Skip to content

Latest commit

 

History

History
631 lines (537 loc) · 14.6 KB

index.adoc

File metadata and controls

631 lines (537 loc) · 14.6 KB
Raw

Kickstart Your Gatling Performance Testing

Introduction

gatling logo

Siegfried Goeschl

  • Senior Software Engineer

  • Writing server-side code

  • Java Meetup Vienna co-organizer

  • Apache Software Foundation member

  • Currently working at Erste Bank Austria

Introducing Gatling

  • Performance testing framework

  • Tests are written in in Scala

  • Developer-centric test tool

  • Development started in 2010

  • Gatling 3.0.1 released now

  • Since V3 there are two license models - free & commercial.

  • As you know some guys have a strong opinion about OSS

What Linus Says

linus torvalds quote

  • Having said that a commercial license could generate more revenue keeping the Open Source version alive.

Money Makes The World Go Round

background

Gatling vs. FrontLine

  • Gatling Open Source is under ASL 2.0

  • Gatling FrontLine is the enterprise edition

    • Annual license or "pay as you go"

    • Web-based,

    • More bells & whistle

    • Real-time reporting

Under The Hood

  • Supports HTTP 1.1/2.0 & JMS protocol

  • Response validation

    • Regular expressions

    • XPath & JSONPath

    • CSS selectors

Under The Hood

  • Provides Domain Specific Language (DSL)

  • Uses asynchronous non-blocking HTTP client

  • Integrates with Maven, SBT & Gradle

  • Test data feeders CSV, JSON, JDBC, Redis

  • Management-friendly HTML reports

  • No more 1:1 mapping between virtual users and worker threads.

When To Use Gatling?

  • Want to write test code in your IDE?

  • Need some integration & performance tests?

  • Want to run those test on your CI server?

  • Do you care about reviews and version control?

Getting Started

background

Getting Started

  • JDK 1.8

  • Apache Maven 3.5.x

  • IntelliJ Community Edition

  • IntelliJ Scala Plugin

Getting Started

  • The official Gatling distributable is not suited for development.

  • The Gatling Maven archetype project does not use Maven Gatling plugin.

    • You can also use SBT & Gradle if you know your way around.

IntelliJ & Gatling

intellig gatling screenshot

  • That is what you see when you import the gatling-maven-plugin-demo into IntelliJ.

Gatling Run Configuration

Parameter Value

Main Class

Engine

VM Options

-Dgatling.core.simulationClass=XXX

  • You need to tell IntelliJ which Gatling tests to execute …​.

Gatling Run Configuration

intellij gatling configuration

Execute Gatling in IntelliJ

intellij running gatling

  • The output of the pre-packaged demo project (computer database)

First Gatling Report

gatling reports 01

  • The report is generated in the target/gatling folder

Execute Gatling Wit Maven

mvn -Dgatling.simulationClass=computerdatabase.BasicSimulation gatling:test

  • Start Gatling from the Maven command line.

  • Please note that different system properties are used!!!

  • Perfect way to integrate with Jenkins

Hello World

background

Gatling Hello World

package postman

import io.gatling.core.Predef._
import io.gatling.http.Predef._

class HelloWorldSimulation extends Simulation {

  val httpProtocol = http.baseUrl("https://postman-echo.com")

  val scn = scenario("Hello World")
    .exec(http("GET").get("/get?msg=Hello%20World"))

  setUp(scn.inject(atOnceUsers(1)).protocols(httpProtocol))
}

Gatling For Rookies

  • Script setup

  • Common HTTP configuration

  • Scenario & load simulation setup

  • Load simulation text report

  • Creating Gatling scripts

Script Setup

package postman

import io.gatling.core.Predef._
import io.gatling.http.Predef._

class PostmanSimulation extends Simulation {

  • Gatling tests are deriving from Simulation

Common HTTP Configuration

val httpProtocol = http
    .baseUrl("https://postman-echo.com")
    .acceptHeader("text/html,application/xhtml+xml,;q=0.9,*/*;q=0.8")
    .acceptEncodingHeader("gzip, deflate")
    .acceptLanguageHeader("en-US,en;q=0.5")
    .userAgentHeader("Gatling/3.0.0")

Scenario Setup

val scn = scenario("Postman")
    .exec(http("GET")
      .get("/get?msg=Hello%20World")
      .check(bodyBytes.transform(_.length > 200).is(true))
    )
    .exec(http("POST")
      .post("/post")
      .formParam("""foo""", """bar""")
      .check(status in (200, 201))
      .check(bodyBytes.exists)
    )

Load Simulation Setup

At Once User

setUp(
  scn.inject(
    atOnceUsers(10) // (1)
  ).protocols(httpConf)
)

  1. Injects a given number of users at once

Rampup Users

setUp(
  scn.inject(
    rampUsers(10) over(5 seconds) // (1)
  ).protocols(httpConf)
)

  1. Start 10 user within 5 seconds ⇒ 10 users

Constant Users

setUp(
  scn.inject(
    constantUsersPerSec(20) during(15 seconds) // (1)
  ).protocols(httpConf)
)

  1. Start 20 users / second for 15 seconds ⇒ 300 users

Heaviside Users

setUp(
  scn.inject(
    heavisideUsers(1000) over(20 seconds) // (1)
  ).protocols(httpConf)
)

  1. Create 1.000 users in 20 seconds using Heaviside step function

Response Time Assertions

setUp(scn)
  .assertions(global.responseTime.max.lt(100)) // (1)

