devstack/tools/xen/README.md

Getting Started With XenServer and Devstack

The purpose of the code in this directory it to help developers bootstrap a XenServer 6.2 (older versions may also work) + OpenStack development environment. This file gives some pointers on how to get started.

Xenserver is a Type 1 hypervisor, so it is best installed on bare metal. The OpenStack services are configured to run within a virtual machine (called OS domU) on the XenServer host. The VM uses the XAPI toolstack to communicate with the host over a network connection (see MGT_BRIDGE_OR_NET_NAME).

The provided localrc helps to build a basic environment.

Introduction

Requirements

• An internet-enabled network with a DHCP server on it
• XenServer box plugged in to the same network This network will be used as the OpenStack management network. The VM Network and the Public Network will not be connected to any physical interfaces, only new virtual networks will be created by the install_os_domU.sh script.

Steps to follow

• Install XenServer
• Download Devstack to XenServer
• Customise localrc
• Start install_os_domU.sh script

Brief explanation

The install_os_domU.sh script will:

• Setup XenAPI plugins
• Create the named networks, if they don't exist
• Preseed-Netinstall an Ubuntu Virtual Machine (NOTE: you can save and reuse it, see Reuse the Ubuntu VM), with 1 network interface:
  • eth0 - Connected to UBUNTU_INST_BRIDGE_OR_NET_NAME, defaults to MGT_BRIDGE_OR_NET_NAME
• After the Ubuntu install process finished, the network configuration is modified to:
  • eth0 - Management interface, connected to MGT_BRIDGE_OR_NET_NAME. Xapi must be accessible through this network.
  • eth1 - VM interface, connected to VM_BRIDGE_OR_NET_NAME
  • eth2 - Public interface, connected to PUB_BRIDGE_OR_NET_NAME
• Start devstack inside the created OpenStack VM

Step 1: Install Xenserver

Install XenServer on a clean box. You can download the latest XenServer for free from: http://www.xenserver.org/

The XenServer IP configuration depends on your local network setup. If you are using dhcp, make a reservation for XenServer, so its IP address won't change over time. Make a note of the XenServer's IP address, as it has to be specified in localrc. The other option is to manually specify the IP setup for the XenServer box. Please make sure, that a gateway and a nameserver is configured, as install_os_domU.sh will connect to github.com to get source-code snapshots.

Step 2: Download devstack

On your XenServer host, run the following commands as root:

wget --no-check-certificate https://github.com/openstack-dev/devstack/zipball/master
unzip -o master -d ./devstack
cd devstack/*/

Step 3: Configure your localrc inside the devstack directory

Devstack uses a localrc for user-specific configuration. Note that the XENAPI_PASSWORD must be your dom0 root password. Of course, use real passwords if this machine is exposed.

cat > ./localrc <<EOF
# At the moment, we depend on github's snapshot function.
GIT_BASE="http://github.com"

# Passwords
# NOTE: these need to be specified, otherwise devstack will try
# to prompt for these passwords, blocking the install process.

DATABASE_PASSWORD=my_super_secret
SERVICE_TOKEN=my_super_secret
ADMIN_PASSWORD=my_super_secret
SERVICE_PASSWORD=my_super_secret
RABBIT_PASSWORD=my_super_secret
SWIFT_HASH="66a3d6b56c1f479c8b4e70ab5c2000f5"
# This will be the password for the OpenStack VM (both stack and root users)
GUEST_PASSWORD=my_super_secret

# XenAPI parameters
# NOTE: The following must be set to your XenServer root password!

XENAPI_PASSWORD=my_xenserver_root_password

XENAPI_CONNECTION_URL="http://address_of_your_xenserver"
VNCSERVER_PROXYCLIENT_ADDRESS=address_of_your_xenserver

# Explicitly set virt driver
VIRT_DRIVER=xenserver

# Explicitly enable multi-host for nova-network HA
MULTI_HOST=1

# Give extra time for boot
ACTIVE_TIMEOUT=45

EOF

Step 4: Run ./install_os_domU.sh from the tools/xen directory

cd tools/xen
./install_os_domU.sh

Once this script finishes executing, log into the VM (openstack domU) that it installed and tail the run.sh.log file. You will need to wait until it run.sh has finished executing.

Appendix

This section contains useful information for running devstack in CI environments / using ubuntu network mirrors.

Use a specific Ubuntu mirror for installation

To speed up the Ubuntu installation, you can use a specific mirror. To specify a mirror explicitly, include the following settings in your localrc file:

UBUNTU_INST_HTTP_HOSTNAME="archive.ubuntu.com"
UBUNTU_INST_HTTP_DIRECTORY="/ubuntu"

These variables set the mirror/http/hostname and mirror/http/directory settings in the ubuntu preseed file. The minimal ubuntu VM will use the specified parameters.

Use an http proxy to speed up Ubuntu installation

To further speed up the Ubuntu VM and package installation, an internal http proxy could be used. squid-deb-proxy has prooven to be stable. To use an http proxy, specify:

UBUNTU_INST_HTTP_PROXY="http://ubuntu-proxy.somedomain.com:8000"

in your localrc file.

Reuse the Ubuntu VM

Performing a minimal ubuntu installation could take a lot of time, depending on your mirror/network speed. If you run install_os_domU.sh script on a clean hypervisor, you can speed up the installation, by re-using the ubuntu vm from a previous installation.

Export the Ubuntu VM to an XVA

Given you have an nfs export TEMPLATE_NFS_DIR:

TEMPLATE_FILENAME=devstack-jeos.xva
TEMPLATE_NAME=jeos_template_for_devstack
mountdir=$(mktemp -d)
mount -t nfs "$TEMPLATE_NFS_DIR" "$mountdir"
VM="$(xe template-list name-label="$TEMPLATE_NAME" --minimal)"
xe template-export template-uuid=$VM filename="$mountdir/$TEMPLATE_FILENAME"
umount "$mountdir"
rm -rf "$mountdir"

Import the Ubuntu VM

Given you have an nfs export TEMPLATE_NFS_DIR where you exported the Ubuntu VM as TEMPLATE_FILENAME:

mountdir=$(mktemp -d)
mount -t nfs "$TEMPLATE_NFS_DIR" "$mountdir"
xe vm-import filename="$mountdir/$TEMPLATE_FILENAME"
umount "$mountdir"
rm -rf "$mountdir"