Service Level Agreement Profiles¶
The SLA Profiles page lets you configure, in one place, the Service Level Agreement of the application tested.
- Create several SLA profiles, for instance to deal with transactions that have very different expected response times.
- Apply several criteria inside a single SLA profile, for instance when you need to control response times and error rate.
Tip
To apply an SLA profile, you can drop an SLA action under a HTTP Request or Container in the Virtual User tree, or at the root of the Virtual User.
SLA profiles list¶
The SLA Profile page uses a layout with the list on the left and edition and usages on the right. The list features all the usual possibilities as described in the dedicated section about lists.
SLA Usage¶
The Usage tree is a visualization of the selected SLA profile. By clicking on a virtual User, you can navigate to it.
SLA Configuration¶
| Name | Description |
|---|---|
| Metric | The metric that this Threshold will apply to. Only raw metrics can be used, calculated metrics like percentiles are not available since they are processed afterwards. The custom metric statistics judge a metric your scripts record, see below. |
| Name | Display name for this SLA. |
| Severity | Passed for no SLA triggered, otherwise pick warning or critical depending on severity. |
| From, To | Define the boundaries of this SLA. |
| During | Number of occurrences or Time waited until a failed SLA is reported. Use this to avoid having too many failed SLA reports because of single hits above th |
Note
An SLA profile can contain several Thresholds and every threshold can contain several SLAs. Use this to create a set of rules that corresponds to your application.
Custom metrics¶
An SLA profile also judges the custom metrics of your tests: the Trend, Counter, Gauge and Rate of a K6 script, or the Core Web Vitals every Playwright test measures.
Pick one of the custom metric statistics as the Metric of a threshold, then:
| Name | Description |
|---|---|
| Custom metric name | The name the script records the metric under, for instance LCP or checkout_duration. The Core Web Vitals are suggested. |
| Unit | The unit of the values: milliseconds, bytes or no unit. The bands are typed in this unit. It only applies to the average, minimum, maximum and total: a count and a rate keep their own unit. |
Naming a Core Web Vital fills the threshold with Google's bands, on the average:
| Metric | Warning (needs improvement) | Critical (poor) |
|---|---|---|
LCP |
over 2.5 s | over 4 s |
INP |
over 200 ms | over 500 ms |
CLS |
over 0.1 | over 0.25 |
FCP |
over 1.8 s | over 3 s |
TTFB |
over 0.8 s | over 1.8 s |
The Core Web Vitals card of the SLA Profiles page creates a profile with the five of them at once.
Info
Google rates the 75th percentile of the page loads of a whole visit. An SLA judges the average of each sampling period instead, three periods in a row by default, so that one slow page load raises nothing on its own. INP is only measured on a page that received a click or a key press: a period without one is not judged.
Warning
A threshold on a custom metric judges the whole Virtual User: drop its SLA action at the root of the Virtual User. Under a container, it is not evaluated.