Kibana CI has dedicated jobs to run performance testing for your Kibana branch or Cloud snapshot.
We are using bare metal machine EX-62 provided by Hezner:
Execution is managed by functional test runner: we have custom FTR config file, that defines how to start Elasticsearch and Kibana servers. We also use custom runner to start Gatling simulation file. At the moment both ES/Kibana and Gatling runner are hosted on the machine.
Note: While running locally a high load test you might face different issues, so we suggest using dedicated machines and make sure you are aware of needed environment tunnings to minimise side effects.
Important: Run Kibana without base path or add a static one to your kibana.yml like server.basePath: "/xfh" before start.
host {
kibana = "http://localhost:5620" // Kibana base url
// "http://localhost:5620/xhf" if you start Kibana with static base path
es = "http://localhost:9220" // ElasticSearch base url
version = "8.0.0" // Stack version
}security { on = true // false for OSS, otherwise - true }
auth { providerType = "basic" providerName = "basic" username = "elastic" // user should have permissions to load sample data and access plugins password = "changeme" }
- start test scenariomvn clean test-compile // if you made any changes to the config or simulations mvn gatling:test -Dgatling.simulationClass=org.kibanaLoadTest.simulation.branch.DemoJourney // could be any other existing simulation class
## Running simulation against existing cloud deployment
- Create [Elastic Cloud](https://cloud.elastic.co/) deployment
- Add a new configuration file in `src/test/resources/config`: `cloud-8.5.0.conf`host { kibana = "https://gcp-8-5-0-def.kb.us-central1.gcp.cloud.es.io:9243" es = "https://gcp-8-5-0-def.es.us-central1.gcp.cloud.es.io" version = "8.5.0" }
security { on = true }
auth {
providerType = "basic"
providerName = "cloud-basic"
username =
- start test scenario with your newly added configmvn clean test-compile mvn gatling:test -Denv=config/cloud-8.5.0.conf -Dgatling.simulationClass=org.kibanaLoadTest.simulation.cloud.LensJourney
## Running simulation against newly created cloud deployment
- Generate API_KEY for your cloud user account
- https://cloud.elastic.co/deployment-features/keys
- Check deployment template at `src/test/resources/config/deploy/default.conf`
- start test scenario, new deployment will be created before simulation and deleted after it is finishedmvn clean test-compile
export API_KEY=
- Optionally create a custom deployment configuration and pass it in command `-DdeploymentConfig=config/deploy/custom.conf`
Follow logs to track deployment status:
09:40:23.535 [INFO ] httpClient - preparePayload: Using Config(SimpleConfigObject({"elasticsearch":{"deployment_template":"gcp-io-optimized","memory":8192},"kibana":{"memory":1024},"version":"7.11.0-SNAPSHOT"})) 09:40:23.593 [INFO ] httpClient - createDeployment: Creating new deployment 09:40:29.848 [INFO ] httpClient - createDeployment: deployment b76dd4a9255a417ca133fe8edd8157a2 is created 09:40:29.848 [INFO ] httpClient - waitForClusterToStart: waitTime 300000ms, poolingInterval 20000ms 09:40:30.727 [INFO ] httpClient - waitForClusterToStart: Deployment is in progress... Map(kibana -> initializing, elasticsearch -> initializing, apm -> initializing) ... 09:46:01.211 [INFO ] httpClient - waitForClusterToStart: Deployment is in progress... Map(kibana -> reconfiguring, elasticsearch -> started, apm -> started) 09:46:21.989 [INFO ] httpClient - waitForClusterToStart: Deployment is ready! ... ... 10:01:08.146 [INFO ] i.g.c.c.Controller - StatsEngineStopped simulation org.kibanaLoadTest.simulation.cloud.DemoJourney completed in 429 seconds 10:01:08.148 [INFO ] httpClient - deleteDeployment: Deployment b76dd4a9255a417ca133fe8edd8157a2 10:01:09.440 [INFO ] httpClient - deleteDeployment: Finished with status code 200
## Adding new simulation
The simplest way is to add new class in `simulation` package:class MySimulation extends BaseSimulation { val scenarioName = s"My new simulation ${appConfig.buildVersion}"
val scn = scenario(scenarioName) .exec( Login .doLogin( appConfig.isSecurityEnabled, appConfig.loginPayload, appConfig.loginStatusCode ) .pause(5 seconds) ) // conbine your simulation using existing scenarios or adding new ones .exec(Discover.doQuery(appConfig.baseUrl, defaultHeaders).pause(5 seconds)) .exec(...) .exec(...)
// Define load model, check https://gatling.io/docs/current/general/simulation_setup/ setUp( scn .inject( rampConcurrentUsers(10) to (250) during (4 minute) ) .protocols(httpProtocol) ).maxDuration(10 minutes) }
In order to run your simulation, use the following command:mvn gatling:test -Dgatling.simulationClass=org.kibanaLoadTest.simulation.MySimulation
## Running simulation from APM traces collected during single user journey run on CI
We created `GenericJourney` simulation in order run scalability testing for a single user journey.
Simulation reads json file with APM traces directly in Gatling runtime and
make api calls defined in the file.
In order to run it, pass json file using the following command:
mvn gatling:test -Dgatling.simulationClass=org.kibanaLoadTest.simulation.generic.GenericJourney -DjourneyPath=
It is possible to override journey config by setting custom values via environment variables:
KIBANA_HOST, ES_URL, AUTH_PROVIDER_TYPE, AUTH_PROVIDER_NAME, AUTH_LOGIN, AUTH_PASSWORD
It is possible to skip unloading kbn & es archives on journey teardown (e.g. you want to inspect Kibana):-DskipCleanupOnTeardown=true
## Test results
Gatling generates html report for each simulation run, available in `<project_root>/target/gatling/<simulation>`path
Open `index.html` in browser to preview the report.
Open `testRun.txt` to find more about Kibana instance you tested.
## Running performance testing from VM
Follow [guide](docs/VM_SETUP.md) to setup VM and run tests on it.
# Delete your deployments on Elastic cloud
Run the following command to delete all existing deploymentsexport API_KEY=
If you don't provide `-Dscope=all` it will delete only the ones with `load-testing` name prefix