#3750·artillery

Final report totals do not match worker logs in distributed Playwright tests on Fargate

Author: rom4ukCreated Jun 25, 2026Updated Jul 14, 2026

Version info:

"devDependencies": {
    "@playwright/test": "1.54.1",
    "artillery": "2.0.32",
    "artillery-plugin-ensure": "1.26.0",
    "playwright": "1.54.1",
    "tsx": "^4.21.0",
  }

Running this command:

bash
npx artillery run-fargate --count 99 --region us-east-2 --environment aws --cpu 8 --memory 16 test.yml

Scenario configuration

environments:
  aws:
    phases:
      - name: warm-up
        duration: 30s
        arrivalCount: 37
        maxVusers: 80
      - name: steady_load
        duration: 110s
        arrivalCount: 88
        maxVusers: 100
      - name: peak
        duration: 210s
        arrivalCount: 91
        maxVusers: 100
      - name: cool-down
        duration: 30s
        arrivalCount: 37
        maxVusers: 90

This configuration should create approximately 25,000 virtual users across all workers.

During the run, the console output contained many warnings such as:

Warning: multiple batches of metrics for period 1782301800000 2026-06-24T11:50:00.000Z

The final aggregated report should contain metrics for all workers and show approximately:

vusers.created: ~25,000

The final report contains significantly fewer metrics than expected:

vusers.completed: 2669
vusers.created: 3514
vusers.failed: 276

However, after aggregating the results directly from the individual worker logs stored in S3, I get:

vusers.completed: 20627
vusers.created: 24794
vusers.failed: 4167

The worker logs appear to contain the expected data, but the final aggregated report is missing a large portion of the metrics.

Given the repeated:

Warning: multiple batches of metrics for period

messages, this may be related to metrics aggregation when multiple workers submit metrics for the same reporting period.

The issue is reproducible across multiple runs and appears consistently in large distributed Fargate tests.