Building Clouds One 1.4

January 2010, CESGA Santiago de Compostela, Spain

Building Clouds with OpenNebula 1.4

Ruben S.Montero
(rubensm@dacya.ucm.es)
University Complutense of Madrid

Course Overview
● Day one (January 18th, Moorning)
● Introduction to Cloud Computing (FAST!!!)
● Installing OpenNebula 1.4
● Configure OpenNebula 1.4 (storage, hypervisor and network)
● Administration of an OpenNebula Cloud (hosts, users)
● Day one (January 18th, Afternoon)
● Basic usage (networks, VMs)
● More on usage (VMs, context and scheduling) Private Cloud
● Configuring an Hybrid Cloud with Amazon EC2 Hybrid Cloud
● Day two (January 19th, Moorning)
● Public Cloud interfaces: The EC2 Query API Public Cloud

● Advance Topics: Adapt OpenNebula 1.4 to your datacenter

Building Clouds with OpenNebula 1.4 Ruben S. Montero

Course Overview
● In this course you will learn to:
● Describe the benefits and characteristics of virtual infrastructures
and IaaS clouds
● Describe the characteristics and architecture of the different
clouds that can be deployed with OpenNebula 1.4
● Plan and architect a private cloud
● Design, Use and Manage Virtual infrastructures (cloud
applications)
● Build public and hybrid clouds
● Adapt OpenNebula 1.4 to your datacenter

This box contains interesting messages

This is a console output, for hands on, checking configuration files...
# This is the root prompt
$ This is oneadmin prompt



PART I: Building your Private Cloud

Ruben S.Montero

Cloud Computing in a Nutshell
What Who
Software as a Service On-demand access End-user
to any application (does not care about hw or sw)

Platform as a Service Platform for building Developer
and delivering web (no managing of the underlying hw
applications & swlayers)

Infrastructure as a
Service ﾺ Delivery of a raw System Administrator
computer (complete management of the
infrastructure computer infrastructure)

In this course

Physical Infrastructure


The IaaS Clouds a Four Point Check List

●
Simple Interface
● Raw Infrastructure Resources
● Total control of the resources
● Capacity leased in the form of
VMs
● Complete Service-HW
decoupling
● Pay-as-you-go
● A single user can not get all the
resources
● Elastic & “infinite” Capacity


The Anatomy of an IaaS Cloud

VM VM VM VM VM VM
Cloud API
(web) Service Service

Virtual Infrastructure Manager (VIM)

Network

Image
Repositories
(Storage)
Physical Infrastructure


Why Virtual Infrastructure Manager
a Virtual Infrastructure Manager?
●
VMs are great!!...but something more is needed
● Where did/do I put my VM? (scheduling & monitoring)
● How do I provision a new cluster node? (clone & context)
● What MAC addresses are available? (networking)
● Provides a uniform view of the resource pool
● Life-cycle management and monitoring of VM
● The VIM integrates Image, Network and Virtualization

VM VM VM VM

OpenNebula (VIM)

hypervisor hypervisor hypervisor hypervisor


Planning the Installation: System Overview

● Executes the OpenNebula Services
● Usually acts as a classical cluster front-end

● Repository of VM images
● Multiple backends (LVM, iSCSI..)

● Modular components to interact
with the cluster services
● Types:
storage, monitoring,
virtualization and network

● Provides physical resources to VMs
The same host can be can be a ● Must have a hypervisor installed
the front-end and a node


Planning the Installation: Working in the Front-End ...
● Choose your installation mode: system wide (/usr, /etc...) or self-
contained (under $ONE_LOCATION)
● Install software dependencies
● Runtime (also ssh and openssl)
# apt-get install ruby
# apt-get install sqlite3
# apt-get install libxmlrpc-c3
● Development libraries and tools (only needed to build OpenNebula)
# apt-get install libxmlrpc-c3-dev libsqlite3-dev libssl-dev
# apt-get install g++ scons
# apt-get install ruby-dev rubygems rake

The following will give you a faster XML parser (faster CLI)

# apt-get install libxml-parser-ruby1.8 libxslt1-dev libxml2-dev
# gem install nokogiri

SET SYSTEM LOCALES TO ENGLISH (oneadmin should be enough...)

# cat /etc/default/locale
LANG="en_US.UTF-8"

● The Users of the private cloud:
● oneadmin: Account to run the daemons, manage the system and do all
the low-level operations (e.g. start VMs, move images...).
● Users: create and manage their own VMs and networks. Need to be
defined in OpenNebula
● Installation layout
● We will use the /srv/cloud directory to place the OpenNebula software
● /srv/cloud/one will hold the OpenNebula installation
● /srv/cloud/images will do our image repository in the course
# tree /srv
/srv/
`-- cloud
|-- images
`-- one
|-- SRC

The oneadmin account must be created system wide (i.e. front-end and all the nodes) you
can use NIS, or a local account with the same ID's in all the hosts. Users do not need a UNIX
account in the nodes, nor in the front-end.

● Hands on...
# groupadd -g 3000 cloud

# mkdir /srv/cloud
# chmod g+ws cloud /srv/cloud
# mkdir /srv/cloud/images

# useradd -d /srv/cloud/one -g cloud -u 3000 -s /bin/bash -m oneadmin

Create the file-system hierarchy with the oneadmin account

$ id
uid=3000(oneadmin) gid=3000(cloud) grupos=3000(cloud)

$ mkdir /srv/cloud/images
$ mkdir SRC

We will place the OpenNebula source code in SRC


● Preparing the storage for the private cloud...
● Image Repository: Any storage medium for the VM images (usually a
high performing SAN)
– OpenNebula supports multiple back-ends (e.g. LVM for fast cloning)
– The front-end must have access to the repository
● VM Directory: The home of the VM in the cluster node
– Stores checkpoints, description files and VM disks
– Actual operations over the VM directory depends on the storage medium
– Should be shared for live-migrations
– You can go on without a shared FS and use the SSH back-end
– Defaults to $ONE_LOCATION/var/$VM_ID

Dimensioning the Storage... Example: A 64 core cluster will typically run around 80VMs,
each VM will require an average of 10GB of disk space. So you will need ~800GB for
/srv/cloud/one, you will also want to store 10-15 master images so ~200GB for
/srv/cloud/images. A 1TB /srv/cloud will be enough for this example setup.


