Skip to content

Latest commit

 

History

History
87 lines (55 loc) · 3.29 KB

README.md

File metadata and controls

87 lines (55 loc) · 3.29 KB
Raw

Cerberus Gatling Performance Tests

Gatling based performance tests

Parameters

Each simulation requires different parameters.

Parameters for IamPrincipalAuthAndReadSimulation

These properties can be supplied to the simulation via environment variables or system properties

Parameter Description
CERBERUS_API_URL This controls the api that will be perfromance tested
CERBERUS_ACCOUNT_ID The account id is needed when creating iam roles for the test and granting kms decrypt for the cerberus account
REGION The region to auth with Cerberus and use KMS in
NUMBER_OF_SERVICES_FOR_SIMULATION This is the number of SDBs with random data will be created for the simulation. Each simulated user will be fed one of these services randomly to be.
CREATE_IAM_ROLES This defaults to false. Setting this to true creates a new iam role that will get deleted at the end of the simulation for each NUMBER_OF_SERVICES_FOR_SIMULATION WARNING: This creates a KMS key for each IAM role that does not get cleaned up automatically. Setting this to false makes each simulated service use the role that is running the tests.
PEAK_USERS The peak number of simulated concurrent users for the test
RAMP_UP_TIME_IN_MINUTES The amount of time to ramp up from peak users to 0 users.
HOLD_TIME_AFTER_PEAK_IN_MINUTES The amount of minutes to hold the peak users for

E.g. in a Bash terminal

   # Set required params
   export CERBERUS_API_URL="https://dev.cerberus-oss.io/"
   export CERBERUS_ACCOUNT_ID=1234567890
   export REGION="us-east-1"
   export NUMBER_LOAD_SERVERS=1 # <- numbers of instances in cluster performing gatling tests
   
   # Set optional parameters
   export NUMBER_OF_SERVICES_FOR_SIMULATION=1
   export CREATE_IAM_ROLES=false
   export PEAK_USERS=1
   export RAMP_UP_TIME_IN_MINUTES=1
   export HOLD_TIME_AFTER_PEAK_IN_MINUTES=1

Running Locally

First export required parameters and AWS credentials.

Use the following gradlew task to run the default simulation:

./gradlew clean cerberus-gatling-perf-tests:runSimulation

To specify the simulation name:

./gradlew cerberus-gatling-perf-tests:runSimulation -Psimulation=IamPrincipalAuthAndReadSimulation

To determine your current account id and role use:

aws sts get-caller-identity

Running the Fat Jar

The following gradle task can create a fat jar containing the tests

./gradlew clean cerberus-gatling-perf-tests:gatlingCompileSimulationFatJar

You can trigger the tests via

java -cp PATH/TO/JAR io.gatling.app.Gatling --simulation SIMULATION NAME -rf PATH/TO/SAVE/REPORT

Additional Gatling configuration parameters are available, e.g.

-Dgatling.http.ahc.connectTimeout=60000 -Dgatling.http.ahc.handshakeTimeout=60000 -Dgatling.http.ahc.requestTimeout=120000 -Dgatling.http.ahc.keepAlive=false

See the Gatling docs for more information

Troubleshooting

MacOS Open File Limits

You may hit open file limits running these tests.

View limits in MacOS

sysctl kern.maxfiles
sysctl kern.maxfilesperproc

Change Limits in MacOS

sudo sysctl -w kern.maxfiles=300000
sudo sysctl -w kern.maxfilesperproc=300000