The purpose of this project is to test the scalability and performance of k4e project using JMeter.
The test plan for testing k4e is using the Parallel Sampling controller that needs to be installed. Instructions for installing plugins can be found on here.
Before the test can be run, the test plan needs to be prepared. The following parameters need to be set:
- K8S_BEARER_TOKEN - the token is used to access the K8s API (see details below)
- HTTP_SERVER - K4E Server address
- HTTP_SERVER_PORT - K4E Server port (default: 8888)
- OCP_API_SERVER - K8S API server address
- OCP_API_SERVER_PORT - K8S API server port (default: 8443)
- NAMESPACE - K4E namespaces aren't supported yet, so the test will be run in the default namespace
After the test plan is prepared, the test can be run:
JVM_ARGS="-Xms1g -Xmx4g -XX:MaxMetaspaceSize=512m" $JMETER_HOME/bin/jmeter.sh -n -t ./test_plans/k4e_test_plan.jmx -l results.csv -eThe basic test plan runs the following scenario:
- Register an edge device
- Label the edge device
- Creates 10 edge deployments for the device
- In Loop:
- Sends heartbeats to the server
- Get updates from the server
For testing purposes, we need to create a token for accessing K8S cluster via RESTFUL API. Once the k4e-operator is deployed on the cluster, the following can be run to create a privileged token:
# Create a service account
kubectl apply -f - <<EOF
apiVersion: v1
kind: ServiceAccount
metadata:
name: k4e-scale-test
EOF
# Attach the service account to a privileged role
kubectl create clusterrolebinding k4e-scale-test-cluster-admin --clusterrole=cluster-admin --serviceaccount=default:k4e-scale-test
# Get the token
kubectl get secret $(kubectl get serviceaccount k4e-scale-test -o json | jq -r '.secrets[].name') -o yaml | grep " token:" | awk {'print $2'} | base64 -d