● In this course we will use NFS to share the VM directories
● The Image Repository is /srv/cloud/images


● Hands on...

# apt-get install nfs-kernel-server

Export /srv/cloud to your nodes
- only need /srv/cloud/one/var
- we also export $HOME of oneadmin for easy SSH key configuration
- No really need to export /srv/cloud/images

# vi more /etc/exports
/srv/cloud 84.21.XX.YY(rw,async,no_subtree_check,no_root_squash)

# /etc/init.d/nfs-kernel-server stop
# /etc/init.d/nfs-kernel-server start


● Networking for the private cloud
● OpenNebula management operations uses a ssh
connections, it does not require a performing NIC
● Image traffic, may require the movement of heavy
files (VM images, checkpoints). Dedicated storage
links may be a good idea
● VM demands, consider the typical requirements of
your VMs. Several NICs to support the VM traffic
may be a good idea
● OpenNebula relies on bridge networking for the
VMs


Planning the Installation: Working in the Nodes ...
● Install software dependencies
● We need SSH daemon running in the cluster nodes (check it!)
● Runtime dependencies:
# apt-get install ruby
● Users
● Create the oneadmin account (use same UID and GID)
# groupadd -g 3000 cloud
# useradd -d /srv/cloud/one -g cloud -u 3000 -s /bin/bash oneadmin

● Storage
● Recreate the installation layout and configure NFS to mount the VM dirs
# mkdir /srv/cloud
# chmod g+ws cloud /srv/cloud

# apt-get install nfs-common
# vi /etc/fstab
x.y.z.w:/srv/cloud /srv/cloud nfs soft,intr,rsize=32768,wsize=32768,rw
0 0

Planning the Installation: SSH Configuration
● Enable password-less SSH access to cluster nodes for the
oneadmin account:
DO NOT PROTECT PRIVATE KEY WITH A PASSWORD
$ ssh-keygen
Generating public/private rsa key pair.
...
Enter passphrase (empty for no passphrase):
Enter same passphrase again:

$ cp ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys

Tell ssh client not to ask to add hosts to known_hosts (optional)

$ cat /srv/cloud/one/.ssh/config
Host *
StrictHostKeyChecking no

TEST!
$ ssh localhost
$ ssh 84.21.xx.yy

You may need to exchange keys with the nodes (not here as we share /srv/cloud/one the
oneadmin home, and so the ~/.ssh directory

Planning the Installation: The Hypervisor ...
● Installing the Hypervisor
● OpenNebula supports KVM, Xen and Vmware (even simultaneously).
This course applies to KVM and Xen
● Refer to the hypervisor documentation for additional (and better
information) on setting up them.
● Setting up KVM and libvirt (Ubuntu Karmic)
● Install the packages

#apt-get install qemu libvirt-bin kvm

● The network set up in this course
peth0, the physical NIC

Physical Host eth0 (bridge)
eth0 (bridge)
84.21.xx.yy
84.21.xx.yy

VM
VM VM
VM
192.168.170.xx
192.168.170.xx 192.168.169.xx
192.168.169.xx


● Hands on...
Remove the default bridge in Ubuntu
#rm /etc/libvirt/qemu/networks/autostart/default.xml
#ifconfig virbr0 down
#brctl delbr virbr0

Set up the network for the course (equivalent to a Xen setup)
- bridge.sh will create the bridge at boot time (see next slide)
- Execute it in the libvirt-bin start script

#vim /etc/libvirt/bridge.sh
#chmod +x /etc/libvirt/bridge.sh

#vi /etc/init.d/libvirt-bin
...
case "$1" in
start)
if check_start_libvirtd_option; then
log_daemon_msg "Starting $DESC" "$NAME"
/etc/libvirt/set_up_bridge.sh
...


#!/bin/sh
create_bridge () {
ip link show p$1 > /dev/null 2>&1
if [ $? -ne 0 ] ; then
ifdown $1

ip link set $1 down
ip link set $1 name p$1
ip link set promisc on dev p$1
ip addr add 0.0.0.0 dev p$1

brctl addbr $1
brctl stp $1 off
brctl addif $1 p$1

ip link set dev p$1 up
ip link set dev $1 up

echo "1" > /proc/sys/net/ipv4/ip_forward
fi
}

create_bridge eth0
route del -net 84.21.xx.0/24 dev peth0
ifup eth0

● Check the network configuration
root@formiga-5:/srv# ifconfig
eth0 Link encap:Ethernet HWaddr 00:1e:4f:e0:5e:9f
...
lo Link encap:Bucle local
...
peth0 Link encap:Ethernet HWaddr 00:1e:4f:e0:5e:9f
...
# brctl show
bridge name bridge id STP enabled interfaces
eth0 8000.001e4fe05e9f no peth0

● Add oneadmin to the libvirt group
# usermod -G kvm,libvirtd oneadmin

● Test the installation for the oneadmin account
$ virsh list
Connecting to uri: qemu:///system
Id Name State
----------------------------------


Planning the Installation: Checklist


Installing OpenNebula 1.4
● Grab the source code and compile it!
~/SRC$ wget http://dev.opennebula.org/attachments/download/103/one-
1.4.0.tar.gz
~/SRC$ tar xzvf one-1.4.0.tar.gz
~/SRC$ cd one-1.4/
~/SRC$ scons

