Code Deploy

Automatic deployment:

• what: code, lambda functions, web files, configurations, executables, packages, scripts, etc.
• to: EC2 instances, on-premises instances (through agent), Lambda, EC2 services
• from: S3, GitHub, Bitbucket
• how:
  • in-place: only for EC2/on-premises each instance, stop, install, start & validate (can use load balancer to de-register / register instance)
  • blue/green: basically first provide replacement instance/environment, install, optionally wait for testing and validation, register replacement instances, remove original instances
    • benefit: simple, no installation, replacement & re-routing; original instances (if not terminated) can be kept for a while to roll-back; avoid problems related to long-running instances (overflown logs?); allow configure how traffic is shifted (canary);
    • blue/green integration with Cloud Formation: check document for more
• benefit: automated, managed, avoid downtime (various strategies), handle complexities (stop, roll back), scalable, center control with integrated with various services, concurrent

Application: name identify the application, combination of revision, deployment configuration & deployment group

Compute platform:

• EC2 / On-premise
• Lambda
• ECS

Deployment configuration: rules determine success/failure conditions

• minimum number of healthy hosts

Blue/green configuration types:

• Canary: first a % of traffic to updated lambda version, then after an interval, shift remaining traffic
• Linear: a gradual steady percentage of traffic is shifted
• All-at-once

(EC2) Deployment Group: a set of individual tagged instances or EC2 instances in EC2 auto scalling groups (or both)

Deployment Type: in-place / blue-green

(EC2) IAM instance profile: an IAM role to attach to EC2 instances for permission to access S3 bucket or GitHub (source)

Revision: different for different environment, but basically an id, configuration & content

Service Role: the role CodeDeploy assume to perform deployments

Target Revision