<![CDATA[because we all share ...]]> 2012-02-18T20:57:35-08:00 http://www.likens.us/ Octopress <![CDATA[a process of making a custom HVM virtualized AMI for EC2]]> 2012-01-07T14:46:00-08:00 http://www.likens.us/blog/2012/01/07/amazon-hvm-custom-ami-on-ec2 Forewarning: This post is about a process; it is not meant to give you a complete working basis to create a hvm virtualization AMI. It is here to give you the information and or tools to build your own.

Required Reading Material.

  • hvmloader

  • pv-on-hvm

  • Note: Usage of a vanilla kernel later then 2.6.37+ will have the pv-on-hvm drivers. Pick a kernel later then that at your descretion.

Requirements.

Step 1

booting an instance.
1
2
3
4

ec2-run-instances ami-0da96764 -k ssh_key -t cc1.4xlarge -H
Type  ReservationID   Owner   Groups  Platform
RESERVATION   r-f5743394  000000000000    default
INSTANCE  i-de453ebc  ami-0da96764            pending ssh_key 0       cc1.4xlarge 2012-01-07T23:01:54+0000    us-east-1c              monitoring-disabled                 ebs                 hvm xen     sg-7baf4812 default

  • Take note of the instance-id and the zone the instance is booted in.

  • additionally Amazon limits us to an 8G root volume; so create an 8gig volume in the zone specifieid.

create your EBS Root volume
1
2

ec2-create-volume -z us-east-1c -s 8
VOLUME    vol-03aff26e    8       us-east-1c  creating    2012-01-07T23:03:25+0000
attach the volume to the instance as /dev/sdp
1
2

ec2-attach-volume vol-03aff26e -d /dev/sdp -i i-de453ebc
ATTACHMENT    vol-03aff26e    i-de453ebc  /dev/sdp    attaching   2012-01-07T23:05:26+0000
  • SSH to the instance in question as ec2-user@$hostname
obtain root priviledges and install screen
1
2
3

sudo -s
yum install screen
screen -S ami
install the amazon api and ami tools
1
2
3
4
5

mkdir -p /mnt/work && cd /mnt/work
wget http://s3.amazonaws.com/ec2-downloads/ec2-ami-tools.zip
unzip ec2-ami-tools.zip
wget http://s3.amazonaws.com/ec2-downloads/ec2-api-tools.zip
unzip ec2-api-tools.zip*

  • Place your X.509 keys in /mnt/work/.ec2

  • Note: the version numbers do change; please update to reflect this.

configure your environment
1
2
3
4
5
6
7
8
9
10
11
12

export EC2_BASE="/mnt/work"
export EC2_AMIHOME="${EC2_BASE}/ec2-ami-tools-1.4.0.5"
export EC2_HOME="${EC2_BASE}/ec2-api-tools-1.5.2.3"
export PATH=${EC2_AMIHOME}/bin:${EC2_HOME}/bin:${PATH}
export EC2_CERT="/mnt/work/.ec2/cert-example.pem"
export EC2_PRIVATE_KEY="/mnt/work/.ec2/pk-example.pem"
export EC2_SECRET_KEY="secret_key"
export EC2_ACCESS_KEY="access_key"
export SOURCE_AMI_NAME="example_ami_name"
export SOURCE_AMI_MANIFEST="${SOURCE_AMI_NAME}.manifest.xml"
export EC2_AMIS="${EC2_BASE}/amis"
export EBS_DEV="/dev/xvdp1"
download your s3 ami locally
1
2
3
4
5

ln -sf $EC2_AMIHOME/lib/ec2 $EC2_HOME/lib/ec2
temp_dir=$(mktemp -d)
mkdir -p ${EC2_AMIS}
cd ${EC2_AMIS}
ec2-download-bundle -b example_bucket -a ${EC2_ACCESS_KEY} -s ${EC2_SECRET_KEY} -k ${EC2_CERT} -m ${SOURCE_AMI_MANIFEST}
unbundle your s3 ami and mount it
1
2
3
4
5
6

ec2-unbundle -k ${EC2_PRIVATE_KEY} -m ${SOURCE_AMI_MANIFEST} -d ${temp_dir}
cd ${EC2_BASE}
export EC2_AMI_IMAGE="${temp_dir}/${SOURCE_AMI_NAME}"
export EC2_SOURCE_MOUNT=$(mktemp -d)
export EC2_DEST_MOUNT=$(mktemp -d)
mount -o loop ${EC2_AMI_IMAGE} ${EC2_SOURCE_MOUNT}
  • Please send suggestions on how to improve this to /dev/null.
partition out your ebs volume
1
2
3
4
5
6
7
8
9
10