● Install the software in /srv/cloud/one (ONE_LOCATION)
$ export ONE_LOCATION=/srv/cloud/one/
$ ./install.sh -d $ONE_LOCATION

Check install.sh -h for other options

● Check and explore the installation tree
~$ ls -F
bin/ etc/ examples.desktop include/ lib/ share/ SRC/ var/


Configuring OpenNebula: The configuration interface
● $ONE_LOCATION/etc/oned.conf
● General configuration

● Defines the drivers used in the private cloud

● Match-making scheduler (default)
● Placement policies configured per VM

$ONE_LOCATION/etc/im_*/im_*.conf
●

● Defines monitoring probes

● $ONE_LOCATION/etc/vmm_*/vmm_*.conf

● Defaults values for the hypervisor

● $ONE_LOCATION/etc/tm_*/tm_*.conf
● Defines action for generic storage operations


Configuring OpenNebula: The oned.conf file
● General configuration attributes
● Monitoring intervals, HOST_MONITORING_INTERVAL
VM_POLLING_INTERVAL
● VM_DIR: Path to the VM directory for all the cluster nodes.
● Network parameters, MAC_PREFIX, NETWORK_SIZE
● PORT : Port where oned will listen for xml-rpc calls
● DEBUG_LEVEL

HOST_MONITORING_INTERVAL = 60
VM_POLLING_INTERVAL = 60

#VM_DIR = /srv/cloud/one/var

MAC_PREFIX = "00:01"
NETWORK_SIZE = 254

PORT = 2633
DEBUG_LEVEL = 3


● Information Drivers, to monitor cluster nodes
● name: identifies the driver
● executable: absolute or relative to $ONE_LOCATION/lib/mads
● arguments: a probe configuration file
IM_MAD = [
name = "im_kvm",
executable = "one_im_ssh",
arguments = "im_kvm/im_kvm.conf" ]

● Transfer Drivers, to interface with the storage
● executable: path to driver executable
● arguments: storage commands configuration file
TM_MAD = [
name = "tm_nfs",
executable = "one_tm",
arguments = "tm_nfs/tm_nfs.conf" ]


● Virtualization Drivers, to interface the hypervisors
● executable: absolute or relative to $ONE_LOCATION/lib/mads
● arguments: (not needed for the distribution drivers)
● default: default values for the hypervisor
● type: format of the VM description file to be passed to the driver: xen, kvm
or xml

VM_MAD = [
name = "vmm_kvm",
executable = "one_vmm_kvm",
default = "vmm_kvm/vmm_kvm.conf",
type = "kvm" ]

● Hooks, custom programs that are executed on specific events, e.g. VM
creation.
● Hands on... Check and adjust the values of oned.conf for your cloud

Configuring OpenNebula: Accounts
● Accounts in OpenNebula
● oneadmin, has enough privileges to perform any operation on any
object. It is created the first time OpenNebula is started using the
ONE_AUTH data
● Regular user accounts must be created by oneadmin and they can only
manage their own objects.
● Configuring the oneadmin account
● Environment variables: ONE_AUTH, ONE_LOCATION and ONE_XMLRPC
$ tail .bashrc
export ONE_LOCATION=/srv/cloud/one
export ONE_AUTH=$HOME/.one/one_auth
export PATH=$PATH:$ONE_LOCATION/bin

● Create the password file
$ mkdir .one
$ cd .one
$ cat one_auth
oneadmin:onecloud


Configuring OpenNebula: Start & Stop
● Use the one script
$ source .bashrc
$ echo $ONE_AUTH
/srv/cloud/one/.one/one_auth

$one start
oned and scheduler started

$ more $ONE_LOCATION/var/oned.log
Thu Jan 14 18:03:11 2010 [ONE][I]: Init OpenNebula Log system
Thu Jan 14 18:03:11 2010 [ONE][I]: Log Level: 3
[0=ERROR,1=WARNING,2=INFO,3=DEBUG]
Thu Jan 14 18:03:11 2010 [ONE][I]: ------------------------------------
Thu Jan 14 18:03:11 2010 [ONE][I]: OpenNebula Configuration File
Thu Jan 14 18:03:11 2010 [ONE][I]: ------------------------------------

Be sure to configure the oneadmin account (specially, create the ONE_AUTH file) before
starting OpenNebula for the first time.

Configuring OpenNebula: Hosts
● Cluster nodes are defined with
● Hostname of the cluster node or IP
● Information Driver to be used to monitor the host
● Storage Driver to clone, delete, move or copy images into the host
● Virtualization Driver to boot, stop, resume VMs in the host
● Cluster nodes are managed with the onehost utility
● Create & delete hosts
● List the hosts in the cluster
● Show detailed information from a host
● Enable/Disable a host


Configuring OpenNebula: Hosts
● Hands on... configure the hosts of your private cloud
$ onehost create 84.21.xx.yy im_kvm vmm_kvm tm_nfs
$ onehost create 84.21.xx.yy im_kvm vmm_kvm tm_nfs

$ onehost list
ID NAME RVM TCPU FCPU ACPU TMEM FMEM STAT
0 84.21.xx.yy 0 0 0 100 0 0 on
1 84.21.xx.zz 0 0 0 100 0 0 on

$ tail -f $ONE_LOCATION/var/oned.log
Thu Jan 14 18:07:39 2010 [InM][I]: Monitoring host 84.21.xx.yy (0)
Thu Jan 14 18:07:39 2010 [InM][I]: Monitoring host 84.21.xx.zz (1)
Thu Jan 14 18:07:43 2010 [InM][D]: Host 0 successfully monitored.
Thu Jan 14 18:07:44 2010 [InM][D]: Host 1 successfully monitored.

