JarCastingjstressy
JStressy
Lightweight framework for prototyping load/stress tools for Web applications. Each framework's component can be separately changed and reimplemented based on your needs. Such modularity with the ability to replace almost everything is achieved by incorporating OSGi framework. Current implementation has hardcoded dependencies on Vert.x and RxJava 2.x (can be changed in future). Apache Felix is used as an OSGi framework implementation. Scenarios are meant to be implemented as OSGi bundles, allowing you to implement hot code reload and other useful OSGi framework's features.
JStressy basic configuration
Basic JStressy configuration consists of global settings related to parameters of the host being tested and exposed JStressy mertics. The number of scenarios and test stages is described in stress plan section:
globals:
host: localhost # required: path to the host
port: 8080 # required: port
stressyMetricsPort: 8089 # required: port for exporting metrics (in Prometheus format)
stressyMetricsPath: /metrics # required: metrics path
useSsl: false # optional: turn on/off ssl (default: false)
insecureSsl: false # optional: trust all certificates (default: false)
maxConnections: 1000 # optional: max connections pool size (default: 1000)
stressPlan: # describes stress plan
stages: # list of stress plan stages
- name: First # optional: stage name
scenarioName: SearchStore # required: scenario name
stageDelay: 1s # required: delay stage execution for the specified time
stageDuration: 20m # required: stage duration
arrivalRate: 1 # required: scenario invocation rate (1 time per second)
rampArrival: 0.5 # optional: ramp scenario invocation rate (2 timer per second)
rampArrivalRate: 1 # optional: ramp rate (increase rampArrival each second)
rampInterval: 10m # optional: ramping interval
scenarioParameters: # optional: parameters map which is passed to the scenario
foo: bar
foo1: bar1
JStressy doesn't restrict changes to the format of the config if it follows the initial structure which is read by JStressy itself
JStressy assembly
JStressy provides Maven plugin which can be used to easily create and distribute custom JStressy builds. Developer just needs to describe the set of dependencies representing OSGi bundles (forming custom JStressy build). JStressy Maven plugin configuration example (full example can be seen in jstressy-assembly-osgi pom.xml):
<plugin>
<groupId>com.github.timofeevda.jstressy</groupId>
<artifactId>jstressy-maven-plugin</artifactId>
<executions>
<execution>
<phase>package</phase>
<goals>
<goal>build</goal>
</goals>
</execution>
</executions>
<configuration>
<target>${project.basedir}/target</target>
<systemBundles>
<systemBundle>
<groupId>com.github.timofeevda.jstressy</groupId>
<artifactId>jstressy-logger</artifactId>
</systemBundle>
<systemBundle>
<groupId>org.apache.felix</groupId>
<artifactId>org.apache.felix.fileinstall</artifactId>
</systemBundle>
</systemBundles>
<scenarioBundles>
<scenarioBundle>
<groupId>com.github.timofeevda.jstressy</groupId>
<artifactId>jstressy-dummy-scenario-kotlin</artifactId>
</scenarioBundle>
</scenarioBundles>
</configuration>
</plugin>
JStressy Maven plugin creates folders structure with all required component needed to run Apache Felix with provided JStressy bundles. Almost everything can be customized, especially configuration folder and run scripts:
|--- bin/
| |--- felix.jar
|--- bundles/
| |--- system/ # basic bundles which are loaded first during Apache Felix start (logback initialization etc.)
| |--- application/ # all bundles representing custom JStressy framework build with all required dependencies
| |--- plugins/ # scenarios bundles, each bundles can be replaced in folder for hot reload
|--- config/
| |--- config.properties # Apache Felix config
| |--- logback.xml # basic logback config
|--- run.sh # basic run script
|--- stressy.yml # basic configuration example
Modelling arrivals
JStressy supports several ways to model arrivals: constant rate, ramping rate, interval rate, Poisson arrival, and their combinations.
Constant rate arrivals
stressPlan:
stages:
- name: ConstRate
....
arrivalRate: 1 # constant rate - 1 per second
...
Ramping rate arrivals
stressPlan:
stages:
- name: RampingRate
....
arrivalRate: 1 # constant rate - 1 time per second
rampArrival: 10 # target constant rate we need to get - 10 times per second
rampArrivalPeriod: 1m # how often we need to adjust arrival rate before getting the target one (1 time per minute)
rampDuration: 10m # how long it should take to achieve target arrival rate (10 times per second after 10 minutes of 10 adjustments)
...
Note: rampArrivalRate can be used instead of rampArrivalPeriod to describe how often we need to adjust arrival rate
Poisson arrivals
JStressy supports modeling of homogeneous Poisson arrivals by using constant arrival rate and poissonArrival option.
stressPlan:
stages:
- name: PoissonArrivals
....
arrivalRate: 1
poissonArrival: true # turns Poisson arrivals mode on
...
Note: by default, JStressy uses [0,1) random numbers interval to calculation the next Poisson arrival. poissonMaxRandom property can be used to set the random number maximum value (exclusive) to control the number of Poisson arrivals and the interval between Poisson arrival. Low max random values give you fewer arrivals with bigger intervals between each arrival.
JStressy also supports modeling of non-homogeneous Poisson arrivals by using ramping arrival rate and poissonArrival option. In this mode, the next arrival time will be determined based on the current arrival rate.
stressPlan:
stages:
- name: NonHomogeneousPoissonArrivals
....
arrivalRate: 1
poissonArrival: true # turns Poisson arrivals mode on
rampArrival: 10 # target constant rate we need to get - 10 times per second
rampArrivalPeriod: 1m # how often we need to adjust arrival rate before getting the target one (1 time per minute)
rampDuration: 10m # how long it should take to achieve target arrival rate (10 timer per second after 10 minutes of 10 adjustments)
...
Interval arrivals
JStressy can combine different arrival types using arrivalIntervals property allowing to model different use cases and make scenario arrivals close to real life.
Below is the example of defining arrival rates for one stage. Several constant rates defined.
stressPlan:
stages:
- name: IntervalArrivals
....
arrivalIntervals:
- id: first
arrivalRate: 1
duration: 5m
- id: second
arrivalRate: 3
delay: 5m
duration: 5m
- id: third
arrivalRate: 1
delay: 10m
duration: 5m
...
Below is the example of mixing different arrivals type for one stage. Constant rate is mixed with Poisson arrivals and ramping rate in the end.
stressPlan:
stages:
- name: IntervalArrivals
....
arrivalIntervals:
- id: first
arrivalRate: 0.016
duration: 10m
- id: second
arrivalRate: 0.016
poissonArrival: true
delay: 10m
duration: 10m
- id: third
arrivalRate: 0.016
rampArrival: 1.0
rampArrivalPeriod: 1m
rampDuration: 10m
delay: 20m
duration: 10m
...
