You can run testing framework upon your testcase to get its performance model as the testing result.
- the performance test with the mocked endpoint will be performed in the github workflow after your code is merged to ADOT Collector repository.
- you need to run the performance test locally with your
real endpoint
and provide the result, this doc is a guideline about how to use the testing framework to get the performance model upon the real endpoint. - you can also follow this document to reproduce the performance test with mock endpoint by just skipping step 3.
please follow setup-basic-components-in-aws-account
- Checkout ADOT Collector in the same folder of the testing framework if you haven't done that.
git clone [email protected]:aws-observability/aws-otel-collector.git
- Build RPM which will be used to install ADOT Collector in the performance test.
cd aws-otel-collector && make package-rpm
skip it if you just want to reproduce the performance test with mock endpoint.
In this step, you need to modify the otconfig which is defined in your testcase folder aws-otel-test-framework/terraform/testcases/{{testcase name}}/otconfig.tpl
, so that the exporter sends data to the real endpoint.
- remove the mocked endpoint config.
- add any required config to your exporter, for example, the apikey or credentials.
You are going to run three rounds of the test, each takes one hour, performance model is presented as a file output/performance.json
after each round.
- Run on rate 100 tps
cd aws-otel-test-framework/terraform/performance
terraform init
terraform apply -var="data_rate=100" -var="testcase=../testcases/{{testcase name}}" -var="install_package_source=local" -var-file="../testcases/{{testcase name}}/parameters.tfvars"
terraform destroy
cd output
cat performance.json
- Run on rate 1000 tps
cd aws-otel-test-framework/terraform/performance
terraform init
terraform apply -var="data_rate=1000" -var="testcase=../testcases/{{testcase name}}" -var="install_package_source=local" -var-file="../testcases/{{testcase name}}/parameters.tfvars"
terraform destroy
cd output
cat performance.json
- Run on rate 5000 tps
cd aws-otel-test-framework/terraform/performance
terraform init
terraform apply -var="data_rate=5000" -var="testcase=../testcases/{{testcase name}}" -var="install_package_source=local" -var-file="../testcases/{{testcase name}}/parameters.tfvars"
terraform destroy
cd output
cat performance.json
- the performance model could be found under
aws-otel-test-framework/terraform/performance/output/performance.json
, and you can open an issue at ADOT Collector with the performance model file, so we can gather it into the performance model readme.
if you want to debug, add a parameter in the apply
command: -var="debug=true"
, for example
terraform apply -var="debug=true" -var="data_rate=1000" -var="testcase=../testcases/{{testcase name}}" -var="install_package_source=local" -var-file="../testcases/{{testcase name}}/parameters.tfvars"
to do so, a private key will be dumped to disk so that you can use it to login instances.
please note don't run terraform destroy
before you finish debugging, otherwise the resources will be clean up.
Basically, performance test launches two ec2 instance:
- An
collector instance
to install ADOT Collector, and CloudWatch Agent to collect the cpu/mem metrics. - An
sample-app instance
to launch the sample app container and the mocked server containers, which sends and receives data, if you are sending data to the real endpoint, you can ignore the mock server container.
cd aws-otel-test-framework/terraform/performance
chmod 400 private_key.pem
ssh -i private_key.pem ec2-user@`terraform output collector_instance`
-
the installed folder:
/opt/aws/aws-otel-collector
-
the logs:
/opt/aws/aws-otel-collector/logs
cd aws-otel-test-framework/terraform/performance
chmod 400 private_key.pem
ssh -i private_key.pem ec2-user@`terraform output sample_app_instance`
sudo docker ps
all the metrics will be showing on the CloudWatch Console in your account. The Namespace is AWSOtelCollector/PerfTest
port 1777 on the collector instance is opened for golang profiling. For example
curl http://127.0.0.1:1777/debug/pprof/heap > /tmp/perf.out