$ onehost list
0 84.21.xx.yy 0 200 184 184 2017004 1848172 on
1 84.21.yy.zz 0 200 200 200 2017004 1857172 on

$ onehost show 0

● Hands on... Explore and test the onehost command in your cloud

Configuring OpenNebula: Users
● Users are defined within OpenNebula by:
● ID unique identifier for the user
● Name of the user, used for authentication
● Password used for authentication
● Users are managed with the oneuser utility
● Create & delete users
● List the users in the cluster
● Hands on... create new users in your private cloud and configure the
“user” UNIX account
$ oneuser create helen mypass
User “Helen” should put helen:mypass in $ONE_AUTH

$ oneuser list
UID NAME PASSWORD ENABLE
0 oneadmin c24783ba96a35464632a624d9f829136edc0175e True
2 helen 34a91f713808846ade4a71577dc7963631ebae14 True

$ oneuser delete helen


Configuring OpenNebula: Log Files
● The operations of the OpenNebula daemon and scheduler are
logged in:
● oned: $ONE_LOCATION/var/oned.log, Its verbosity is set by
DEBUG_LEVEL in $ONE_LOCATION/etc/oned.conf.
● Scheduler (mm_sched): All the scheduler information is collected
into the $ONE_LOCATION/var/sched.log file.
● VM logs and files are in $ONE_LOCATION/var/<VM_ID>,
more in a few slides...
● Drivers can activate ONE_MAD_DEBUG in the associated RC file
(or in $ONE_LOCATION/etc/defaultrc)



PART II: Using your Private Cloud

Ruben S.Montero

Using the Private Cloud: Virtual Networks
● A Virtual Network in OpenNebula
● Defines a separated MAC/IP address space to be used by VMs
● Each virtual network is associated with a physical network through
a bridge
● Virtual Networks can be isolated (at layer 2 level) with ebtables
and hooks
● Virtual Network definition
● Name, of the network
● Type
– Fixed, a set of IP/MAC leases
– Ranged, defines a network range
● Bridge, name of the physical bridge in the physical host where
the VM should connect its network interface.
● Virtual Networks are managed with the onevnet utility
Networks created by oneadmin are public, i.e. can be used by VMs of any other user


● Hands on...
$ vi red.net
NAME = "Red LAN"
TYPE = RANGED
BRIDGE = eth0
NETWORK_SIZE = C
NETWORK_ADDRESS = 192.168.169.0

$ vi blue.net
NAME = "Blue LAN"
TYPE = FIXED
BRIDGE = eth0
LEASES = [IP=192.168.170.5]
LEASES = [IP=192.168.170.10]
LEASES = [IP=192.168.170.15]
LEASES = [IP=192.168.170.20]
LEASES = [IP=192.168.170.25]

$ onevnet -v create blue.net
$ onevnet -v create red.net

● Hands on... explore the use of onevnet list, show, delete


● Using a Virtual Network with your VMs
● Define NICs attached to a given virtual network. The VM will get a
NIC with a free MAC in the network and attached to the
corresponding bridge
#A VM with two interfaces each one in a different vlan
NIC=[NETWORK="Blue LAN"]
NIC=[NETWORK="Red LAN"]

#Ask for a specific IP/MAC of the Red vlan
NIC=[NETWORK="Red LAN", IP=192.168.0.3]

● Prepare the VM to use the IP. Sample scripts to set the IP based
on the MAC are provided for several Linux distributions.
IP-MAC address correspondence

I P: 10. 0. 1. 2

MAC: 02: 01: 0A: 00: 01: 02

oned.conf IP Address


Using the Private Cloud: Virtual Machines
● Preparing a VM to be used with OpenNebula
● You can use any VM prepared for the target hypervisor
● Hint I: Place the vmcontext.sh script in the boot process to make
better use of vlans
● Hint II: Do not pack useless information in the VM images:
– swap. OpenNebula can create swap partitions on-the-fly in the target
host
– Scratch or volatile storage. OpenNebula can create plain FS on-the-
fly in the target host
● Hint III: Install once and deploy many; prepare master images
● Hint IV: Do not put private information (e.g. ssh keys) in the
master images, use the CONTEXT
● Hint V: Pass arbitrary data to a master image using CONTEXT


● Virtual Machine Life-cycle


● A Virtual Machine in OpenNebula
● A capacity in terms memory and CPU
● A set of NICs attached to one or more virtual networks
● A set of disk images, to be “transfered” to/from the execution host.
● A state file (optional) or recovery file, with the memory image of a
running VM plus some hypervisor specific information.
● Virutal Machines are defined in a VM template
● Each VM has an unique ID in OpenNebula the VM_ID
● All the files (logs, images, state files...) are stored in
$ONE_LOCATION/var/<VM_ID>


● Virtual Machine Definition File (VM templates)
#---------------------------------------
# Name of the VM
#---------------------------------------
NAME = "vm-example" # Optional, Default: one-$VMID

#---------------------------------------
# Capacity
#---------------------------------------
CPU = "amount_of_requested_CPU"
MEMORY = "amount_of_requested_MEM"
VCPU = "number of virtual cpus"

#---------------------------------------
# OS and boot options
#---------------------------------------
OS = [
kernel = "path_to_os_kernel", # para-virtualization
initrd = "path_to_initrd_image", # para-virtualization
kernel_cmd = "kernel_command_line",
root = "device to be mounted as root"
bootloader = "path to the boot loader exec”
boot = "device to boot from" ]


#---------------------------------------
# Network Interfaces
#---------------------------------------

NIC = [
network = "name_of_the_virtual_network",
ip = "ip_address",
bridge = "name_of_bridge_to_bind_if",
target = "device_name_to_map_if",
mac = "HW_address",
script = "path_to_script_to_bring_up_if",
Model = "NIC model"]

#---------------------------------------
# I/O Interfaces
#---------------------------------------

INPUT = [
type = "mouse|tablet",
bus = "usb|ps2|xen" ]


