Immutable Infrastructure with Packer

Dirk and I talked about our usage of HashiCorp's Packer at the AWS Meetup in Cologne (2015-09-25). We outline why we favor the approach of handling servers "like cattle" and immutable over conventional configuration management methods.

Transcript

1. Immutable Infrastructure with Packer
2. Sebastian StormForger.com @tisba Dirk Freelancer @railsbros_dirk Who?
3. Challenges • Increasing complexity • Multiple target environments • Multi/hybrid cloud • Evolving infrastructure • Communication & Documentation
4. Opportunities • Infrastructure (not just servers) on demand • Flexible APIs for all infrastructure components • Affordable but powerful hardware and services • Virtualization
5. Today: Handling Servers
6. Solution SCRIPTING SCRIPTING
9. Circle of (Server) Life • initial (hardware) provisioning & basic setup • continuous updates, applying changes • decommissioning
10. Change & Configuration Management • Manual (aka "Turnschuhadministration") • Automatic, convergent (chef, puppet, …)
11. These tools and approaches don't have the modern data center tooling & infrastructure in mind!
12. Pets vs Cattle
14. Pets vs Cattle • Pets have names, you care about them • Cattle is… "replaceable", you have many of them • Most servers should be cattle, not pets
15. "Immutable Infrastructure?"
16. Immutable Infrastructure • Idea: Servers do not change! • If you need to change something, build a new server and terminate the existing one http://martinfowler.com/bliki/ImmutableServer.html
17. Servers do not change!
18. What do we need to
 treat Servers as Cattle immutable?
19. Immutable Servers • Treat entire servers as build artifacts • Always build them from a known start state • Deploy servers, not applications • Inject configuration on deploy
20. Solutions? Tools?
21. – @bascht “For any (dev)ops related problem HashiCorp most likely already has a tool.”
23. – packer.io "Packer is a tool for creating machine and container images for multiple platforms from a single source configuration." Packer
24. Packer • Packer helps to create pipelines for building servers as artifacts for multiple target platforms • It also integrates with multiple provisioning tools
25. Builders Artifacts
26. Provisioners Builders Artifacts
27. Provisioners Builders Post-Processors Artifacts
28. Dev Production Packer Config Provisioners Builders Amazon EC2 VMWare Post-Processors Ansible Ansible Ansible Atlas Vagrant Atlas Artifacts AMI Vagrant Box
29. Dev Production Builders Artifacts Builders Builders
30. Builders Amazon EC2 (AMI) DigitalOcean Docker Google Compute Engine Null OpenStack Parallels QEMU VirtualBox VMware Custom *currently
31. Provisioners Remote Shell Local Shell File Uploads PowerShell Windows Shell Windows Restart Ansible Chef Client Chef Solo Puppet Masterless Puppet Server Salt Custom *currently
32. Post-Processors Artifice Atlas compress docker-import docker-push docker-save docker-tag Vagrant Vagrant Cloud vSphere *currently
33. { "variables": { "aws_access_key": "", "aws_secret_key": "" }, "builders": [{ "type": "amazon-ebs", "access_key": "{{user `aws_access_key`}}", "secret_key": "{{user `aws_secret_key`}}", "region": "us-east-1", "source_ami": "ami-de0d9eb7", "instance_type": "t1.micro", "ssh_username": "ubuntu", "ami_name": "packer-example {{timestamp}}" }] }
35. { "variables": ["..."], "builders": ["..."], "provisioners": [{ "type": "shell", "inline": [ "sleep 30", "sudo apt-get update", "sudo apt-get install -y redis-server" ] }] }
36. Q&A Sebastian Cohnen stormforger.com @tisba Dirk Breuer tfcl.de @railsbros_dirk
37. NOT SURE TO USE PACKER NOT SURE TO USE PACKER OR JUST GO WITH DOCKER OR JUST GO WITH DOCKER