fdisk /dev/xvdp
n
p
1
<enter>
<enter>
a
1
w
q
format the EBS volume and mount it. No filesystem checks desired. 
1
2
3
4

blockdev --rereadpt /dev/xvdp
mkfs -t ext4 -L rootfs ${EBS_DEV}
tune2fs -i 0 -c 0 ${EBS_DEV}
mount ${EBS_DEV} ${EC2_DEST_MOUNT}
copy your local S3 AMI to the EBS volume
1

rsync -aHAS ${EC2_SOURCE_MOUNT}/ ${EC2_DEST_MOUNT}/
chroot into your EBS root and download the linux kernel
1
2
3
4
5

cp /boot/config* ${EC2_DEST_MOUNT}
mount -o bind /dev ${EC2_DEST_MOUNT}/dev
chroot ${EC2_DEST_MOUNT}
cd /usr/src && curl http://www.kernel.org/pub/linux/kernel/v2.6/linux-2.6.39.4.tar.bz2 | tar jxf -
ln -s /usr/src/linux-2.6.39.4 linux
configure your kernel
1
2
3
4
5
6

cd /usr/src/linux
cp /config* .config
make bzImage -j16
make modules -j16
make modules_install
cp arch/x86/boot/bzImage /boot/vmlinuz-hvm

  • Configure /boot/grub/menu.lst similar to how you see provided by amazon; but replace the kernel and initrd with your own.

  • Use your favorite package manager to install grub. The following should work provided no ephemeral disks are attached.

install grub on the disk
1
2
3
4

grub
root (hd1,0)
setup (hd1)
quit
  • Exit out of your chroot and unmount everything
unmount your ebs root
1
2

umount ${EC2_DEST_MOUNT}/dev
umount ${EC2_DEST_MOUNT}
  • Now locally create a snapshot of your EBS root in case something goes wrong and you want to re-attach it; fix it; re-register.
snapshot your ebs-root
1
2

ec2-create-snapshot vol-03aff26e
SNAPSHOT  snap-9239e6f6   vol-03aff26e    pending 2012-01-07T23:37:08+0000        621517282122    8
stop your instance while it’s snapshotting
1
2

ec2-stop-instances i-de453ebc
INSTANCE  i-de453ebc  running stopping
  • Verify the instance is stopped
list your instance stopping
1
2
3
4
5
6

ec2-describe-instances i-de453ebc -H
Type  ReservationID   Owner   Groups  Platform
RESERVATION   r-f5743394  621517282122    default
INSTANCE  i-de453ebc  ami-0da96764            stopped ssh_key 0       cc1.4xlarge 2012-01-07T23:01:54+0000    us-east-1c              monitoring-disabled                 ebs                 hvm xen     sg-7baf4812 default
BLOCKDEVICE   /dev/sda1   vol-a9acf1c4    2012-01-07T23:39:34.000Z    
BLOCKDEVICE   /dev/sdp    vol-03aff26e    2012-01-07T23:39:34.000Z
detach the old root volume provided by amazon and your ebs root volume you made
1
2
3
4

ec2-detach-volume vol-a9acf1c4
ATTACHMENT    vol-a9acf1c4    i-de453ebc  /dev/sda1   detaching   2012-01-07T23:39:34+0000
ec2-detach-volume vol-03aff26e
ATTACHMENT    vol-03aff26e    i-de453ebc  /dev/sdp    detaching   2012-01-07T23:39:34+0000
attach volume
1
2

ec2-attach-volume vol-03aff26e -i i-de453ebc -d /dev/sda1
ATTACHMENT    vol-03aff26e    i-de453ebc  /dev/sda1   attaching   2012-01-07T23:42:38+0000
create your AMI
1
2

ec2-create-image i-de453ebc -n "fail"
IMAGE ami-ed9d4a84
  • Once it’s done preparing the AMI; boot it and hope for success. If you have a problem review the Console log and research. I suggest reviewing Amazon’s Kernel configuration and Ubuntu’s Kernel configuration for the HVM instance types if you have any problems; most of all good luck!
]]> <![CDATA[Learn to Respect the Problem]]> 2011-07-24T12:06:00-07:00 http://www.likens.us/blog/2011/07/24/learn-to-respect-the-problem There is a time in your life when you start learning to respect the problem. Of course you can fix a problem easily by doing x. But if you spend a little more time you can fix x,y and z and in a way that you feel proud of.

