Analyze results

In this tutorial, learn how to:

Apply tags to filter specific results
Learn about k6 metrics
Use jq to filter JSON results
Define groups to organize the test
Create custom metrics

Context: k6 result outputs

k6 provides many result outputs. By default, the end-of-test summary provides the aggregated results of the test metrics.

checks.........................: 50.00% ✓ 45       ✗ 45  
data_received..................: 1.3 MB 31 kB/s
data_sent......................: 81 kB  2.0 kB/s
group_duration.................: avg=6.45s    min=4.01s    med=6.78s    max=10.15s   p(90)=9.29s    p(95)=9.32s   
http_req_blocked...............: avg=57.62ms  min=7µs      med=12.25µs  max=1.35s    p(90)=209.41ms p(95)=763.61ms
http_req_connecting............: avg=20.51ms  min=0s       med=0s       max=1.1s     p(90)=100.76ms p(95)=173.41ms
http_req_duration..............: avg=144.56ms min=104.11ms med=110.47ms max=1.14s    p(90)=203.54ms p(95)=215.95ms
  { expected_response:true }...: avg=144.56ms min=104.11ms med=110.47ms max=1.14s    p(90)=203.54ms p(95)=215.95ms
http_req_failed................: 0.00%  ✓ 0        ✗ 180 
http_req_receiving.............: avg=663.96µs min=128.46µs med=759.82µs max=1.66ms   p(90)=1.3ms    p(95)=1.46ms  
http_req_sending...............: avg=88.01µs  min=43.07µs  med=78.03µs  max=318.81µs p(90)=133.15µs p(95)=158.3µs 
http_req_tls_handshaking.......: avg=29.25ms  min=0s       med=0s       max=458.71ms p(90)=108.31ms p(95)=222.46ms
http_req_waiting...............: avg=143.8ms  min=103.5ms  med=109.5ms  max=1.14s    p(90)=203.19ms p(95)=215.56ms
http_reqs......................: 180    4.36938/s
iteration_duration.............: avg=12.91s   min=12.53s   med=12.77s   max=14.35s   p(90)=13.36s   p(95)=13.37s  
iterations.....................: 45     1.092345/s
vus............................: 1      min=1      max=19
vus_max........................: 20     min=20     max=20

For simplicity to learn about k6 metric results, this tutorial uses the JSON output and jq to filter results.

For other options to analyze test results such as storage and time-series visualizations in real-time, refer to:

Write time-series results to a JSON file

To output results to a JSON file, use the --out flag.

k6 run --out json=results.json api-test.js

Then run this jq command to filter the latency results; http_req_duration metric.

jq '. | select(.type == "Point" and .metric == "http_req_duration")' results.json

k6 results have a number of built-in tags. For example, filter results to only results where the status is 200.

jq '. | select(.type == "Point" and .data.tags.status == "200")' results.json

Or calculate the aggregated value of any metric with any particular tags.

jq '. | select(.type == "Point" and .metric == "http_req_duration") | .data.value' results.json | jq -s 'add/length'

Apply custom tags

You can also apply Tags to requests or code blocks. For example, this is how you can add a tags to the request params.

  const params = {
    headers: {
      "Content-Type": "application/json",
    },
    tags: {
      "my-custom-tag": "auth-api",
    },
  };

Create a new script named "tagged-login.js", and add a custom tag to it.

import http from "k6/http";

export default function () {
  const url = "https://test-api.k6.io";
  const payload = JSON.stringify({
    username: "test_case",
    password: "1234",
  });

  const params = {
    headers: {
      "Content-Type": "application/json",
    },
    //apply tags
    tags: {
     "my-custom-tag": "auth-api"
    },
  };

  //Login with tags
  http.post(`${url}/auth/basic/login`, payload, params);
};

Run the test:

k6 run --out json=results.json tagged-login.js

Filter the results for this custom tag:

jq '. | select(.type == "Point" and .metric == "http_req_duration" and .data.tags."my-custom-tag" == "auth-api")' results.json

Organize requests in groups

You can also organize your test logic into Groups. Test logic inside a group tags all requests and metrics within its block. Groups can help you organize the test as a series of logical transactions or blocks.

Context: a new test to group test logic

Results filtering isn't very meaningful in a test that makes one request. And the API test script is getting long. To learn more about how to compare results and other k6 APIs, write a test for the following situation:

A dummy example: your development team wants to evaluate the performance of two user-facing flows.
visit an endpoint, then another one
A GET request to https://test.k6.io/contacts.php
A GET to https://test.k6.io/
play the coinflip game:
A POST request to https://test.k6.io/flip_coin.php with the query param bet=heads
Another POST to https://test.k6.io/flip_coin.php with the query param bet=tails

Can you figure out how to script the requests? If not, use the following script.

