# Scraping Load Testing - Test #3

## Summary

The load test involved setting up an autoscaling option and adjusting the hard and soft limits for the Fly.io configuration. The test environment consisted of 5 machines, with 3 machines automatically scaling up during the test. Despite the scaling, there were 653 timeouts (7.3%) and 2 HTTP 502 responses (0.02%). The average response time was 3037.2 ms, with a peak response time of 9941 ms. Further adjustments to the soft limit are recommended to improve performance and reduce errors.

## Table of Contents

- [Scraping Load Testing - Test #3](#scraping-load-testing---test-3)
  - [Summary](#summary)
  - [Table of Contents](#table-of-contents)
  - [Test environment](#test-environment)
    - [Machines](#machines)
  - [Load Test Phases](#load-test-phases)
    - [Configuration](#configuration)
    - [Results](#results)
    - [Metrics](#metrics)
  - [Conclusions and Next Steps](#conclusions-and-next-steps)
    - [Conclusions](#conclusions)
    - [Next Steps](#next-steps)

## Test environment
### Machines

| Machine | Size/CPU | Status |
|---|---|---|
| e286de4f711e86 mia (app) | performance-cpu-1x@2048MB | always on |
| 73d8dd909c1189 mia (app) | performance-cpu-1x@2048MB | always on |
| 6e82050c726358 mia (app) | performance-cpu-1x@2048MB | paused |
| 4d89505a6e5038 mia (app) | performance-cpu-1x@2048MB | paused |
| 48ed6e6b74e378 mia (app) | performance-cpu-1x@2048MB | paused |

---

## Load Test Phases

### Configuration

```toml
# fly.staging.toml
[http_service.concurrency]
  type = "requests"
  hard_limit = 100
  soft_limit = 75
```
```yml
# load-test.yml
- duration: 60
arrivalRate: 10  # Initial load
- duration: 120
arrivalRate: 20  # Increased load
- duration: 180
arrivalRate: 30  # Peak load
- duration: 60
arrivalRate: 10  # Cool down
```


### Results
Date: 14:53:32(-0300)

| Metric                                      | Value   |
|---------------------------------------------|---------|
| errors.ETIMEDOUT                            | 653     |
| errors.Failed capture or match              | 2       |
| http.codes.200                              | 8345    |
| http.codes.502                              | 2       |
| http.downloaded_bytes                       | 0       |
| http.request_rate                           | 11/sec  |
| http.requests                               | 9000    |
| http.response_time.min                      | 979     |
| http.response_time.max                      | 9941    |
| http.response_time.mean                     | 3037.2  |
| http.response_time.median                   | 2059.5  |
| http.response_time.p95                      | 7709.8  |
| http.response_time.p99                      | 9416.8  |
| http.responses                              | 8347    |
| vusers.completed                            | 8345    |
| vusers.created                              | 9000    |
| vusers.created_by_name.Scrape a URL         | 9000    |
| vusers.failed                               | 655     |
| vusers.session_length.min                   | 1044.5  |
| vusers.session_length.max                   | 9998.8  |
| vusers.session_length.mean                  | 3109.7  |
| vusers.session_length.median                | 2143.5  |
| vusers.session_length.p95                   | 7709.8  |
| vusers.session_length.p99                   | 9416.8  |

### Metrics 

![](./assets/metrics-test-3.png)

---

## Conclusions and Next Steps

### Conclusions
1. **Performance:** The system handled 9000 requests with a mean response time of 3037.2 ms. There were 653 timeouts and 2 HTTP 502 responses.
2. **Autoscaling:** Three machines automatically scaled up during the test, but the scaling was not sufficient to prevent all errors.
3. **Response Times:** The peak response time was 9941 ms, indicating that the system struggled under peak load conditions.

### Next Steps

1. **Adjust Soft Limit:** Change the soft limit to 100 and the hard limit to 50 to test if machines will start faster and reduce the number of 502 errors.
2. **Further Load Tests:** Conduct additional load tests with the new configuration to assess improvements.

By following these steps, we can enhance the system's performance and reliability under varying load conditions.