GitHub Actions test app (Node.js)

A simple Node.js app to test and try App Service's integration with GitHub Actions.

SRE Agent lab scenarios

This fork is also used as the workload for the Azure SRE Agent Level 200 labs. It adds structured request logging, opt-in fault-injection endpoints, and a traffic generator, so the agent has real failures to investigate. See docs/ARCHITECTURE.md, docs/RUNBOOK.md, and the infrastructure in infra/main.bicep.

Fault-injection endpoints

The endpoints return 404 unless the app setting CHAOS_ENABLED=true is set. Every fault is capped.

Endpoint	Effect	Limits
GET /api/chaos	Shows whether faults are enabled, plus current CPU and memory load	Always available
GET /api/slow?ms=3000	Delays the response	Max 10 s
GET /api/error	Throws an error: HTTP 500 plus a stack trace in the logs	—
GET /api/cpu?seconds=10	Busy-loops in the background	Max 30 s, at most 2 at once
GET /api/memory?mb=50	Holds memory until released	Max 200 MB in total
GET /api/memory/release	Frees the held memory	—

To turn faults on or off (no deployment slot, so this targets production):

az webapp config appsettings set -g rg-testapp-node -n testapp-node-2rqjmu --settings CHAOS_ENABLED=true
az webapp config appsettings set -g rg-testapp-node -n testapp-node-2rqjmu --settings CHAOS_ENABLED=false

Changing an app setting restarts the app, which also releases any held memory.

Generating traffic

bash scripts/traffic.sh https://testapp-node-2rqjmu.azurewebsites.net [normal|errors|slow|cpu|memory|mixed] [duration-seconds]

Logs

Each request is logged to stdout as one JSON line, with method, path, status, durationMs and requestId. Unhandled errors are logged with their stack trace and the same requestId. The x-request-id response header lets you match a client response to its log lines.