Easy Deployment of Tools with Complex Dependencies using Docker and Galaxy on the Cloud

+ =

Simon Belluzzo - Ira Cooke

Galaxy for the Uninitiated

Tool Dependency Asteroid Fields

Containers & Docker

Containers

• Other than the kernel, containers are isolated from each other and the host
• Sit in between chroot and full virtualization

And Docker..?

• Convenience wrapper around container management
• Mixes in Git/DVCS ideas in dealing with image changes/layers
• Docker Hub provides hosting and automated building
• For us: Enhances accessibility and reproducibility

Docker is not the only player, expect competition

Project Aims

1. Create a ProtK Docker image
2. Adapt Galaxy wrappers to use the image
3. Test deployment and execution on standalone and cloud Galaxy instances (for use with GVL)

Docker Image Building

Galaxy Integration

GVL/Galaxy on the Cloud Integration

• Dockerised tools run fine on either master or worker nodes (with the usual scaling abilities)
• Not compatible with the current GVL (new GVL version will be compatible)
• Don't have to wait; You can use our custom Galaxy on the Cloud image & bucket (NeCTAR only currently)
• Launchable by GVL Launcher (with custom settings) or by CLI tool
• Quickstart Guide available

Advantages

• Transparent to end-user
• Simple to deploy & update
• Aids reproducibility
• Isolates tools and their dependencies from each other and the host operating system

Disadvantages

• Some Docker bugs (transient, and fixed in new versions)
• Duplication of dependencies & large image sizes
• Nascent tech, patterns and best practices still forming
• Docker, Inc uncertainty
• Security (at least in Galaxy's use of Docker)

Further Work

• Inclusion of a Docker registry mirror
• Enhanced Galaxy-Docker integration
  • Allowing use of multiple containers in a single wrapper
  • Container sharing between tools
• Docker image and container management options
• Better image build process (Dockerfile or otherwise)
• Wine in Docker for Windows-based tools