#---------------------------------------
# I/O Interfaces
#---------------------------------------

GRAPHICS = [
type = "vnc|sdl",
listen = "IP-to-listen-on",
port = "port_for_VNC_server",
passwd = "passwor_for_VNC_server" ]

#---------------------------------------
# Raw Hypervisor attributes
#---------------------------------------

RAW = [
type = "xen|kvm",
data = "raw_domain_configutarion"]

Not all the parameters are supported for each hypervisor. Complete reference and
examples for all sections in
http://www.opennebula.org/doku.php?id=documentation:rel1.4:template


● Hands on... define a ttylinux VM
NAME = ttylinux
CPU = 0.1
MEMORY = 64

DISK = [
source = "/srv/cloud/images/ttylinux/ttylinux.img",
target = "hda",
readonly = "no" ]

NIC = [ NETWORK = "Red LAN" ]

FEATURES = [ acpi="no" ]

#This may be useful to debug your VMs (can use also console)

GRAPHICS = [
type = "vnc",
listen = "loclahost",
port = "5902",
keymap="es"]


● Fix the apparmor libvirt issue for the cluster nodes (bug in 9.10)

# tail /etc/apparmor.d/abstractions/libvirt-qemu
# https://launchpad.net/bugs/457716
#include <abstractions/private-files-strict>
owner @{HOME}/ r,
owner @{HOME}/** rw,
/srv/cloud/one/var/** rw,

# /etc/init.d/apparmor restart


● Virtual Machines are managed with the onevm utility
● Operations: create, deploy shutdown, livemigrate, stop, cancel,
resume, suspend, delete, restart
● Information: list, show, top, history
$ onevm create ttylinux.one

$ onevm list
ID USER NAME STAT CPU MEM HOSTNAME TIME
1 oneadmin ttylinux pend 0 0 00 00:00:28

$ onevm top

● Hands on...
● Create a couple of VMs and check virtual network usage (onevnet)
● Try control operations with the VMs stop, shutdown, migrate, resume...
● Modify the template
– Add one more NIC for the Blue network
– Add another DISK for VM data (type=”fs”, format=”ext2”)
Building Clouds with OpenNebula 1.4
● Ruben S. Montero

● Context contains data to be passed to the VM at boot time

#---------------------------------------
# Context for the VM
# values can be:
# $<template_variable>
# $<template_variable>[<attribute>]
# $<template_variable>[<attribute>, <attribute2>=<value2>]
# $<vm_id>.<context_var>
#---------------------------------------
CONTEXT = [
var_1 = "value_1",#Will be in context.sh as var_1=”val_1” (sh syntax)
var_n = "value_n",#Will be in context.sh as var_N=”val_N” (sh syntax)
files = "space-separated list of paths to include in context device",
target= "device to attach the context device" ]

● Hands on... Add custom ssh keys the VM image
● Check boot process of the ttylinux VM (systemrc.local) it will
– mount iso (do it yourself and see the ISO layout...)
– Source context.sh
– In this example it will execute init.sh so you can try anything
CONTEXT = [
files = "/srv/cloud/images/ttylinux/init.sh
/srv/cloud/one/.ssh/id_rsa.pub",
target = "hdc",
root_pubkey = "id_rsa.pub"
]

$ more init.sh
#!/bin/bash
if [ -f /mnt/context/context.sh ]
then
. /mnt/context/context.sh
fi
if [ -f /mnt/context/$ROOT_PUBKEY ]; then
cat /mnt/context/$ROOT_PUBKEY >> /root/.ssh/authorized_keys
fi


● Tunning the placement of VMs with the Match-making scheduler
● First those hosts that do not meet the VM requirements are filtered
out (REQUIREMENTS)
● RANK is evaluated for the remaining hosts
● That with the highest RANK is used for the VM
● Placement policies are specified per VM

#---------------------------------------
# Scheduler
#---------------------------------------
# Use Host Monitor attributes
REQUIREMENTS = "Bool_expression_for_reqs"
RANK = "Arith_expression_to_rank_hosts"

● Hands on... try a simple VM pinning
REQUIREMENTS = "HOSTNAME=”...”"

● Hands on... try a simple load-aware policy
RANK = FREECPU


PART III: Using External Cloud Providers
(Hybrid Cloud)

Ruben S.Montero

Hybrid Cloud Computing: Overview
● VMs can be local or remote
● VM connectivity has to be configured, usually VPNs

● External Clouds are like any other host
● Placement constraints


Installing the Hybrid Cloud Components
● OpenNebula distribution includes drivers to build hybrid clouds with
Amazon EC2 and Elastic Hosts
● Additional requirements: EC2 libraries and tools
# apt-get install ec2-api-tools ec2-ami-tools

● Hands on... try the EC2 tools (ec2-*)
$ export EC2_PRIVATE_KEY=/srv/cloud/one/ec2/pk.pem

$ export EC2_CERT=/srv/cloud/one/ec2/cert.pem

$ ec2-describe-images
IMAGE ami-0742a66e /rubensm-
amis.s3.amazonaws.com/image.manifest.xml 418314910487
available private i386 machine
IMAGE ami-e142a688 rubensm-
amis.s3.amazonaws.com/image.manifest.xml 418314910487
available private i386 machine


Configuring the EC2 Hybrid Cloud Driver
● Hands on... Add the following drivers to oned.conf
IM_MAD = [
name = "im_ec2",
executable = "one_im_ec2",
arguments = "im_ec2/im_ec2.conf" ] # No. of instances of each type

VM_MAD = [
name = "vmm_ec2",
executable = "one_vmm_ec2",
arguments = "vmm_ec2/vmm_ec2.conf", # Defaults, e.g. keypair
type = "xml" ]

