CloudBioLinux is a build and deployment system which installs an easily customizable selection of bioinformatics and machine learning libraries on a linux container, bare virtual machine (VM) image, freshly installed PC, or in the cloud. CloudBioLinux is a curated and community developed set of instructions for tools provided by operating system packages (debs and RPMs), external packaging efforts (homebrew-science) and language specific library installers (Python, R, Perl and Ruby).
CloudBioLinux included software packages are fully customizable. In addition to the default configuration, we support custom configuration builds through flavors. Flavors support overriding default package installations, making it simple to create derived installs for specific purposes.
CloudBioLinux is a single install route for Docker containers ,desktop VMs such as VirtualBox, cloud providers such as Amazon EC2 or desktop machines. This works equally well for other virtual machines and private cloud environments, including XEN, Linux KVM, Eucalyptus and Openstack.
See the 'Getting Started with CloudBioLinux' guide on the CloudBioLinux website for a detailed description. The short version for users familiar with Amazon is:
- Login to the Amazon EC2 console.
- Click Launch Instance, and choose the latest CloudBioLinux AMI from the website in the community AMI section (search for 'CloudBioLinux').
- After launching the instance, find the host details of your running instance from the Instances section.
- Connect to your machine via ssh or VNC (using the Amazon PEM keys)
The install process for CloudBioLinux is fully automated through a Fabric build file written in Python. Everything is fully configurable through plain text YAML configuration files, and custom build targets allow installation of a subset of the total available packages.
Retrieve the CloudBioLinux code base and install fabric:
pip install fabric git clone git://github.com/chapmanb/cloudbiolinux.git cd cloudbiolinux
The basic usage specifies the hostname of a machine accessible via ssh or the local machine:
fab -f fabfile.py -H localhost install_biolinux
Fabric contains some other useful commandline arguments for customizing this to your environments:
-c your_fabricrc.txt-- Specify the path to a fabricrc configuration files. This allows customization of install directories and other server specific details. See the defaultconfig/fabricrc.txtfor a full list of options.-u username-- The username on a remote machine, overriding the default of your current username.
In most cases you want to customize a specific set of packages, or install into an isolated directory without root access, using flavors:
fab -f fabfile.py -H localhost install_biolinux:flavor=my_flavor
my_flavor can be the name of an existing flavor in
contrib/flavor or the path to a directory with customization
information. The files in your flavor directory replace those in the
standard config directory, allowing replacement of any of the
configuration files like main.yaml with customized copies.
If you desire even more control, flavors allow custom python hooks. See
doc/hacking.md for more details.
The best place to get started is the demo flavor
included with CloudBioLinux. This installs a small number of common packages
into an isolated directory (~/tmp/cbl_demo by default), without root access.
Run the example with:
fab -f fabfile.py -H localhost install_biolinux:flavor=demo
You can substitute install_biolinux with more specific targets to
only build portions of CloudBioLinux:
install_biolinux:packages-- Install all of the defined system packages.install_biolinux:libraries-- Install all libraries for various programming languages.install_biolinux:brew-- Install homebrew packages only.install_libraries:language-- Install libraries for a specific language.install_biolinux:custom-- Install all custom programs.install_brew:a_package_name-- Install a specific brew package.install_custom:a_package_name-- Install a specific custom program.
Homebrew and Linuxbrew provide a Ruby-based environment for installing packages on MacOSX and Linux. The active homebrew-science packaging community maintains a number of common scientific tools. We also maintain a homebrew-cbl repository with tools not yet integrated into homebrew-science.
CloudBioLinux manages installation of the Linuxbrew or Homebrew framework and
pulls in the homebrew/science and chapmanb/cbl taps, as well as
injecting your current compilers into the homebrew build scripts. To install a
supported package
using CloudBioLinux:
fab -f fabfile.py -H localhost install_custom:bedtools
The custom directory contains installation instructions for programs that are not available from standard package repositories, written in Python using the Fabric remote deployment tool. To install individual custom packages:
fab -f fabfile.py -H localhost install_custom:your_package_name
We prefer using the Homebrew framework for new packages over writing custom packages.
We manage a repository of useful public biological data on an Amazon S3 bucket. Currently this includes whole genomes pre-indexed for a number of popular aligners. Downloading and installing these saves a ton of time over running the indexing steps yourself, and eases running next-generation analyses on cloud machines.
A Fabric build script is provided to install this data on your local
machine. A biodata configuration file in YAML
format,
config/biodata.yaml, specifies the genomes of interest and the
aligner indexes to use. The config/fabricrc.txt file specifies
details about the system and where to install the data.
The basic commandline is:
fab -f data_fabfile.py -H your_machine install_data_s3
and you can pass in custom biodata and fabricrc files with:
fab -f data_fabfile.py -H your_machine -c your_fabricrc.txt install_data_s3:your_biodata.yaml
In addition to downloading and preparing the data, the script will integrate these files with a Galaxy instance by updating appropriate Galaxy configuration files. This makes it useful for installing data to a local or cloud-based Galaxy server.
