Edit me

Local Monitoring with Kruize

Kruize Local provides services to generate recommendations for both single and bulk experiments.

  • For Bulk Services: For detailed instructions, refer this guide.
  • For Demo of Individual Experiments: Continue with the steps below.
  • For Advanced Local Monitoring Options: Explore advanced testing details here

Prerequisites

Ensure you have one of the clusters: kind, minikube, or openShift.

Getting Started with the Demo

To begin exploring local monitoring capabilities, follow these steps:

Run the Demo

Clone the demo repository:
git clone git@github.com:kruize/kruize-demos.git
Change directory to the local monitoring demo:
cd kruize-demos/monitoring/local_monitoring

Note : We support Kind, Minikube and Openshift clusters. By default, it runs on the Kind cluster.

Execute the demo script on kind as:
./local_monitoring_demo.sh
Execute the demo script in openshift as:
./local_monitoring_demo.sh -c openshift

Usage: ./local_monitoring_demo.sh [-s|-t] [-c cluster-type] [-f]
c = supports minikube, kind and openshift cluster-type
s = start (default), t = terminate
f = create fresh environment setup if cluster-type is minikube or kind

Understanding the Demo

This demo covers the steps to install Kruize, create an experiment, and generate recommendations.

  • By default, it creates an experiment for a container which is long running in a cluster and generates recommendations for the same.
  • If user creates an environment set-up for minikube/kind, benchmark ‘sysbench’ is installed and that container is used to create experiment and generate recommendations.

Using kruize UI

Refer this video on how to create experiments and generate recommendations!

Recommendations for different load Simulations observed on Openshift

TFB (TechEmpower Framework Benchmarks) benchmark is simulated in different load conditions and below are the different recommendations observed from Kruize-Autotune.

IDLE

  • Experiment: monitor_tfb-db_benchmark
    • Shows an IDLE scenario where CPU recommendations are not generated due to minimal CPU usage (less than a millicore). idle

Over Provision

  • Experiment: monitor_tfb_benchmark_multiple_import
    • Highlights over-provisioning where CPU recommendations are lower than the current CPU requests. This scenario also demonstrates over-provisioning in memory usage. overprovision

Under Provision

  • Experiment: monitor_tfb-db_benchmark_multiple_import
    • Illustrates under-provisioning where CPU recommendations exceed the current CPU requests, suggesting adjustments for improved efficiency. underprovision