network
namespaces, Source Network Address Translation (SNAT), and floating IPs.

The basics of routing in Neutron

If you recall from the previous chapter, users can create and manage networks
known as tenant networks within their respective project without any knowledge of
the underlying infrastructure. By default, instances connected to tenant networks are
isolated from other networks and are unable to access external resources such as the
Internet. Neutron provides connectivity to instances in tenant networks by way of
virtual routers.

Network namespaces
In a reference implementation, virtual routers created in Neutron are implemented
as network namespaces that reside on nodes running the Neutron L3 agent service.
Network namespaces are similar in function to Virtual Routing and Forwarding
(VRF) domains, where multiple instances of a routing table exist to provide
complete network segregation in a single device. The use of network namespaces
allows Neutron to support overlapping subnets across networks. In Linux, network
namespaces can be managed using the ip netns command as the root user or a
user with sudo privileges. Router namespaces follow the qrouter-<router_id>
naming convention.

[ 71 ]

Routing

In addition to providing dedicated routing tables to each namespace,

Linux allows processes like dnsmasq to be run and contained within
namespaces. Neutron uses network namespaces to isolate DHCP services
between networks. DHCP namespaces follow the qdhcp-<network_
id> naming convention.

The Neutron L3 agent service usually runs on controller nodes or dedicated network
nodes, but as we'll see later, the service can also run on compute nodes to help
provide a smaller failure zone when using distributed virtual routers.

Connectivity through a router

At its most basic level, a Neutron router acts as a default gateway for one or more
connected tenant networks, as shown in the following diagram:

Instances in Tenant Network A can use the Neutron router as their default
gateway to communicate with instances in Tenant Network B, and vice versa.
When an external provider network is attached to the Neutron router, the router
can route traffic upstream to its respective gateway device. In most cases, a physical
routing device will be the gateway for a Neutron router. The following diagram
demonstrates a physical gateway and the Neutron router connected to a common
provider network:

[ 72 ]

Chapter 6

In the diagram, a Neutron router is connected to multiple tenant networks and

serves as the default gateway for those networks. The Neutron router is also
connected to a provider network that provides access to external networks,
including the Internet.

[ 73 ]

Routing

Outbound connectivity
By default, Neutron routers will apply Source Network Address Translation
(SNAT) to all outbound traffic from connected tenant networks. This means that, as
traffic exits the virtual router and heads upstream, the router modifies each packet
and changes the source IP address to that of its own external interface. This ensures
that return traffic gets directed back to the virtual router, where the destination
IP address is modified from the router's address back to the original client.
The following diagram demonstrates a router performing SNAT for a virtual
machine instance:

In the diagram, the outbound traffic from VM 1 is modified as it traverses the

Neutron router towards its destination. As each packet leaves the router, the
source address is modified. As the inbound response traffic enters and traverses
the router, the destination address is changed from that of the router to that of the
virtual machine.

Inbound connectivity
In a SNAT scenario, all traffic leaving the router appears to come from the same
address. In addition, inbound connections cannot be made directly to a SNAT
address, which means that that address cannot be used to reach the instance directly.

[ 74 ]

Chapter 6

A floating IP is an address that is used to provide a 1:1 static NAT mapping to

a single fixed IP. In Nova/Neutron-speak, a fixed IP is an IP address associated
with an instance and, by definition, a Neutron port. Floating IPs provide a unique
outbound and inbound address; this allows clients to reach individual virtual
machine instances and other devices. The following diagram demonstrates
a router performing an address translation using a floating IP:

In the diagram, the outbound traffic from VM 1 is modified as it traverses the

Neutron router towards its destination. As each packet leaves the router, the source
address is modified to that of the floating IP. As the inbound response traffic enters
and traverses the router, the destination address is changed from that of the router to
that of the virtual machine. This is similar in operation to the earlier SNAT example.
Rather than using a shared address, however, the floating IP is dedicated to traffic
associated with the fixed IP of the instance. Inbound connections to the floating IP
from an external network are translated to the respective fixed IP and directed to the
appropriate resource or instance.

[ 75 ]

Routing

Types of routers
Neutron routers act as the default gateway for connected tenant networks
and provide outbound and inbound connectivity to the instances they service.
Neutron provides three types of routers to users:

Standalone

Highly available

Distributed

Routers can be created in both the Horizon dashboard and via the Neutron CLI.
As an ordinary user, the type of router that is created via the API is predetermined,
based on a combination of settings found in the Neutron server and L3 agent
configuration files. Users with the admin role are free to define the type of router
to be created using the router-create command, using the following flags:
--distributed {true | false}
--ha {true | false}

Neutron does not expose the router type to users via the API, even with the
router-show command. Users with the admin role, however, can see those details.
The Horizon dashboard limits all users, including administrative users, to the
default router type specified in the configuration. This behavior may change
in future releases.