Since this example simulates a human user rather than an API call, it has a sleep between each request. Run with k6 run multiple-flows.js.

1import http from "k6/http";
2import { group, sleep } from "k6";
3
4//set URL as variable
5const baseUrl = "https://test.k6.io";
6
7export default function () {
8  // visit contacts
9  http.get(`${baseUrl}/contacts.php`);
10  sleep(1);
11  // return to the home page
12  http.get(`${baseUrl}/`);
13  sleep(1);
14
15  //play coinflip game
16  http.get(`${baseUrl}/flip_coin.php?bet=heads`);
17  sleep(1);
18  http.get(`${baseUrl}/flip_coin.php?bet=tails`);
19  sleep(1);
20}

Add Group functions

Wrap the two endpoints in different groups. Name one group Contacts flow and another Coinflip game.

1//import necessary modules
2import http from "k6/http";
3import { group, sleep } from "k6";
4
5//set baseURL
6const baseUrl = "https://test.k6.io";
7
8export default function () {
9
10  // visit some endpoints in one group
11  group("Contacts flow", function () {
12    http.get(`${baseUrl}/contacts.php`);
13    sleep(1);
14    // return to the home page
15    http.get(`${baseUrl}/`);
16    sleep(1);
17  });
18
19  // Coinflip players in another group
20  group("Coinflip game", function () {
21    http.get(`${baseUrl}/flip_coin.php?bet=heads`);
22    sleep(1);
23    http.get(`${baseUrl}/flip_coin.php?bet=tails`);
24    sleep(1);
25  });
26}

Run and filter

Inspect the results for only the Coinflip game group. To do so:

Save the preceding script as multiple-flows.js.

Run the script with the command:

k6 run multiple-flows.js --out json=results.json --iterations 10

Inspect the results with jq. Group names have a :: prefix.

jq '. | select(.data.tags.group == "::Coinflip game")' results.json

Add a custom metric

As you have seen in the output, all k6 tests emit metrics. However, if the built-in metrics aren't enough, you can create custom metrics. A common use case is to collect metrics of a particular scope of your test.

As an example, create a metric that collects latency results for each group:

Import Trend from the k6 metrics module.
Create two duration trend metric functions.
In each group, add the duration time to the trend for requests to contacts and the coin_flip endpoints.

1//import necessary modules
2import http from "k6/http";
3import {  group, sleep } from "k6";
4import { Trend } from "k6/metrics";
5
6//set baseURL
7const baseUrl = "https://test.k6.io";
8
9// Create custom trend metrics
10const contactsLatency = new Trend("contacts_duration");
11const coinflipLatency = new Trend("coinflip_duration");
12
13export default function () {
14  // Put visits to contact page in one group
15  let res;
16  group("Contacts flow", function () {
17    // save response as variable
18    res = http.get(`${baseUrl}/contacts.php`);
19    // add duration property to metric
20    contactsLatency.add(res.timings.duration);
21    sleep(1);
22
23    res = http.get(`${baseUrl}/`);
24    // add duration property to metric
25    contactsLatency.add(res.timings.duration);
26    sleep(1);
27  });
28
29  // Coinflip players in another group
30
31  group("Coinflip game", function () {
32    // save response as variable
33    let res = http.get(`${baseUrl}/flip_coin.php?bet=heads`);
34    // add duration property to metric
35    coinflipLatency.add(res.timings.duration);
36    sleep(1);
37
38    res = http.get(`${baseUrl}/flip_coin.php?bet=tails`);
39    // add duration property to metric
40    coinflipLatency.add(res.timings.duration);
41    sleep(1);
42  });
43}

Run the test with small number of iterations and output the results to results.json.

k6 run multiple-flows.js --out json=results.json --iterations 10

Look for the custom trend metrics in the end-of-test console summary.

coinflip_duration..............: avg=119.6438  min=116.481  med=118.4755 max=135.498  p(90)=121.8459 p(95)=123.89565
contacts_duration..............: avg=125.76985 min=116.973  med=120.6735 max=200.507  p(90)=127.9271 p(95)=153.87245

You can also query custom metric results from the JSON results. For example, to get the aggregated results as.

jq '. | select(.type == "Point" and .metric == "coinflip_duration") | .data.value' results.json | jq -s 'add/length'

Next steps

In this tutorial, you looked at granular output and filtered by built-in and custom tags. Then you made a new script with groups. Finally, you added a new metric for each group. A next step would be to create a Custom end-of-test summary or to stream the results to a database.

For ongoing operations, you can modularize your logic and configuration. That's the subject of the next step of this tutorial.

Examples

Tutorials