Overnight soak test

import { Steps, Aside } from ‘@astrojs/starlight/components’;

Problem: Some failures only appear after hours under load — memory leaks, connection pool exhaustion, log file rotation failures, slow garbage-collection pressure. A 10-minute load test will never find these. A soak test holds moderate load for 8–24 hours and looks for drift in key metrics over time.

Test type: Soak / endurance test.

Prerequisites

• A MaxoPerf account with a test that has passed a standard load test at the target VU/RPS level.
• Awareness that the test will consume runner quota for the full duration — verify your plan limits before starting.

Soak tests are long. Make sure you are testing a staging environment with enough storage headroom, that logs are rotated, and that the target database has space for the writes your scenario generates across 8+ hours of continuous load.

Step by step in MaxoPerf

1. Write the soak scenario

execution:
  - scenario: api-soak
    concurrency: 50          # moderate — 50 % of your peak
    ramp-up: 5m
    hold-for: 8h             # the soak window
    throughput: 100          # optional: cap RPS for predictability

scenarios:
  api-soak:
    requests:
      - label: GET /api/health
        url: https://api.example.com/api/health
        method: GET

      - label: GET /api/products
        url: https://api.example.com/api/products
        method: GET
        headers:
          Authorization: Bearer ${API_TOKEN}

      - label: POST /api/sessions
        url: https://api.example.com/api/sessions
        method: POST
        headers:
          Content-Type: application/json
          Authorization: Bearer ${API_TOKEN}
        body: '{"user_id": "u-soak-01"}'

Keep soak test scenarios simple — the goal is sustained pressure, not scenario coverage. A handful of representative endpoints is enough.

2. Upload and configure the test

1. In the MaxoPerf console, open the test and upload the YAML under Files.
2. Bind API_TOKEN under Settings → Secrets.
3. Under Configuration, confirm the location and VU count.

3. Schedule the soak to start overnight

Rather than clicking Run now and staying up, schedule it:

1. Switch to the Schedules tab.
2. Click New schedule.
3. Enter a cron expression for tonight — for example 0 22 * * * (10 PM every night).
4. Pick your timezone.
5. Name it overnight-soak and save.

See Schedule recurring tests for the full scheduling workflow.

4. Review the results in the morning

When you open the run the next day:

1. Open the run and go to the Overview tab.
2. Zoom the time range to the full 8-hour window.
3. Look at p95 latency over time — a healthy system shows a flat line. An upward trend indicates drift (memory pressure, pool exhaustion).
4. Check error rate — any sustained error rate after the first few minutes of ramp-up is a finding.
5. Switch to the Runners tab to verify all runners stayed healthy throughout.

Before the soak, create a [reporting dashboard](/docs/academy/cookbook/build-a-reporting-dashboard/) with charts for p95 latency over time, RPS over time, and error rate over time. Apply it to the soak run in the morning for a clean one-glance view.

Verify

• Latency is flat or has a negligible trend over the full 8-hour window (< 5 % drift from the 1-hour mark to the 8-hour mark).
• Error rate stays near zero after ramp-up.
• RPS tracks the target steadily — no progressive drop that would indicate pool exhaustion starving new requests.
• The run status is Finished.

Variations

• 24-hour soak: extend hold-for: 24h for platforms where daily cycles (midnight jobs, log rotation, certificate renewal) are expected.
• Recurring weekly soak: leave the cron schedule running on a weekly cadence so every release is validated against the soak profile automatically.
• Parameterise with CSV data: supply a data entity with enough rows for the full 8-hour run so unique users cycle in — see CSV data-driven test.

Where to go next

• Soak / endurance test — how to read a soak result and the patterns that indicate common failure modes.
• Scheduled regression test — combine scheduling with a weekly baseline comparison.
• Build a reporting dashboard — set up the dashboard before the soak so the morning review is fast.
• Schedule recurring tests — the full scheduling how-to.