Storage Benchmark for K8s

Prerequisites

Access to cluster
kubestr
Python 3.8
- matplotlib for rendering plots of results

Usage

Call bench.py -s mystorageclass to run full set of benchmarks for StorageClass mystorageclass. Check bench.py -h for available options. The benchmark script will save the results into JSON file results_%Y_%m_%d_%H%M%S.json (Python strftime format).

You can visualize the results by running ./render.py <resultfile>.json. This command will produce a PDF file with plots for the benchmark results.

To extract statistical information from a results file, you can run ./data.py <results.json>. This command prints the same statistical information which is printed during the benchmark run.

Local setup

We recommend that you setup a virtualenv to run the scripts locally.

git clone https://github.com/vshn/k8s-storage-bench.git
cd k8s-storage-bench
python3.8 -m venv .virtualenv
. .virtualenv/bin/activate
pip install -r requirements.txt
# Show help
./bench.py -h
# Run all benchmarks for storage class mystorageclass
./bench.py -s mystorageclass

Container image and deploy manifests

The container image is available on quay.io/vshn/k8s-storage-bench.

Building the container image

The included Dockerfile can be used to build a container image based on python:3.8-slim.

Running the benchmark as a K8s job

Adjust deploy/job.yaml to customize the benchmark configuration. If you want to run the benchmark in a namespace other than default, adjust the RoleBinding in deploy/rbac.yaml accordingly.

Then you can execute the provided run.sh to run the benchmark with the default settings.