TM_MAD = [ #No actual transfers are made by OpenNebula to EC2
name = "tm_dummy",
executable = "one_tm",
arguments = "tm_dummy/tm_dummy.conf" ]

● Hands on... Check the values of the driver configurations files


● Configure the account to be used with Amazon EC2

$ vim $ONE_LOCATION/etc/vmm_ec2/vmm_ec2rc
#----------------------------------------------------------------------
# EC2 API TOOLS Configuration.
#----------------------------------------------------------------------
EC2_HOME=/usr
EC2_PRIVATE_KEY="/srv/cloud/one/ec2/pk.pem"
EC2_CERT="/srv/cloud/one/ec2/cert.pem"

● Restart the OpenNebula daemon, and check that the new drivers are
loaded
$ one stop; one start
$ more $ONE_LOCATION/var/oned.log
Fri Jan 15 18:16:46 2010 [VMM][I]: Loading Virtual Machine Manager driv
Fri Jan 15 18:16:46 2010 [VMM][I]: Loading driver: vmm_kvm (KVM)
Fri Jan 15 18:16:47 2010 [VMM][I]: Driver vmm_kvm loaded.
Fri Jan 15 18:16:47 2010 [VMM][I]: Loading driver: vmm_ec2 (XML)
Fri Jan 15 00:16:47 2010 [InM][I]: Loading Information Manager drivers.
Fri Jan 15 00:16:47 2010 [InM][I]: Loading driver: im_kvm
Fri Jan 15 00:16:47 2010 [InM][I]: Driver im_kvm loaded
Fri Jan 15 00:16:47 2010 [InM][I]: Loading driver: im_ec2

● Amazon EC2 cloud is manage by OpenNebula as any other cluster
node
● You can use several accounts by adding a driver for each account (use
the arguments attribute, -k and -c options). Then create a host that uses
the driver
● You can use multiple EC2 zones, add a driver for each zone (use the
arguments attribute, -u option), and a host that uses that driver
● You can limit the use of EC2 instances by modifying the IM file
● Hands on... Create your EC2 hybrid cloud by adding a new host
$ onehost create ec2 im_ec2 vmm_ec2 tm_dummy

$ onehost list
0 84.21.x.y 0 200 200 200 2017004 1667080 on
1 84.21.x.z 1 200 200 200 2017004 1681676 on
2 ec2 0 500 500 500 8912896 8912896 on


Using the EC2 Hybrid Cloud
● Virtual Machines can be instantiated locally or in EC2
● The template must provide a description for both instantiation
methods.
● The EC2 counterpart of your VM (AMI_ID) must be available for
the driver account
● The EC2 VM template attribute:
EC2 = [
AMI = "ami_id for this VM",
KEYPAIR = "the keypair to use the instance",
AUTHORIZED_PORTS = "ports to access the instance",
INSTANCETYPE = "m1.small...",
ELASTICIP = "the elastic ip for this instance",
CLOUD = "host (EC2 cloud) to use this description with"
]


● Hands on... Add an EC2 counterpart to the ttylinux image
$vi ttylinux.one
#EC2 template machine, this will be use wen submitting this VM to EC2
EC2 = [ AMI="ami-ccf615a5",
KEYPAIR="keypair",
AUTHORIZED_PORTS="22",
INSTANCETYPE=m1.small]

#Add this if you want to use only EC2 cloud
REQUIREMENTS = "HOSTNAME = "ec2""

● Hands on... Create the VM and check progress
$ onevm create ttylinux.one
$ onevm list
16 oneadmin one-16 runn 0 0 ec2 00 00:00:35
$ ec2-describe-instances
RESERVATION r-5eff7536 418314910487 default
INSTANCE i-bac3f0d2 ami-0572946c pending
keypair0 m1.small 2010-01-14T23:32:35+0000 us-
east-1a aki-a71cf9ce ari-a51cf9cc monitoring-
disabled


● Hands on... Log in the EC2 instance when running
$ onevm show 17
...
VIRTUAL MACHINE TEMPLATE
CPU=0.5
...
EC2=[
AMI=ami-ccf615a5,
AUTHORIZED_PORTS=22,
INSTANCETYPE=m1.small,
KEYPAIR=keypair ]
IP=ec2-72-44-62-194.compute-1.amazonaws.com
...
REQUIREMENTS=HOSTNAME = "ec2"
VMID=17

$ ssh -i keypair.pem root@ec2-72-44-62-194.compute-1.amazonaws.com
Linux ip-10-212-134-128 2.6.21.7-2.fc8xen-ec2-v1.0 #2 SMP Tue Sep 1
10:04:29 EDT 2009 i686
root@ip-10-212-134-128:~#

This costs money!
$ onevm shutdown 17
$ onehost disable ec2
$ onehost list


PART IV: Share your Cloud!
(Cloud Interfaces)

Ruben S.Montero

The Public Cloud: Overview
● You can use multiple interfaces for the Cloud
● Transparent to your setup:

● Hypervisor

● Storage Model

● Hybrid configuration

● Client tools uses EC2 libraries
● Potential integration with EC2 tools

(EC2_URL problems for example)
● Provided in the OpenNebula

distribution
● Includes a simple S3 replacement

● Supports HTTP and HTTPS protocols
● EC2 authentication based on OpenNebula

credentials
● Public Cloud users need an OpenNebula

account


Installing the Public Cloud Components
● OpenNebula distribution supports two Cloud interfaces: the EC2
Query API and OCCI
● Additional requirements: EC2 development library, web server and
web framework
# gem install amazon-ec2
# sudo gem install sinatra
# sudo gem install thin

# gem install sequel
# apt-get install curl libcurl3 libcurl4-gnutls-dev libopenssl-ruby
# gem install curb
# gem install sqlite3-ruby

Add a “FQDN” for our Public Cloud
# vim /etc/hosts
127.0.0.1 localhost
#127.0.1.1 formiga-z

