When streaming the results to the k6 Cloud, the machine - where you execute the k6 CLI command - runs the test and uploads the results to the k6 Cloud. Then, you will be able to visualize and analyze the results on the web app in real-time.
Don't confuse k6 --out cloud script.js (what this page is about) with k6 cloud script.js. While the former means that you run k6 locally and stream the result to the cloud, the latter means that you upload your script to k6 cloud and tell the cloud infrastructure to run the entire test for you. In this case you'll only see status updates in your CLI, but in all cases you'll be able to see your test results at https://app.k6.io.
1 (Optional) - Log in to the k6 Cloud
Assuming you have installed k6, the first step is to log in to k6 Cloud. You can use your API token or username and password:
2 - Run the tests and upload the results
Now, k6 will authenticate you against the k6 Cloud, and you can use the --out option to send the k6 results to the k6 Cloud as:
Alternatively, you could skip the k6 login command when using your API token with the k6 run command as:
After running the command, the console shows an URL. Copy this URL and paste it in your browser's address bar to visualize the test results.
When you send the results to the k6 Cloud with k6-out, data will be continuously streamed to the cloud. While this happens the state of the test run will be marked as Running. A test run that ran its course will be marked Finished. The run state has nothing to do with the test passing any thresholds, only that the test itself is operating correctly.
If you deliberately abort your test (e.g. by pressing Ctrl-C), your test will end up as Aborted by User. You can still look and analyze the test data you streamed so far. The test will just have run shorter than originally planned.
Another possibility would be if you lose network connection with the k6 Cloud while your test is running. In that case the k6 Cloud will patiently wait for you to reconnect. In the meanwhile your test's run state will continue to appear as Running on the web app.
If no reconnection happens, the k6 Cloud will time out after two minutes of no data, setting the run state to Timed out. You can still analyze a timed out test but you'll of course only have access to as much data as was streamed before the network issue.
There are a few special environment variables specifically for controlling k6 when streaming with k6 -o cloud.
When streaming, k6 will collect all data and send it to the cloud in batches.
|K6_CLOUD_METRIC_PUSH_INTERVAL||How often to send data to the k6 cloud (default '6s').|
k6 can also aggregate the data it sends to the k6 cloud each batch. This reduces the amount of data sent to the cloud. Aggregation is disabled by default.
When using aggregation, k6 will collect incoming test data into time-buckets. For each data-type collected in such a bucket, it will figure out the dispersion (by default the interquartile range) around the median value. Outlier-data far outside the lower and upper quartiles are not aggregated in order to not lose potentially important testing information.
|K6_CLOUD_AGGREGATION_PERIOD||>0s to activate aggregation (default disabled, '3s' is a good example to try)|
|K6_CLOUD_AGGREGATION_CALC_INTERVAL||How often new data will be aggregated (default '3s').|
|K6_CLOUD_AGGREGATION_WAIT_PERIOD||How long to wait for period samples to accumulate before aggregating them (default '5s').|
|K6_CLOUD_AGGREGATION_MIN_SAMPLES||If fewer samples than this arrived in interval, skip aggregation (default 100).|
|K6_CLOUD_AGGREGATION_OUTLIER_IQR_RADIUS||Outlier sampling from median to use for Q1 and Q3 quartiles (fraction) (default 0.25 (i.e. IQR)).|
|K6_CLOUD_AGGREGATION_OUTLIER_IQR_COEF_LOWER||How many quartiles below the lower quartile are treated as non-aggregatable outliers (default 1.5)|
|K6_CLOUD_AGGREGATION_OUTLIER_IQR_COEF_UPPER||How many quartiles above the upper quartile are treated as non-aggregatable outliers (default 1.3)|
When running a test entirely in the cloud with k6 cloud, k6 will always aggregate. For that case the aggregation settings are however set by the cloud infrastructure and are not controllable from the CLI.