Creating an AMI on EC2
Notes About Creating a New EC2 AMI
Back to parent: GotCloud
The following are notes taken when creating the Amazon Machine Instance used for the CSG pipeline process. These notes assume you have already created an EC2 account and have the certificates and keys set up properly.
Launch an instance
Login to https://console.aws.amazon.com/ec2 # EC2 Management Console
Pay attention to the region you are using, at least for now it seems any StarCluster activity must be in us-east-1. Launch a new instance which we will use to set up the software and ultimately save it as an AMI.
EC2 DashBoard -> Launch Instance
Class Wizard
Ubuntu Server 12.04.1 LTS 64 bit
Instance type -> Micro, EC2, no preference # Memory size does not matter
Advanced Instance Options (take defaults)
Storage Device Configuration -> Edit
Change volume to 30G -> Save -> Continue # Storage size does not matter
Key Name = GotCloud 1.06a
Create Key/Pair if you need to, Name the PEM and save the pem file for access by ssh
Choose a Security Group (take default)
Launch
No need to Create Status Check Alarms
No need to Create EBS Volumes
Set Up Swap Space
Issue the command swapon -s to see if there is swap space. If there is only a header line, you need to add a swap file like this:
df -h # Be sure there's enough space, decide on swap size
# Create a file /swap to use (assuming / is large enough)
sudo bash # Run these commands as root
swap=/swap
dd if=/dev/zero of=$swap bs=1073741824 count=15 # 15GB swap
chown root:root $swap
mkswap $swap
chmod 0600 $swap
swapon $swap
echo "$swap none swap sw 0 0" >> /etc/fstab
swapon -s # Should show the swap device
Install the Software
(1) There are a number of additional Debian packages that you may well need, so we make sure they are all installed.
sudo apt-get update
sudo apt-get upgrade # Apply maintenance
sudo apt-get install java-common default-jre make libssl0.9.8
sudo apt-get install libnet-amazon-ec2-perl s3cmd
sudo apt-get install make g++ libcurl4-openssl-dev libssl-dev libxml2-dev libfuse-dev
(2) S3fs allows one to access S3 storage as a conventional file system. This can be quite handy, if it is set up properly. Our recent experience is that the 1000 Genomes data is has many files with incorrect permissions. Still if you're lucky, your data will be useful. Install the software like this:
mkdir -p ~/src
cd ~/src
wget http://s3fs.googlecode.com/files/s3fs-1.68.tar.gz
tar xzvf s3fs-1.68.tar.gz
cd s3fs*
./configure
sudo make install
(3) Follow the instructions to install the GotCloud Debian packages Run the tests to be sure everything is OK.
Configure the Host to be Usable
It is useful to configure /etc/rc.local to do most things you need at boot time. There are many other ways to do this, but here's one simple way - create the file /etc/rc.local (as root). The following example sets up access details for s3cmd and s3fs (use your own credentials).
ubuntu@ip-10-254-60-210:~$ sudo more /etc/rc.local
#!/bin/sh
USER=ubuntu
THOUSANDG=/mnt/1000g
FILES3=passwd-s3fs
S3ERR=/tmp/s3fs.err
# Set up for GotCloud Assumes /dev/xvdf has reference files for GotCloud
mkdir -p /gotcloud
mount /dev/xvdf /gotcloud
if [ ! -d /gotcloud/gotcloud.ref ]; then
echo "#######################################################"
echo "# GotCloud is not set up on /gotcloud"
echo "#######################################################"
fi
# Setup 1000g access by s3fs
usermod -aG fuse $USER
echo 'AKIAIW5TQEUWZ3YCZF2Q:ft1eJa3WGzNE8iitNlbA08x/G8iMqkMIkJjFCIGf' > /etc/$FILES3
chown root.root /etc/$FILES3
chmod 640 /etc/$FILES3
mkdir -p $THOUSANDG
chown $USER.$USER $THOUSANDG
# It is tempting to use caching with -o use_cache=/tmp 1000genomes
# But s3fs cache is exceedingly dumb and does not use a least recently used
# mechanism -- which will guarantee your root volume will fill up
/usr/local/bin/s3fs -o allow_other 1000genomes $THOUSANDG > $S3ERR 2>&1
if [ ! -r $THOUSANDG/alignment.index ]; then
echo "#######################################################" >> $S3ERR
echo "# 1000genomes is not set up on $THOUSANDG" >> $S3ERR
echo "#######################################################" >> $S3ERR
fi
df -h
exit 0
Create the AMI
Once your instance is all ready with the files you want, swap space etc, then create the AMI. In your browser at the EC2 Management Console do the following:
Create Image
Image Name csg-biopipe_instance
Image Description: Image for CSG Biopipe instance
Volume Size: 30GB
Take defaults otherwise
This will take several minutes to complete. In the EC2 Dashboard, you can monitor the progress. When it is done, you'll see a new AMI under the list of AMIs.
Your new AMI should look pretty much like this:
AMI: Ubuntu Cloud Guest AMI ID ami-3d4ff254 (x86_64)
Name: Ubuntu Server 12.04.1 LTS
Description: Ubuntu Server 12.04.1 LTS with support available from Canonical (http://www.ubuntu.com/cloud/services).
Number of Instances: 1
Availability Zone: No Preference
Instance Type: Micro (t1.micro)
Instance Class: On Demand Edit Instance Details
EBS-Optimized: No
Monitoring: Disabled Termination Protection: Disabled
Tenancy: Default
Kernel ID: Use Default Shutdown Behavior: Stop
RAM Disk ID: Use Default
Network Interfaces:
Secondary IP Addresses:
User Data:
IAM Role: Edit Advanced Details
Key Pair Name: CSG Edit Key Pair
Security Group(s): sg-a098e9c8 Edit Firewall
Test the new AMI
Launch a new AMI instance and check that files are in the correct places. In the EC2 Management Console do:
EC2 DashBoard -> AMIs -> Select CSG instance -> Launch Instance
Launch Instances (take defaults)
Advanced Instance Options (take defaults)
Storage Device Configuration -> Edit
Change volume to 30G or whatever -> Continue # Defaults are OK
Instance Details
Key Name = test of instance
Create Key/Pair if you need to, most likely you can use one you have created
Choose a Security Group -> sg-a098e9c8 - quick-start-1
Review -> Launch
