- Maven 3.3.9+
- Java (JDK) 8+
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:
- Intel® Core™ i9-9900K (8 cores)
- 128 GB DDR4 RAM
- 2 TB SSD
- 1 GBit/s-Port connection
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.
- Start ES and Kibana instances.
Important: Run Kibana without base path or add a static one to your kibana.yml like server.basePath: "/xfh" before start.
- Update Kibana configuration in /resources/config/local.conf file
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 scenario
mvn 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
- Create Elastic Cloud 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 = <username>
password = <password>
}
- start test scenario with your newly added config
mvn clean test-compile
mvn gatling:test -Denv=config/cloud-8.5.0.conf -Dgatling.simulationClass=org.kibanaLoadTest.simulation.cloud.LensJourney
- Generate API_KEY for your cloud user account
- 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 finished
mvn clean test-compile
export API_KEY=<your_cloud_key>
mvn gatling:test -DcloudStackVersion=7.11.0-SNAPSHOT -Dgatling.simulationClass=org.kibanaLoadTest.simulation.cloud.DemoJourney
- 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
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
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=<path_to_json_file>
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
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.
Follow guide to setup VM and run tests on it.
Run the following command to delete all existing deployments
export API_KEY=<your_key>
mvn exec:java -Dexec.mainClass=org.kibanaLoadTest.deploy.DeleteAll -Dexec.classpathScope=test -Dscope=all
If you don't provide -Dscope=all it will delete only the ones with load-testing name prefix