Automating Drudgery

Overview

Built and deployed a system to automate the process of provisioning cloud resources based on customer requests.

    graph LR
    U[User Request] ---- System
    Cd[Config Data] --- System
    R[Packages] --- System
    System --- AWS
    System --- GCP
    System --- Azure
    AWS --- Agent
    GCP --- Agent
    Azure --- Agent

    

Processes

This was a full stack system. Not only did this system support connecting to a customer’s private cloud, but would spin up servers, included a concept of bootstrapping servers, running and installing software of multiple variations.

• package deployment (shipit -> depo)
• package and release tracking
• network creation and deployment
• multi-cloud
• local daemon to pull updates (depot-agent)
• local cli to automate remote http actions
• fault tolerent
• instance abstraction
• bootstrapping
• configuration data (by environment)
• cloud
  • load balancer management
  • white labeled DNS

Releasing new Code

    graph LR
    Code --> Package --- Registry
    Code --> Metadata --- Registry

    

• software packaging and a repository (in-house format)

Configuration Data

    graph LR
    User --> Config --- ConfigStore

    

Alternatives would have been etcd, counsul, cloud native options like Azure Key Vault, AWS SSM Parameter Store. We wanted something that was multi-cloud, and wouldn’t require a cloud specific security credential.