Standalone routers
A standalone router is a single logical router that is implemented as a single network
namespace on a host running the Neutron L3 agent. Most often, the L3 agent runs
on dedicated network nodes or the controller nodes themselves. By its very nature,
a standalone router is a single point of failure for directly connected networks. If
the node hosting the network namespace experiences issues, connectivity through
the namespace can become limited or completely unavailable. Needless to say, the
failure of a standalone router can result in an unhappy user experience.
Standalone routers have been the default router type since the Folsom release
of OpenStack and are supported by both the Open vSwitch and LinuxBridge
mechanism drivers and agents.

[ 76 ]

Chapter 6

Highly available routers

A Highly Available (HA) router is a single logical router that is implemented as
two or more network namespaces on hosts running the Neutron L3 agent. Like its
standalone counterpart, an HA router is likely to be spread across dedicated network
or controller nodes. An HA router utilizes the keepalived service and the Virtual
Routing Redundancy Protocol (VRRP) between network namespaces to provide
high availability. Only one of the network namespaces acts as a master virtual router
at any given time while the others remain in a backup state awaiting a failover event.
If the active router fails, a backup router will take over quickly. While HA routers
provide redundancy not found with standalone routers, pushing all traffic through
a subset of nodes may still be seen as a bottleneck that can result in poor network
performance.
Highly available routers have been available since the Juno release of OpenStack
and are supported by both the Open vSwitch and LinuxBridge mechanism drivers
and agents.

Distributed virtual routers

A Distributed Virtual Router (DVR) is a single logical router that is implemented
as multiple network namespaces on network and compute nodes. The model
of distributing virtual routers across compute nodes is similar to the multihost
functionality of Nova Network. It offers high availability of networking by limiting
single points of failure to individual compute nodes rather than network nodes.
Distributed virtual routers have been available since the Juno release of OpenStack
and, as of the Liberty release, are supported only by the Open vSwitch mechanism
driver and agent.
Highly available and distributed virtual routers require the use of more
than one host and won't be created as part of the exercises demonstrated
in this book. For a more in-depth look at HA and distributed virtual
routers, check out Learning OpenStack Networking (Neutron), Second Edition,
available through Packt Publishing at https://www.packtpub.
com/virtualization-and-cloud/learning-openstacknetworking-neutron-second-edition.

[ 77 ]

Routing

Managing routers in the dashboard

Like networks, virtual routers can be created and managed within the Horizon
dashboard and by using the Neutron command-line client.

Creating routers within a project

To create a router, follow these steps:
1. Navigate to the Project | Network | Routers section of the Horizon
dashboard and click on the Create Router button in the upper right-hand
corner of the screen, as shown here:

2. A single-step router creation wizard will appear. Name the router in the
Router Name field, as shown here:

3. Click on the Create Router button to complete the wizard and return to the
Routers screen, as shown in the following screenshot:
[ 78 ]

Chapter 6

Congratulations, you just created a virtual router! Behind the scenes, Neutron has
determined the type of router to create and may have implemented one or more
network namespaces as a result. In this environment, the default router type is
standalone, which means a single network namespace will be created that will
serve as the virtual router. In Chapter 7, Building Networks and Routers, we'll
build some common network topologies and observe traffic flow through a
Neutron router.

Viewing the network topology

Now that we have created a router, let's view the resulting network topology. Here
are the steps:
1. From the Project | Network menu, choose Network Topology:

[ 79 ]

Routing

2. Right now, the topology consists of a single network object and a single
router object. Notice that the router and network are not connected. To
connect the router to the network, click on the Add Interface button,
shown here:

3. An interface wizard will appear. From the Subnet menu, select the subnet to
attach the router to:

[ 80 ]

Chapter 6

4. When adding an interface to a router, the router will take on the IP address
defined in the gateway_ip attribute of the selected subnet. Instead, you can
specify a different IP in the IP Address field. Click on the Add interface
button to complete the wizard and return to the network topology screen
pictured here:

The router is now connected to the network! This means that an instance in the
MySimpleNetwork network can use the router as its default gateway.
Additional interfaces can be added to the router, but only from other
networks. A Neutron router should not be connected to the same network
more than once.

[ 81 ]

Routing

Managing routers as an administrator

From the Admin | System | Routers window, administrators can view all routers in
the cloud as well as edit and delete selected routers, as shown here:

Here, we can see the previously created router, MySimpleRouter. Clicking on the
router name provides a limited subset of the actions that are available through the
Project | Network | Routers pane:

Administrators can only delete routers, mark their administrative state UP

or DOWN, view static routes, and add or delete interfaces. All other router
management functions must be done from within the respective Project panel.

