As a part of this ticket, we tested liftwing API from a home connection. The datacenter hosting liftwing API server has a peering connection with aws. So, we are supposed to see better response times when the test script is run from aws.

In this ticket, by re-running the test from aws, we would like to verify the following for liftwing revertrisk-language-agnostic endpoint:

• we are able to send upto 250 K requests in a 1 hour duration (status code 200). Beyond this, requests should error out (status code 4xx)
• for large percentage of requests, the response time is <= 500 ms.

Notes:

1. Use a user agent as prescribed by the ML team for the requests.
2. During the span of an hour, 250K requests should be processed with status code 200. Capture the error returned (with other status codes), if any, for these requests. Get the percentage of error'd out (other status code) requests.
3. During the span of the same hour, requests beyond 250K should return an error (status code 4xx). Capture the error for a couple of these kind of request.
4. Gather similar response time statistics as in aforementioned ticket.
5. Run a few iterations of the test for 2 or 3 hours.

To-Do

• create a GitLab repository that will hold the testing code
• create a CI/CD pipeline that will run the code from the repository (pay attention to the runner tag)
• trigger the pipeline to collect the test results

Things to consider:

• DB dependency