Skip to content

Latest commit

 

History

History
106 lines (79 loc) · 3.49 KB

README.md

File metadata and controls

106 lines (79 loc) · 3.49 KB
Raw

ghz

ghz is a gRPC benchmarking and load testing tool.

Kangal supports ghz as a loadtest backend using a custom docker image: hellofresh/kangal-ghz published to dockerhub.

For more information, please check ghz official site.

Usage

To create a loadtest, simply send a request to Kangal proxy with the ghz configuration in the testFile field:

$ curl -X POST http://${KANGAL_PROXY_ADDRESS}/load-test \
  -H 'Content-Type: multipart/form-data' \
  -F distributedPods=1 \
  -F [email protected] \
  -F type=Ghz \
  -F targetURL=http://my-app.example.com

Example config.json:

{
  "call": "helloworld.Greeter.SayHello",
  "total": 2000,
  "concurrency": 50,
  "data": {
    "name": "Joe"
  },
  "metadata": {
    "foo": "bar",
    "trace_id": "{{.RequestNumber}}",
    "timestamp": "{{.TimestampUnix}}"
  },
  "max-duration": "10s",
  "host": "0.0.0.0:50051"
}

Providing a protobuf schema

While ghz accepts .proto files to not depend on server reflection, Kangal currently only supports .protoset files.

To do so, use the testData form field to provide the .protoset file:

$ curl -X POST http://${KANGAL_PROXY_ADDRESS}/load-test \
  -H 'Content-Type: multipart/form-data' \
  -F distributedPods=1 \
  -F [email protected] \
  -F testData=@/path/to/testdata.protoset \
  -F type=Ghz \
  -F targetURL=http://my-app.example.com

and add the following key to your JSON/TOML configuration file:

{
  "protoset": "/data/testdata.protoset",
  ...
}

For information about how to create .protoset files and the complete list of configuration parameter, please check the ghz documentation.

Since ghz does not use the master-worker pattern, distributedPods simply creates replicas of the load-generating pod.
This means that a distributedPods value of 5 would mean that it creates 5 identical pods, generating 5x the load with 5x concurrency, etc.

Configuring resource limits and requirements

By default, Kangal does not specify resource requirements for loadtests run with ghz backend.

You can specify resource limits and requests for the containers to ensure that when the load generator runs, it has sufficient resources and will not fail before the service does.

The following environment variables can be specified to configure this parameter:

GHZ_CPU_LIMITS
GHZ_CPU_REQUESTS
GHZ_MEMORY_LIMITS
GHZ_MEMORY_REQUESTS

You have to specify these variables on Kangal Controller, read more at charts/kangal/README.md.

More information regarding resource limits and requests can be found in the following pages:

Notes

Kangal overrides the following options:

  • The output format is always set to HTML
  • Output directory is always set to /results
  • This is done so Kangal is able to pick up the results and persist the results
  • Because they are set as container arguments, this cannot be overridden with the configuration file