[ 82 ]

Chapter 6

Managing routers with the Neutron client

The Neutron command-line client provides additional functionality not found in the
Horizon dashboard, including the ability to specify the type of router to create.

Creating and listing routers

Listing networks with the Neutron client is as easy as using the neutron router-

list command, shown here:

Running the router-list command as an administrator will return all routers

known to Neutron, while running the command as an ordinary user will only return
routers associated with the user's tenant or project. As an example, let's authenticate
as the demo user and run the same command to see what is returned:

As expected, no routers were returned. Routers, like other OpenStack resources, are
associated with a single tenant or project and can only be viewed or managed by the
respective users of those projects or by administrators.

Creating a router
Creating a router with the Neutron client can be accomplished with the neutron
router-create command. In this example, the demo user is logged in and the router
to be created is named MyDemoRouter:
[root@allinone ~(keystone_demo)]# neutron router-create MyDemoRouter

[ 83 ]

Routing

The operation returns a response that can be seen in the following output:

Notice that the output did not return information regarding the router type. As an
ordinary user, that information is not exposed by Neutron and is only available to
users with the admin role.
Copy the id and name values from the output. You will use the router ID when you
add an interface to the router or perform certain other network activities. In some
cases, the router name can be used in lieu of the ID, but only when the name is
unique. In this example, the ID is dfa617ad-3ded-4962-9484-a5e9ce138172,
but the value will be unique in your response.

Adding an interface
Adding an interface to a router with the Neutron client can be accomplished with the
neutron router-interface-add command. To add an interface, you must specify
the router and subnet name or ID.
Users who consume all available addresses in a subnet may find it
necessary to add additional subnets to a network. Neutron allows
multiple subnets to be associated with a single network, and routers
should be connected to each subnet using the router-interface-add
command.

[ 84 ]

Chapter 6

If you recall from Chapter 4, Interfacing with Neutron, we created a network and
subnet as the demo user, named MyDemoNetwork and MyDemoSubnet, respectively. As
the demo user, perform a neutron net-list command to retrieve a list of networks
and associated subnets, like this:

Using the neutron router-interface-add command, add an interface to the

router and attach it to the MyDemoSubnet subnet:

That's it! As a result of adding an interface to the router via the API, Neutron created
a logical Neutron port for the virtual network interface used by the router, created
the virtual interface inside the respective network namespace, attached the interface
to the virtual switch or bridge, and configured an IP address on the virtual interface
that corresponds with the address defined by the gateway_ip attribute of the subnet.
Thanks, Neutron!

Listing router interfaces

To obtain a list of the Neutron ports associated with a router, use the neutron

router-port-list command, seen here:

All ports currently associated with the router will be listed in the output.
As expected, the fixed IP of the port corresponds to the gateway_ip of the
MyDemoSubnet subnet.

[ 85 ]

Routing

Examining the routers

Virtual routers are implemented as network namespaces on one or more nodes. In
our single-node environment, the two routers that have been created so far can both
be found on the same node. Using the ip netns command as root or a user with
sudo privileges, you can list all network namespaces on a host, like so:

The two qrouter namespaces in the list correspond to the two routers we've created
so far:

Using the ip netns exec <namespace> command, you can specify a command to
execute within the specified namespace. Useful commands such as ip, netstat, ps,
and iptables provide details within the scope of the namespace they're executed in.
A quick look at the MySimpleRouter network namespace shows the virtual interface
created by Neutron when we attached the router to the MySimpleSubnet subnet
earlier in this chapter:

Using the ovs-vsctl show command, we can see the qr-2e525501-5f interface
attached to the integration bridge in a local VLAN mapped to MySimpleNetwork:
[ 86 ]

Chapter 6

Additional router interfaces will be connected to the same integration bridge, but
they will be tagged with different local VLAN tags and names based on the network,
subnet, and port the interfaces are associated with.

Summary
In this chapter, we learned that Neutron routers can route between directly
connected tenant networks and external networks using network address translation.
Neutron routers can be configured in a redundant or distributed manner, and
they trade simplicity in their implementation for high availability. Like the virtual
switching infrastructure covered in the previous chapter, users without access to the
underlying infrastructure will be unable to observe how Neutron implements virtual
routers and their respective features. The logical network diagram provided within
the Horizon dashboard, coupled with an understanding of the concepts outlined in
this book, will help you understand what is happening behind the scenes.
In the next chapter, we will take a look at some common virtual network
architectures that can be built by users and will showcase the traffic flow from
client workstations to virtual machine instances using floating IPs.
[ 87 ]

Get more information OpenStack Networking Essentials

Where to buy this book

You can buy OpenStack Networking Essentials from the Packt Publishing website.
Alternatively, you can buy the book from Amazon, BN.com, Computer Manuals and most internet
book retailers.
Click here for ordering and shipping details.

www.PacktPub.com

Stay Connected: