K8s_Load_Metrics_Simulator

This repo includes the yaml files for simulating the pressure over the nodes this includes the metrics like CPU , memory etc

K8s_Load_Metrics_Simulator

Overview

K8s_Load_Metrics_Simulator is a collection of Kubernetes manifest files designed to simulate high resource usage (CPU, memory, disk latency) and trigger OOM (Out of Memory) kills for pods. This allows for testing node resilience, performance under stress, and validating autoscaling policies.

By default, the manifests are configured to simulate pressure on Azure Kubernetes Service (AKS) nodes with SKU Standard_D2as_v4 (2 vCPUs, 8 GiB RAM).
You can adjust the manifests to match your node specifications by modifying the parameters according to your Azure VM SKU.

---

Features

• CPU Pressure: Simulates high CPU utilization.
• Memory Pressure: Forces memory exhaustion on nodes.
• Disk Latency: Creates artificial I/O latency.
• Pod OOM Kill: Triggers pod termination by exceeding memory limits.

---

Getting Started

Prerequisites

• Kubernetes cluster (AKS preferred)
• kubectl configured for your cluster
• Node names or labels for targeting specific nodes

Installation

1. Clone the repository:

   git clone https://github.com/your-repo/K8s_Load_Metrics_Simulator.git
   cd K8s_Load_Metrics_Simulator
   
   
   

2. Modify the nodeSelector in the manifest files to point to your target node:

nodeSelector: kubernetes.io/hostname:

3. Apply the manifests to your cluster:

kubectl apply -f cpu-pressure.yaml kubectl apply -f memory-pressure.yaml kubectl apply -f disk-pressure.yaml kubectl apply -f pod-oom-kill.yaml Customization If you are targeting nodes with different SKUs, adjust the parameters in the manifest files to reflect the node capacity. Example:

Standard_D8s_v4 (8 vCPUs, 32 GiB RAM) Standard_E16_v3 (16 vCPUs, 128 GiB RAM) Azure SKU Load Approximation Table Use this table to estimate the resource load required for different Azure VM SKUs. Adjust the manifest requests and limits based on these values.

Azure SKU vCPUs Memory (GiB) Recommended CPU Pressure (millicores) Recommended Memory Pressure (Mi) Standard_D2as_v4 2 8 1800 6500 Standard_D4as_v4 4 16 3600 13000 Standard_D8as_v4 8 32 7200 26000 Standard_E16_v3 16 128 15000 105000 Standard_D32_v3 32 128 30000 110000

Troubleshooting

Pod Stuck in Pending: Check if there are enough resources available on the node. Use:

kubectl describe pod CrashLoopBackOff: If the pod keeps restarting, reduce the resource limits in the manifest or allocate more resources to the node. Contribution Feel free to submit pull requests to improve the manifest files or add new types of stress simulations.

License

MIT License

---

Pressure Load Calculator (Concept)

You can create a Python or Bash script to automate the calculation of pressure parameters based on node SKU. Here's a simple example in Python:

def calculate_pressure(vcpus, memory_gb):
    cpu_pressure = (vcpus * 900)  # 900 millicores per vCPU
    memory_pressure = (memory_gb * 800)  # 800 Mi per GiB of memory
    return cpu_pressure, memory_pressure

sku_data = {
    "Standard_D2as_v4": (2, 8),
    "Standard_D8as_v4": (8, 32),
    "Standard_E16_v3": (16, 128),
}

for sku, specs in sku_data.items():
    cpu, mem = calculate_pressure(*specs)
    print(f"{sku}: CPU {cpu}m | Memory {mem}Mi")