Managing Your AWS Infrastructure at Scale

You know how to manage one instance - can you manage a thousand?

Transcript

1. MANAGING YOUR AWS INFRA AT SCALE S U M M E R O F T E C H N O L O G Y 2 5 / 0 8 / 2 0 1 5 1
2. INTRODUCTION David Mat Cloud Architecture & DevOps Consultant at Archimiddle ✉️ david.mat@archimiddle.com @davidmat 2
3. INTRODUCTION • Instances x 1000 = Cost & complexity x 1000 ? • Processes & best practices • Tools AS-IS TO-BE 3
4. INTRODUCTION • First steps – Instances ~ Regular VMs in a DC • Manual • Diverges • Solution – Specify & manage the desired end state of your infrastructure independently of the EC2 instances & other running components 4
5. INTRODUCTION • Benefits of dynamic cloud environments 1. Software-defined infrastructure 2. Auto scaling & self healing 3. Fast environment provisioning 4. Reduce costs 5. Blue-green deployments 5
6. INTRODUCTION • Required capabilities 1. Automatically provisioned 2. Stateless 6
7. PROVISIONING NEW EC2 INSTANCES • New instances ~ external events – New instances/environments – Scaling events – Replace failed instances – New version of software stack • Difficult to predict ⇒ process needed: – Automated – Repeatable – Consistent 7
8. PROVISIONING NEW EC2 INSTANCES • Bootstrapping – “The process of automatically provisioning new instances & bringing them into service” • 2 most popular approaches: 1. Golden Images 2. Dynamic Configuration • Many possible combinations & hybrid models 8
9. GOLDEN IMAGE • Amazon Machine Image (AMI) – “Template that provides all of the information required to launch an Amazon EC2 instance” • Options – AWS-Provided – Community – Marketplace – Custom AMI 9
10. GOLDEN IMAGE • Advantages of custom AMIs – Increases speed – Reduces external dependencies – Removes the reliance on complex configuration scripts at launch time • Disadvantages of custom AMIs – Loss of agility – Complexity – Run-time configuration requirements 10
11. GOLDEN IMAGES • Managing AMI Builds: Automate! – Speed – Version control – Repeatable, auditable • Process 1. Launch the latest version of the AMI 2. Execute the automated configuration 3. Create a new AMI based on the running instance • Tools – Packer – Aminator 11
12. HYBRID MODELS • Decision Factors – Frequency of deployments – Reduction on external dependencies – Requirements to scale quickly 12
13. HYBRID MODELS 13
14. DYNAMIC CONFIGURATION • Many tools & techniques – Simple scripts → centralized configuration management tools • Scripting your own solution – User Data & Cloud-Init • Shell script or cloud-init directives • Limited to 16kb • Ideal for simple use cases – Leveraging EC2 metadata • http://169.254.169.254/latest/meta-data/<key> – Resource Tagging 14
15. DYNAMIC CONFIGURATION 15
16. DYNAMIC CONFIGURATION • Using configuration management tools – Push model • Master server needs info on nodes to manage • Load on master server • Need incoming ports on node opened – Pull model • Nodes scale up & down independently • Less load on master, more on node • Agent-based • No specific ports needed inbound 16
17. RESOURCES • Building Fault Tolerant Applications on AWS (Whitepaper) Jeff Barr, Attila Narin, and JineshVaria • Managing Your AWS Infrastructure At Scale (Whitepaper) Shaun Pearce and Steven Bryen • ARC402 – Deployment Automation (reInvent 2014) Chris Munns 17