Verify your thresholds are CI-ready

Before you connect k6 to a pipeline, confirm that your thresholds behave as gates. A threshold that passes means the deploy is acceptable. A threshold that fails means k6 returns a non-zero exit code, which the pipeline interprets as a failure. If your thresholds are too loose, every deploy passes regardless of performance. If you have no thresholds at all, k6 always exits with code 0 and the gate is meaningless.

To verify your thresholds are CI-ready, complete the following steps:

1. Open your k6 test script in your code editor and check the thresholds section in the options block. You should have at least one threshold defined. For example:

   JavaScript Copy

   export const options = {
     thresholds: {
       http_req_duration: ['p(95)<500'],
       http_req_failed: ['rate<0.01'],
     },
   };

   If your script doesn’t have thresholds, add them now. At minimum, include a latency threshold (http_req_duration) and an error rate threshold (http_req_failed).

2. Run your test and confirm it passes with exit code 0:

   Bash Copy

   k6 run test.js
   echo $?

   The output of echo $? should be 0, meaning all thresholds passed. If the thresholds failed, adjust either your thresholds or your test target so you have a known-passing state.

3. Temporarily tighten a threshold to force a failure. For example, change the latency threshold to an impossibly low value:

   JavaScript Copy

   thresholds: {
     http_req_duration: ['p(95)<1'],
     http_req_failed: ['rate<0.01'],
   },

4. Run the test again and confirm it fails with a non-zero exit code:

   Bash Copy

   k6 run test.js
   echo $?

   The exit code should be non-zero (for threshold failures, k6 commonly returns 99). k6 prints which thresholds were breached in the end-of-test summary. This is the signal your pipeline uses to block a deploy.

5. Revert the threshold to its original value so the test passes again.

Tip: Good CI thresholds are based on your actual performance baseline, not arbitrary numbers. If you’ve established a performance baseline, use the p95 latency and error rate from your baseline runs as starting points. To learn more, refer to Establish a performance baseline with k6 and Load test types.

In the next milestone, you parameterize your script so it can run against any environment without code changes.