中文

scheduler-stress-test

This is a stress testing tool for the scheduler in a large-scale scenario.

Preparatory Work

To simulate a large-scale scheduling scenario, you can create a large number of nodes as needed using kwok. The nodes you create may be in a NotReady state. In order to use the nodes to schedule pods, the "pod.yaml" file must add a toleration to tolerate all NoSchedule taints.

To prepare for this, you should do the following:

1. Install kwok on your k8s cluster. Refer to https://kwok.sigs.k8s.io/docs/user/kwok-in-cluster/.
2. Create fake nodes on your k8s cluster.

fake node example:

cat << EOF > node.yaml 
apiVersion: v1
kind: Node
metadata:
  annotations:
    node.alpha.kubernetes.io/ttl: "0"
    kwok.x-k8s.io/node: fake
  labels:
    beta.kubernetes.io/arch: amd64
    beta.kubernetes.io/os: linux
    kubernetes.io/arch: amd64
    kubernetes.io/hostname: {NODE_NAME}
    kubernetes.io/os: linux
    kubernetes.io/role: agent
    node-role.kubernetes.io/agent: ""
    type: kwok
  name: {NODE_NAME}
spec:
  taints: # Avoid scheduling actual running pods to fake Node
    - effect: NoSchedule
      key: kwok.x-k8s.io/node
      value: fake
status:
  allocatable:
    cpu: "64"
    ephemeral-storage: 1Ti
    hugepages-1Gi: "0"
    hugepages-2Mi: "0"
    memory: 250Gi
    pods: "110"
  capacity:
    cpu: "64"
    ephemeral-storage: 1Ti
    hugepages-1Gi: "0"
    hugepages-2Mi: "0"
    memory: 250Gi
    pods: "128"
  nodeInfo:
    architecture: amd64
    bootID: ""
    containerRuntimeVersion: ""
    kernelVersion: ""
    kubeProxyVersion: fake
    kubeletVersion: fake
    machineID: ""
    operatingSystem: linux
    osImage: ""
    systemUUID: ""
  phase: Running
EOF

# create nodes as you needed
for i in {0..99}; do sed "s/{NODE_NAME}/kwok-node-$i/g" node.yaml | kubectl apply -f -; done

How to use

Download the code and build it:

git clone https://github.com/k-cloud-labs/scheduler-stress-test.git
make build

Two commands are supported: create and wait.

The create command creates a specified number of pods from a template file in the k8s cluster with a specified level of concurrency.

The wait command waits for all the pods created above to be scheduled and continuously prints the results.

Example:

# create 1000 pod by 1000 concurrency (namespace: scheduler-stress-test)
sst create --kubeconfig=/root/.kube/config --count 1000 --concurrency 1000 --pod-template=pod.yaml
# wait result
sst wait --kubeconfig=/root/.kube/config --namespace=scheduler-stress-test

The example above uses "pod.yaml" as a template and creates 1000 pods in the "scheduler-stress-test" namespace of the k8s cluster. It then waits and continuously prints the results. You can modify the "pod.yaml" file as needed.