Skip to content

walkthrough.md

Latest commit

 

History

History
107 lines (77 loc) · 2.5 KB

walkthrough.md

File metadata and controls

107 lines (77 loc) · 2.5 KB
Raw

Walkthrough

This walkthrough builds a realistic service, runs preflight checks, executes it in a virtualized microVM sandbox, and exposes it through the HTTP API.

1) Create Service

ignite init data-processor
cd data-processor

2) Implement Logic

Edit index.js:

const input = process.env.IGNITE_INPUT ? JSON.parse(process.env.IGNITE_INPUT) : {"data":[],"operation":"sum"};

function calculate(data, operation) {
  if (data.length === 0) return 0;

  switch (operation) {
    case 'sum':
      return data.reduce((a, b) => a + b, 0);
    case 'average':
      return data.reduce((a, b) => a + b, 0) / data.length;
    case 'max':
      return Math.max(...data);
    case 'min':
      return Math.min(...data);
  }
}

const output = {
  result: calculate(input.data, input.operation),
  operation: input.operation,
  count: input.data.length,
  timestamp: new Date().toISOString(),
};

console.log(JSON.stringify(output));

3) Configure Service

Edit service.yaml:

service:
  name: data-processor
  runtime: bun@1.3
  entry: index.js
  memoryMb: 128
  cpuLimit: 1
  timeoutMs: 10000
  env:
    NODE_ENV: production

4) Run Preflight Checks

Run preflight checks to validate configurations and memory limits:

ignite preflight .

If checks return FAIL, optimize the configurations or dependencies before executing the service.

5) Execute Sandbox VM

Run the service inside the hardware-isolated microVM:

ignite run . --input '{"data":[1,2,3,4],"operation":"sum"}'

To see trace timings for the formatting of the virtual block device images and KVM launch steps, add the --verbose flag:

ignite run . --input '{"data":[1,2,3,4],"operation":"sum"}' --verbose

6) Expose Service via HTTP REST API

Run the server pointing to the folder containing the services:

cd ..
ignite serve --services ./services --port 3000

Now, invoke execution from a client using curl:

curl -X POST http://localhost:3000/services/data-processor/execute \
  -H 'Content-Type: application/json' \
  -d '{"input":{"data":[5,10,15],"operation":"max"}}'

7) Operational Recommendations

  • Pin Versions: Always keep runtime versions pinned in service.yaml (e.g., bun@1.3, node@20) to guarantee deterministic block mounting on the host.
  • Resource Constraints: Set conservative memoryMb and timeoutMs for each service to bound untrusted execution loops.
  • Single Purpose: Keep services modular; decouple large tasks into separate services.