Docs

Docs

Load Testing Thresholds

How to set and control failure thresholds for the load tests

Overview

k6 thresholds define pass/fail criteria for your load tests.

When a threshold is breached, k6 marks the test as failed, and the Maven build fails if k6.failOnThreshold=true (the default).

The TestBench k6 converter plugin generates some default thresholds automatically in every k6 script produced. You can configure the default thresholds, disable individual ones, or add custom thresholds for any k6 metric.

Default Thresholds

Every generated k6 script includes these thresholds out of the box:

Metric Expression Default

http_req_duration

95th percentile response time

p(95)<2000 (2 seconds)

http_req_duration

99th percentile response time

p(99)<5000 (5 seconds)

checks

All checks must pass

rate==1 with abortOnFail: true

This generates the following block in the k6 script:

Source code
JavaScript
export const options = {
  thresholds: {
    checks: [{ threshold: 'rate==1', abortOnFail: true, delayAbortEval: '5s' }],
    http_req_duration: ['p(95)<2000', 'p(99)<5000'],
  },
}

Editing Default Thresholds

You can change the default threshold values with Maven properties or through configuration in the POM.

Via Command Line

Source code
bash
# Tighter response time requirements
mvn k6:convert -Dk6.harFile=recording.har \
    -Dk6.threshold.httpReqDurationP95=1000 \
    -Dk6.threshold.httpReqDurationP99=3000

# Allow checks to fail without aborting the test
mvn k6:convert -Dk6.harFile=recording.har \
    -Dk6.threshold.checksAbortOnFail=false

# Disable the p95 threshold entirely (set to 0)
mvn k6:convert -Dk6.harFile=recording.har \
    -Dk6.threshold.httpReqDurationP95=0

Via POM Configuration

Source code
XML
<plugin>
    <groupId>com.vaadin</groupId>
    <artifactId>testbench-converter-plugin</artifactId>
    <configuration>
        <!-- 95th percentile response time in ms (0 to disable) -->
        <httpReqDurationP95>1000</httpReqDurationP95>
        <!-- 99th percentile response time in ms (0 to disable) -->
        <httpReqDurationP99>3000</httpReqDurationP99>
        <!-- Set to false to continue the test even when checks fail -->
        <checksAbortOnFail>false</checksAbortOnFail>
    </configuration>
</plugin>

Default Threshold Parameters

Parameter Default Description

k6.threshold.httpReqDurationP95

2000

95th percentile response time threshold in ms. Set to 0 to disable.

k6.threshold.httpReqDurationP99

5000

99th percentile response time threshold in ms. Set to 0 to disable.

k6.threshold.checksAbortOnFail

true

When true, the test aborts immediately on the first check failure. When false, failures are recorded but the test continues.

Adding Custom Thresholds

You can add pass/fail thresholds for any k6 built-in metric using the k6.threshold.custom property.

The format is a comma-separated list of metric:expression pairs:

Source code
metric1:expression1,metric2:expression2,...

Via Command Line

Source code
bash
# Fail if more than 1% of requests fail
mvn k6:run -Dk6.testFile=test.js \
    -Dk6.threshold.custom="http_req_failed:rate<0.01"

# Multiple custom thresholds
mvn k6:run -Dk6.testFile=test.js \
    -Dk6.threshold.custom="http_req_failed:rate<0.01,http_req_waiting:p(95)<500,http_reqs:count>100"

Via POM Configuration

Source code
XML
<plugin>
    <groupId>com.vaadin</groupId>
    <artifactId>testbench-converter-plugin</artifactId>
    <configuration>
        <!-- Custom thresholds in metric:expression format -->
        <customThresholds>http_req_failed:rate&lt;0.01,http_req_waiting:p(95)&lt;500</customThresholds>
    </configuration>
</plugin>
Note
 In XML, use < instead of < inside threshold expressions.

Combining Defaults and Custom Thresholds

Custom thresholds are added alongside the defaults. If you add a custom http_req_duration expression, it is merged with the default p95/p99 thresholds:

Source code
bash
# Adds a median threshold alongside the existing p95 and p99
mvn k6:convert -Dk6.harFile=recording.har \
    -Dk6.threshold.custom="http_req_duration:p(50)<1000"

This produces:

Source code
JavaScript
export const options = {
  thresholds: {
    checks: [{ threshold: 'rate==1', abortOnFail: true, delayAbortEval: '5s' }],
    http_req_duration: ['p(95)<2000', 'p(99)<5000', 'p(50)<1000'],
  },
}

Common Threshold Examples

Below are commonly used threshold configurations for Vaadin load tests.

Strict Performance Requirements

Source code
bash
mvn k6:run -Dk6.testFile=test.js \
    -Dk6.threshold.httpReqDurationP95=500 \
    -Dk6.threshold.httpReqDurationP99=1000 \
    -Dk6.threshold.custom="http_req_failed:rate<0.001,http_req_waiting:p(95)<300"

Generated output:

Source code
JavaScript
thresholds: {
  checks: [{ threshold: 'rate==1', abortOnFail: true, delayAbortEval: '5s' }],
  http_req_duration: ['p(95)<500', 'p(99)<1000'],
  http_req_failed: ['rate<0.001'],
  http_req_waiting: ['p(95)<300'],
},

Relaxed Thresholds for Stress Testing

Source code
bash
mvn k6:run -Dk6.testFile=test.js \
    -Dk6.threshold.httpReqDurationP95=10000 \
    -Dk6.threshold.httpReqDurationP99=0 \
    -Dk6.threshold.checksAbortOnFail=false \
    -Dk6.threshold.custom="http_req_failed:rate<0.05"

Generated output:

Source code
JavaScript
thresholds: {
  checks: ['rate==1'],
  http_req_duration: ['p(95)<10000'],
  http_req_failed: ['rate<0.05'],
},

Throughput Minimum

Source code
bash
mvn k6:run -Dk6.testFile=test.js \
    -Dk6.threshold.custom="http_reqs:count>1000"

k6 Metric Reference

The following k6 built-in metrics can be used with custom thresholds. For the full reference, see the k6 metrics documentation.

Metric Type Description

http_req_duration

Trend

Total time for the request (send + wait + receive)

http_req_waiting

Trend

Time spent waiting for response (Time to First Byte)

http_req_connecting

Trend

Time spent establishing TCP connection

http_req_sending

Trend

Time spent sending data

http_req_receiving

Trend

Time spent receiving data

http_req_failed

Rate

Rate of failed requests

http_reqs

Counter

Total number of requests

checks

Rate

Rate of successful checks

Trend metrics support: p(N)<value, avg<value, min<value, max<value, med<value
Rate metrics support: rate<value, rate>value
Counter metrics support: count<value, count>value

Updated –