84.21.x.y formiga-z formiga-z.opennebula.org


Configuring the Public Cloud
● The EC2 service is configured in $ONE_LOCATION/etc/econe.conf
● Hands on... Study the configuration file and adjust it to your cloud
$ more econe.conf
# OpenNebula administrator user, the one_auth contents
USER=oneadmin
PASSWORD=onecloud

# OpenNebula sever contact information
ONE_XMLRPC=http://localhost:2633/RPC2

# Host and port where econe server will run keep FQDNs
SERVER=formiga-y.opennebula.org
PORT=4567

# Configuration for the image repository
# IMAGE_DIR will store the Cloud images, check space!
DATABASE=/srv/cloud/one/var/econe.db
IMAGE_DIR=/srv/cloud/public_repo/

# VM types allowed and its template file
VM_TYPE=[NAME=m1.small, TEMPLATE=m1.small.erb]


● You have to define the correspondence between types (simple) and
local instantiation of VMs (hard, you should be fine by now)
● Capacity allocated by this VM type (CPU, MEMORY)
● Your cloud requirements, e.g. force to use a given kernel (OS) or place
public VMs in a given set of cluster nodes (REQUIREMENTS)
● The network used by Public VMs (NIC)
● VM Types are defined in econe.conf. Templates for the VM
templates are in $ONE_LOCATION/etc/ec2query_templates
● Templates for VM Types are erb files <% Ruby code here >, you
should not need to modify that.


● Hands on... Prepare the m1.small type of your cloud use ttylinux.one
as a reference
$ more m1.small.erb

NAME = eco-vm

CPU = 0.1
MEMORY = 64

DISK = [ source = <%= erb_vm_info[:img_path] %>,
clone = yes,
target = hda,
readonly = no]

#You have to create this network, and it should be owned by oneadmin

NIC = [ NETWORK = "Red LAN" ]

IMAGE_ID = <%= erb_vm_info[:img_id] %>
INSTANCE_TYPE = <%= erb_vm_info[:instance_type ]%>


● Hands on... start the econe server
$ unset EC2_URL
$ econe-server start

$ lsof -i

Check $ONE_LOCATION/var/econe-server.log for errors


Using the Public Cloud
● The econe-tools are a subset of the functionality provided by the
onevm utility, and resembles the ec2-* cli
● Image related commands are:
● econe-upload, place an image in the Cloud repo and returns ID
● econe-describe-images, lists the images
● econe-register, register an image not really needed in 1.4
● Instance related commands are:
● econe-run-instances, starts a VM using an image ID
● econe-describe-instances, lists the VMs
● econe-terminate-instances, shutdowns a VM
● User authentication is based in the OpenNebula credentials
● AWSAccessKeyId is OpenNebula's username
● AWSSecretAccessKey is OpenNebula's password


● Pass your credentials to the econe-tools by (in this order)
● Commands arguments (--access-key <username>, --secret-key <pass>)
● Environemt EC2_ACCESS_KEY and EC2_SECRET_KEY
● Environment ONE_AUTH
● Point econe-tools to your target cloud
● Command arguments (--url <http|https>://<fqdn>:<port>) port needed in
not the default for the protocol
● EC2_URL environment
●
Hands on... upload the ttylinux image, and start it using the public
cloud interface.
● Compare the econe-* (public view) and onevm (local view) evolution and
information
● Check the template build by the econe server (onevm show)


$ econe-upload -U http://formiga-x.opennebula.org:4567 --access-key
oneadmin --secret-key onecloud /srv/cloud/images/ttylinux/ttylinux.img
Success: ImageId 1

$ export EC2_URL="http://formiga-15.opennebula.org:4567"
$ econe-describe-images -H
Owner ImageId Location
-----------------------------------------------------------------------
oneadmin 1 /srv/cloud/public_repo/1

$ econe-run-instances 1
oneadmin 1 18 m1.small

$ econe-describe-instances
oneadmin 18 1 pending
192.168.169.5 m1.small

This is the local view not accessible to public cloud users
$ onevm list
19 oneadmin eco-vm runn 0 65536 84.21.x.y 00 00:01:34

$ onevm show 19


Configuring SSL access for the Public Cloud
● SSL security is handle by a proxy that forwards the request to the
EC2 Query Service and takes back the answer to the client
● Requirements:
● A server certificate for the SSL connections
● An HTTP proxy that understands SSL
● EC2Query Service configuration to accept petitions from the proxy
● Hands on... Install the proxy (lighttpd) and get the certificates for your
cloud
# apt-get install lighttpd
# apt-get instatall ssl-cert

# /usr/sbin/make-ssl-cert generate-default-snakeoil
# cat /etc/ssl/private/ssl-cert-snakeoil.key /etc/ssl/certs/ssl-cert-
snakeoil.pem > /etc/lighttpd/server.pem


● Hands on... configure the lighttpd proxy
# vim /etc/lighttpd/lighttpd.conf
server.modules = (
"mod_access",
"mod_alias",
"mod_accesslog",
"mod_compress",
"mod_proxy"
...
## bind to port (default: 80)
server.port = 8443
...
#### proxy module
proxy.server = ( "" =>
("" =>
(
"host" => "127.0.0.1",
"port" => 4567
)
)
)
#### SSL engine
ssl.engine = "enable"
ssl.pemfile = "/etc/lighttpd/server.pem"
server.reject-expect-100-with-417 = "disable"

● Hands on... configure the econe server
$ vim /srv/cloud/one/etc/econe.conf

#SERVER=formiga-15.opennebula.org
SERVER=127.0.0.1
PORT=4567

# SSL proxy that serves the API (set if is being used)
SSL_SERVER=formiga-15.opennebula.org

● Hands on... by pass the EC2 library URL checking
# sudo vim /var/lib/gems/1.8/gems/amazon-ec2-0.7.9/lib/AWS/EC2.rb
Comment out line 12