Not all of the genomes are hosted on the S3 bucket, but are still supported. If your genome fails to install with install_data_s3, you might be able to download the genome from from Ensembl, etc and prepare it:
fab -f data_fabfile.py -H your_machine -c your_fabricrc.txt install_data:your_biodata.yaml
Docker provides lightweight local containers for Linux machines, allowing isolation without the associated overhead of full virtual machines. Include any of the standard CloudBioLinux commands inside a Dockerfile to use CloudBioLinux to build up the set of tools on your instance. See the Dockerfile examples for information how to write Dockerfiles.
To use a pre-built Docker image made with CloudBioLinux infrastructure, using this bcbio-nextgen Dockerfile, you can import the bcbio-nextgen container into your local docker environment:
docker import https://s3.amazonaws.com/bcbio_nextgen/bcbio-nextgen-docker-image.gz chapmanb/bcbio-nextgen-cbl
A bare Linux image launched in Amazon EC2 is configured from another machine, i.e. your local desktop, using ssh and cloudbiolinux. See the Installation section for installing CloudBioLinux with fabric.
Any cloudbiolinux distribution can be used, including Ubuntu, Debian Linux and CentOS. We recommend using m1.medium or better instance for building a CloudBioLinux image from scratch, due to resource usage while compiling software.
- Go to the cloudbiolinux source and edit the
config/fabricrc.txt, to match the system you plan to install on. Specifically,distributionanddist_nameparameters specify details about the type of target. - Start an Amazon EC2 base instance and retrieve it's DNS hostname:
From your local machine, have CloudBioLinux install your Amazon instance:
fab -f fabfile.py -H hostname -u username -i private_key_file install_biolinux
When finished, use the Amazon console to create an AMI. Thereafter make it public so it can be used by others.
Vagrant allows easy deploying and connecting to VirtualBox images. The setup is ideal for runnig CloudBioLinux on a desktop computer. Install VirtualBox and vagrant. Then add a pre-built CloudLinux VirtualBox images and start it up:
vagrant box add biolinux_$VERSION https://s3.amazonaws.com/cloudbiolinux/biolinux_$VERSION.box mkdir tmp/biolinux cd tmp/biolinux vagrant init biolinux_version
(note with vagrant you need disk space - at least 3x the image size). The created ./Vagrantfile can be edited to get a full GUI, extra RAM, and a local IP address. Start and log into the image with:
vagrant up vagrant ssh sudo bash
Through Vagrant additional facilities are available, such as a shared network drive. It is also possible to tweak the image (e.g. RAM/CPU settings, and getting the all important guest additions) by firing up virtualbox itself. For more information, see the documentation on the Vagrant website.
See the VirtualBox and Vagrant documentation for details on creating a local virtual machine from scratch with CloudBioLinux.
As long as there is an 'ssh' entry to an running VM, CloudBioLinux can install itself.
For more on private Cloud and CloudBioLinux see ./doc/private_cloud.md.
This provides a quick cheat sheet of commands for getting up and running on EC2 using Amazon's command line tools.
The first time using EC2, you'll need to install the toolkit and credentials for connecting on your local machine, following the getting started guide.
Login to your Amazon EC2 account and
go to Security Credentials/X.509. Create a new certificate and download
the public cert-*.pem and private pk-*.pem files. Put these in
~.ec2.
Install the ec2 api tools, which require java.
Set up .zshrc/.bashrc:
export EC2_PRIVATE_KEY=~/.ec2/pk-UBH43XTAWVNQMIZRAV3RP5IIBAPBIFVP.pem export EC2_CERT=~/.ec2/cert-UBH43XTAWVNQMIZRAV3RP5IIBAPBIFVP.pem export AWS_ACCESS_KEY_ID=<your access key> export AWS_SECRET_ACCESS_KEY=<your secret access key>
To test, you should be able to run the command:
% ec2-describe-regions
Now generate a privatekey for logging in:
% ec2-add-keypair yourmachine-keypair
This will produce an RSA private key. You should copy and paste this to your .ec2 directory for future use:
% vim ~/.ec2/id-yourmachine.keypair % chmod 600 ~/.ec2/id-yourmachine.keypair
Allow ssh and web access to your instances:
% ec2-authorize default -p 22 % ec2-authorize default -p 80
Each time you'd like to use EC2, you need to create a remote instance to work with; the AWS console is useful for managing this process.
When building from scratch with Alestic images, you will need to increase the size of the root filesystem to fit all of the CloudBioLinux data and libraries. This is done by starting the instance from the commandline with:
% ec2-run-instances ami-1aad5273 -k kunkel-keypair -t m1.large
-b /dev/sda1=:20
% ec2-describe-instances i-0ca39764
On Ubuntu 10.04, you then need to ssh into the instance and resize the filesystem with:
% sudo resize2fs /dev/sda1
On 11.04 the resize happens automatically and this is not required.
BioLinux comes with an integration testing frame work - currently based on Vagrant. Try:
cd test ./testing_vagrant --help
Target VMs can be listed with
./testing_vagrant --list
Build a minimal VM
./testing_vagrant Minimal
Additional documentation can be found in the ./doc directory in the BioLinux source tree.
The code is freely available under the MIT license.