  1. Max response time of all requests is less than 100 ms

Simulation Text Report

=============================================================
2018-11-16 20:43:51                        2s elapsed
---- Requests -----------------------------------------------
> Global                                (OK=2      KO=0     )
> GET                                   (OK=1      KO=0     )
> POST                                  (OK=1      KO=0     )

---- Postman ------------------------------------------------
[#######################################################]100%
       waiting: 0      / active: 0      / done: 1
=============================================================

Simulation Text Report

---- Global Information -------------------------------------
> request count                       2 (OK=2      KO=0     )
> min response time                 118 (OK=118    KO=-     )
> max response time                 604 (OK=604    KO=-     )
> mean response time                361 (OK=361    KO=-     )
> std deviation                     243 (OK=243    KO=-     )
> response time 50th percentile     361 (OK=361    KO=-     )
> response time 75th percentile     483 (OK=483    KO=-     )
> response time 95th percentile     580 (OK=580    KO=-     )
> response time 99th percentile     599 (OK=599    KO=-     )
> mean requests/sec                   2 (OK=2      KO=-     )
---- Response Time Distribution -----------------------------
> t < 800 ms                          2 (100%)
> 800 ms < t < 1200 ms                0 (  0%)
> t > 1200 ms                         0 (  0%)
> failed                              0 (  0%)
=============================================================

Creating Gatling Scripts

  • Gatling Web Proxy Recorder

  • Start from the scratch

    • More initial work

    • Clean test code

  • Import HTTP Archive Format

  • Since I’m testing REST APIs I’m crafting my Gatling scripts from the documentation.

Beyond Hello World

background

Things Not Being Told In Tutorials

Please Note That The Following Problems Are Not Specific To Gatling!

Hard-coded Server Address

val httpConf = http
    .baseURL("http://computer-database.gatling.io") // (1)
    .acceptHeader("text/html,application/xhtml+xml,application/xml")
    .doNotTrackHeader("1")
    .acceptLanguageHeader("en-US,en;q=0.5")
    .acceptEncodingHeader("gzip, deflate")
    .userAgentHeader("Mozilla/5.0 (Macintosh; Intel Mac OS X 10.8; rv:16.0)")

  1. Need to support staging environments?

  • You might start writing your test hitting you local box or DEV environment.

  • later you want to switch to FAT, UAT & PROD.

Hard-coded CSV Files

val feeder = csv("users.csv").random // (1)

  1. Different users / passwords for staging environments?

Hard-coded User Injection

setUp(
  users.inject(rampUsers(10) over (10 seconds)), // (1)
  admins.inject(rampUsers(2) over (10 seconds))
).protocols(httpConf)

  1. Different load for staging environments?

  • Your load testing environment might be a lot smaller than PROD.

Configuration Overload

Http(getURL("identity", "oauth/token"))
    .postForm(Seq(
        "scope" -> identityScope, 	      // (1)
        "grant_type" -> identityGranType,
        "client_id" -> identityClientId,
        "client_secret" -> identityClientSecret,
        "resource" -> identityResource
      ))

  1. Tons of configurable properties?

  • How to pass the configuration properties which might be dependent on your staging environment?

How To Pass Settings

theres more than one way to skin a cat

How To Pass All The Settings

  • System properties

  • Maven profiles

  • Custom Scala class

  • Unhappy with those approaches

  • I came up with Gatling Blueprint Project

Gatling Blueprint Project

background

Gatling Blueprint Project

  • Gatling Blueprint Project - a recipe of how to do things with Gatling

  • Use it like a cooking recipe - try it and change it to your personal taste

Simulation Coordinates

background

Simulation Coordinates

Tenant

The tenant to test (AT, CZ, SK)

Application

Application to simulate (web, mobile)

Site

Staging site to be tested (dev, prod)

Scope

Scope of test (smoke, performance)

Why Did I Write The Gatling Blueprint Project?

Why Did I Write The Gatling Blueprint Project?

george online banking logo

  • I had a problem - it was called George

Introducing George

george online banking

  • George is Erste Bank Austria’s Online Banking

  • It became a group-wide solution for Online Banking

Introducing George

background

George International Team

  • Erste Bank Austria,

  • Česká spořitelna,

  • Slovenská sporiteľňa

  • Banca Comercială Română

  • I was part of George International Team for 2 years

  • George is Erste Bank Austria’s Online Banking

  • It became a group-wide solution for Online Banking

George & Gatling

  • Many moving parts & staging sites

  • Gatling for automated integration tests

  • Internal performance testing

  • Continuous performance testing?

  • Other teams use JMeter & Neoload

  • Continuous performance testing is a cultural problem not a technical

When Are We Using Gatling?

  • Integration tests across tenants & sites

  • Developer driven performance testing

  • Elastic Search server migration & tuning

  • Desaster recovery tests

  • Detecting changes between releases

Real Test Code

read the source luke

Real Test Code

gatling production code

  • This is real code being used for George API performance testing

  • CSV file being used is resolved dynamically

  • HTTP configuration hidden behind a factory

  • Test steps are also create by a factory method

  • Load scenario configurable using external properties

Is Gatling For You?!

Is Gatling For You?!

  • Gatling’s DSL is elegant & powerful

    • Programming power at your finger tips

  • Scala & DSL learning curve

    • Requires solid development skills

  • Works on Windows, Linux & OS X

Is Gatling For You?!

  • Developer-friendly tool

  • Code only, IDE support & refactoring

  • Integrates nicely into your build process

  • Do you need to onboard your test team?

Is Gatling For You?!

background

Questions?!

background