● Hands on... restart services (lighttpd and econe-server) and try your
new SSL cloud access (https://formiga-x.opennebula.org:8443)



PART V: Customizing your Cloud

Ruben S. Montero

Customizing and Extending your Cloud
● You can customize your cloud by:
● Tunning or adapting the transfer operations to your storage back-end
● Adding new monitorization probes to improve the VM placement
● Adjusting VM operations to your hypervisor installation
● Trigger custom actions on specific VM events (e.g. “on VM creation
update the accounting DB” or “on VM shutdown send an email”)
● You can extend your cloud by:
● Developing new drivers for other hypervisors
● Developing new drivers for other storage back-ends
● Developing Cloud applications using the OpenNebula API or the Cloud
APIs

OpenNebula is very scripting friendly, drivers can be written in any language. You can
modify the current ones or use them as templates for new ones.


Customizing the Storage of your Cloud
OpenNebula issue generic storage operations (check
var/<vm_id>/transfer.*)

Transfer Driver process the core script

Action programs interpret the semantics
of generic actions depending on the
storage back.-end

● OpenNebula requests the following abstract operations over a
VM image
● CLONE: This action will basically make a copy of the image from
ORIGIN to DESTINATION.
● LN: Creates a symbolic link in DESTINATION that points to
ORIGIN
● MKSWAP: Generates a swap image in DESTINATION. The size is
given in ORIGIN in MB.
● MKIMAGE: Creates a disk image in DESTINATION and populates
it with the files inside ORIGIN directory.
● DELETE: Deletes ORIGIN file or directory.
● MV: Moves ORIGIN to DESTINATION.


● Actions are defined in
$ONE_LOCATION/etc/tm_<storage>/tm_<storage>.conf
$ more /srv/cloud/one/etc/tm_ssh/tm_ssh.conf
CLONE = ssh/tm_clone.sh
LN = ssh/tm_ln.sh
MKSWAP = ssh/tm_mkswap.sh
MKIMAGE = ssh/tm_mkimage.sh
DELETE = ssh/tm_delete.sh
MV = ssh/tm_mv.sh

● Actions scripts are placed in
$ONE_LOCATION/lib/tm_commands/<storage>/
$ ls /srv/cloud/one/lib/tm_commands/ssh/
tm_clone.sh tm_delete.sh tm_mkimage.sh tm_mv.sh
tm_context.sh tm_ln.sh tm_mkswap.sh


● Hands on... Take a look to the tm_clone.ssh

. $TMCOMMON
...
log "Creating directory $DST_DIR"
exec_and_log "ssh $DST_HOST mkdir -p $DST_DIR"
...
case $SRC in
http://*)
log "Downloading $SRC"
exec_and_log "ssh $DST_HOST wget -O $DST_PATH $SRC"
;;

*)
log "Cloning $SRC"
exec_and_log "scp $SRC $DST"
;;
esac

exec_and_log "ssh $DST_HOST chmod a+w $DST_PATH"

● Hands on... Check the semantics of other operations for the ssh
storage, e.g. tm_ln.ssh


Customizing the Information System
● OpenNebula gets host information by executing an arbitrary number of
probes
● A probe is a program that returns the monitorization metric in the form
METRIC_NAME = VALUE
● Probes are configured in
$ONE_LOCATION/etc/im_<hypervisor>/im_<hypervisor>.conf
And placed in
$ONE_LOCATION/lib/im_probes
● Probe information is mainly used for VM placement


Customizing the Information System
● Hands on... Take a look to the default probes defined for KVM
$ more /home/ruben/Virtual/one/etc/im_kvm/im_kvm.conf
cpuarchitecture=architecture.sh
nodename=name.sh
cpu=cpu.sh
kvm=kvm.rb

$ more name.sh
#!/bin/sh

echo HOSTNAME=`uname -n`

● Hands on... Create a new probe that returns the number of VMs in
RUNNING_VMS (e.g. you can use virsh, pgrep kvm...). Use the new
metric to pack VMs (RANK=RUNNING_VMS).


Customization with Hooks: Network Isolation

● IN: Only Ethernet frames from a MAC in Red LAN
● OUT: Only Ethernet frames from the MAC assigned by

OpenNebula

● Networks are isolated at layer 2
● You can put any TCP/IP service as part of the VMs (e.g.

DHCP, nagios...)


● Requirements (this has to be done in all the cluster nodes)
● Check that ebtables package is installed
● Allow oneadmin to use the ebtables command through sudo
#visudo
...
oneadmin ALL=(ALL) NOPASSWD: /sbin/ebtables *
...

● Configure the hooks for OpenNebula
VM_HOOK = [
name = "ebtables-start",
on = "running",
command = "/srv/cloud/one/share/hooks/ebtables-kvm",
arguments = "one-$VMID",
remote = "yes" ]
VM_HOOK = [
name = "ebtables-flush",
on = "done",
command = "/srv/cloud/one/share/hooks/ebtables-flush",
arguments = "",
remote = "yes" ]


● Hands on... Start a couple of VMs in Networks Red and Blue.
● Check the ebtables rules in the hosts
● Check connectivity between VMs
● Change the network mask of the VMs and check connectivity
● Shutdown and check the ebtables rules


Summary
● You should be able to:
● Describe the benefits and characteristics of virtual infrastructures
and IaaS clouds
● Describe the characteristics and architecture of the different
clouds that can be deployed with OpenNebula 1.4
● Plan and architect a private cloud
● Design, Use and Manage Virtual infrastructures (cloud
applications)
● Build public and hybrid clouds
● Adapt OpenNebula 1.4 to your datacenter


Building Clouds One 1.4

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to Building Clouds One 1.4 (20)

Recently uploaded (20)

Building Clouds One 1.4