Yes I am likely twisting the definition for my own gains.

]]> <![CDATA[Getting Chef 0.8 Bootstrapped and Up in Screen]]> 2009-12-14T11:00:00-08:00 http://www.likens.us/blog/2009/12/14/getting-chef-08-bootstrapped-and-up-in-screen Simple post, just to detail if you want .8 and don’t want to use bootstrap or runit or init and you just want it working.

  • Get the base system up
1
2
3
4
5

sed -i 's/universe/multiverse universe/' /etc/apt/sources.list
apt-get update
env DEBIAN_FRONTEND=noninteractive apt-get upgrade -y
echo 'sun-java6-jdk shared/accepted-sun-dlj-v1-1 boolean true' | sudo debconf-set-selections
apt-get install -y ruby ruby1.8-dev libopenssl-ruby1.8 rdoc ri irb build-essential zlib1g-dev libxml2-dev wget ssl-cert git-core couchdb sun-java6-jdk sun-java6-jre
  • Setup gem to not annoy me
1

echo "gem: --no-rdoc --no-ri" >> ~/.gemrc
  • Now install rubygems,
1
2
3
4
5
6

cd /tmp
wget http://rubyforge.org/frs/download.php/60718/rubygems-1.3.5.tgz
tar zxf rubygems-1.3.5.tgz
cd rubygems-1.3.5
sudo ruby setup.rb --no-ri --no-rdoc
sudo ln -sfv /usr/bin/gem1.8 /usr/bin/gem
  • Install RabbitMQ
1
2

wget http://www.rabbitmq.com/releases/rabbitmq-server/v1.7.0/rabbitmq-server_1.7.0-1_all.deb
dpkg -i rabbitmq-server_1.7.0-1_all.deb
  • Make sure you have rake/etc installed
1

gem install rake cucumber rspec libxml-ruby jeweler thin unicorn merb-core merb-slices
  • now setup RabbitMQ so that chef can talk to it
1
2
3

rabbitmqctl add_vhost /chef
rabbitmqctl add_user chef testing
rabbitmqctl set_permissions -p /chef chef ".*" ".*" ".*"
  • Grab Chef .8 from master and start work
1

mkdir git;git clone git://github.com/opscode/chef.git
  • Install chef from master
1

rake install

  • You may catch a few missing dependencies, gem install them if I missed it (but I think this is right)

  • Now you need to install the chef-solr bins (the gemspec isn’t fixed to do this yet)

1

cd git/chef/chef-solr/bin;cp * /usr/bin
  • Grab mixlib-authentication from git
1

git clone git://github.com/opscode/mixlib-authentication.git
  • Install mixlib-authentication
1

cd mixlib-authentication && rake install
  • Grab mixlib-log from git
1

git clone git://github.com/opscode/mixlib-log
  • Install that also
1

cd mixlib-log && rake install

  • Now fire up screen and start it up

  • In not the perfect order, but.

  • Start chef-solr

1

chef-solr
  1. Start chef-solr-indexer
1

chef-solr-indexer
  1. Grab some configs that should work,
1
2
3

mkdir -p /etc/chef
cd /etc/chef;wget http://likens.us/server.rb  
cd /etc/chef;wget http://likens.us/client.rb
  1. Start chef-server
1

chef-server
  1. Now configure knife to work with the webui key and you should be solid.
1

knife configure

  1. Run chef-client over localhost and it should register and work hopefully

  2. Enjoy

  3. Thanks to Bryan Helmkamp for updates to this post.

]]> <![CDATA[Why you should never change API on a minor release]]> 2009-11-02T10:33:00-08:00 http://www.likens.us/blog/2009/11/02/why-you-should-never-change-api-on-a-minor-release In this case, my pain is Chef; as 0.7.0 came out we were rocking with metadata.rb, as I wasn’t paying attention and reading release documentation properly suddenly that doesn’t do squat, and everything is in metadata.json. I’m sure there was something in one of the release notes that I omitted, however it brings up a valid point that Opscode has broken many times (I am only mentioning these 2) in the .7 release cycle

They have repeatedly broken the basic idea of not making major changes in a minor release, such as this, or when they re-wrote the bootstrap system; and then neutered the old system so you had to use the old bootstrap tarball (which is the only reason I am not up in arms).

… However, if you want to do things like this, you should release a new major version and allow people to get used to the major changes. There are always growing pains with anything new, different, … however the past has taught us that you should not make stupid major changes to a structure without a major release.

Opscode has made this much smoother since the original post of this message and they have delivered a much better product then this post originally ranted about.

]]>