DEV Community

George-Adaba
George-Adaba

Posted on

SwiftDeploy: Building a Self-Configuring DevOps Engine with Observability, Policy Enforcement & Auditing

#automation #devops #infrastructure #tutorial

Introduction

This is not just a project—it’s a mini DevOps platform.
In this guide, you will build SwiftDeploy, a system that:

  • Generates its own infrastructure from a single file
  • Monitors itself using real-time metrics
  • Enforces deployment safety using policy-as-code
  • Logs and audits every decision

By the end, you will be able to replicate the entire system locally from scratch.

1. The Design: A Tool That Writes Its Own Infrastructure

Core Idea

Everything is driven by manifest.yaml
This is your single source of truth.

Instead of manually writing:

docker-compose.yml
nginx.conf

Your CLI (SwiftDeploy) generates them.

2. Prerequisites

Install:

  • Docker
  • Docker Compose
  • Git
  • Python or Go

3. Project Structure

swiftdeploy/
├── manifest.yaml
├── swiftdeploy
├── app/
├── templates/
├── policies/
├── history.jsonl
├── README.md

4. Example manifest.yaml 

services:
  image: swift-deploy-1-node:latest
  port: 3000

nginx:
  image: nginx:latest
  port: 8080

network:
  name: swiftdeploy-net
  driver_type: bridge
Enter fullscreen mode Exit fullscreen mode

This file controls everything.

5. Setup & Run (Step-by-Step)

  • git clone https://github.com/YOUR_USERNAME/swiftdeploy.git
  • cd swiftdeploy
  • swiftdeploy init
  • swiftdeploy validate
  • swiftdeploy deploy

Expected Output (Deploy)

  • Manifest valid
  • Docker image found
  • Nginx config valid
  • Services starting...
  • Health check passed
  • Deployment successful

6. Observability: Metrics (/metrics)

Your API exposes metrics in Prometheus format.
Access Metrics
http://localhost:8080/metrics

Example Metrics Output
http_requests_total{method="GET",path="/",status="200"} 120
http_requests_total{method="GET",path="/",status="500"} 5

http_request_duration_seconds_bucket{le="0.5"} 100
http_request_duration_seconds_bucket{le="1"} 110
http_request_duration_seconds_bucket{le="+Inf"} 125

app_uptime_seconds 360
app_mode 1
chaos_active 2

Metrics Interpretation
Total requests = 125
Errors = 5
Error rate = 5 / 125 = 4%

This feeds directly into policy decisions.

7. Policy Enforcement (OPA)

SwiftDeploy uses Open Policy Agent.

Policy Flow
CLI → Collect Data → Send to OPA → Receive Decision → Act

Example Input to OPA (Pre-Promote)
{
"error_rate": 0.04,
"p99_latency": 600,
"mode": "canary"
}

Example OPA Response
{
"allow": false,
"reason": "Error rate exceeds 1% threshold"
}

Policies Implemented

Infrastructure Policy
Block deploy if:
Disk < 10GB
CPU > 2.0

Canary Safety Policy
Block promote if:
Error rate > 1%
P99 latency > 500ms

8. Chaos Testing (Failure Simulation)

Trigger Chaos

Slow Mode

POST /chaos
{
  "mode": "slow",
  "duration": 3
}
Enter fullscreen mode Exit fullscreen mode

Error Mode

POST /chaos
{
  "mode": "error",
  "rate": 0.5
}
Enter fullscreen mode Exit fullscreen mode

Reproduce Failure Scenario

  1. Deploy system
  2. Trigger chaos
  3. Run swiftdeploy status
  4. Run swiftdeploy promote

Expected Result
DENIED: P99 latency exceeds 500ms

9. Live Dashboard (swiftdeploy status)

swiftdeploy status

Example Output
Mode: canary
Requests/sec: 15
P99 latency: 620ms
Error rate: 3%

Policy Status:

  • Infrastructure: PASS
  • Canary Safety: FAIL (latency too high)

10. Logging (history.jsonl)

Each snapshot is stored as:

{
"timestamp": "2026-05-06T12:00:00Z",
"mode": "canary",
"req_per_sec": 15,
"p99_latency": 620,
"error_rate": 0.03,
"policy": "FAIL"
}

11. Audit Report

Generate:
swiftdeploy audit

Example Output (audit_report.md)

## Timeline
12:00 - Mode switched to canary
12:02 - Chaos injected
Enter fullscreen mode Exit fullscreen mode 
## Violations
12:03 - Error rate exceeded threshold
12:04 - Promotion denied
Enter fullscreen mode Exit fullscreen mode

12. Failure Testing

Test Disk Failure
Fill disk → run deploy
DENIED: Disk space below 10GB

Test Error Rate Failure
Inject chaos → run promote
DENIED: Error rate exceeds threshold

13. Security Requirement

OPA:
Must NOT be exposed via NGINX
Only accessible internally

14. Lessons Learned

1. Metrics Design is Critical
Bad metrics = wrong decisions

2. Policy Separation Improves Safety
OPA removes logic from the CLI.

3. Observability Enables Automation
No metrics → no intelligence

4. Failure Handling Matters
The system must never crash

5. Build Incrementally
Each layer depends on the previous

System Architecture

manifest.yaml
↓

swiftdeploy CLI
↓

Generated configs (Docker + Nginx)
↓

Containers (App + Nginx + OPA)
↓

Metrics → CLI → OPA → Decision
↓
history.jsonl → audit_report.md

Conclusion

SwiftDeploy is now a system that

  • Writes its own infrastructure
  • Observes its own behaviour.
  • Enforces its own safety
  • Records its own history

Final Thought

If your system can:

  • See itself
  • Evaluate itself
  • Protect itself

Then you’re not just deploying apps…
You’re building intelligent, resilient systems.

Top comments (0)