https://github.com/yakdriver/basic-ami-pipeline

Test AMI pipeline
https://github.com/yakdriver/basic-ami-pipeline

Last synced: 11 days ago
JSON representation

Test AMI pipeline

• Host: GitHub
• URL: https://github.com/yakdriver/basic-ami-pipeline
• Owner: YakDriver
• License: mpl-2.0
• Created: 2020-03-02T18:00:41.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-03-02T19:01:10.000Z (over 4 years ago)
• Last Synced: 2024-10-05T22:21:13.279Z (about 1 month ago)
• Size: 7.81 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# basic-ami-pipeline

Test AMI pipeline

## 1. Identify starting AMI (s-ami)

* Automatic lookup?

  - Pro: Convenient

  - Pro: Automatically up-to-date

  - Con: Increased chanced of manual intervention to fix issues with new OS releases

* Manual lookup?

  - Pro: Pipeline less likely to fail as time passes

  - Con: Falls out of date without manual intervenion depending on OS

  - Con: Inconvenient

* Potential solutions for automatic lookup

  - AWS CLI

  - Hashicorp Terraform

## 2. Launch and mod EC2 based on s-ami

* Credentials

  - Okta/STS

  - Instance role

  - Hardcoded

* Platform for launch and mod

  - AWS CodeBuild

  - Other CI (Travis, Azure Pipelines) - potential for timeout

* Tools for launch and mod

  - Hashicorp Terraform

  - AWS CloudFormation

* Modifications

  - Launch EC2 instance

  - Install patches

  - Install IT agents

  - Apply OS/disk/etc. configurations

## 3. Create modified AMI (m-ami)

* Stop EC2

* Create AMI

  - Hashicorp Packer

  - Hashicorp Terraform

  - AWS CLI

## 4. Test m-ami

* Provide basic testing only if time runs short

* Launch EC2 instance from m-ami

* Run tests on instance

## 5. Publish m-ami

* Extract info used in release (e.g., rpm list) to document

* Copy AMI to supported regions

* Share AMI to supported accounts

* Publish AMI release

## 6. Cleanup

* Ensure AWS resources used in pipelines are deleted

* CloudBuild, Terraform, Packer should handle

* Consider error cases where orphan resources might remain