Flux Operator Benchmark

Mean Time To Production benchmark for Flux Operator and the enterprise distribution.

Prerequisites

Running the benchmark requires an EKS Auto Mode cluster with a dedicated node pool for Flux:

• Karpenter NodePool
• Flux Operator configuration
• Flux Instance configuration

The benchmark measurements are taken with kube-prometheus-stack. The Grafana dashboards are available at:

• Monitoring configuration

The following CLI tools are required to run the benchmark:

• kubectl
• flux-operator

Running the Benchmark

Generate 1K HelmReleases and 1K Kustomizations in the benchmark namespace:

COUNT=1000 make benchmark

Rerunning the benchmark will trigger an update of all existing HelmReleases and Kustomizations.

To trigger an update of the first 50 HelmReleases and Kustomizations via OCIRepository semver change, run:

VER=6.9.1 make set-version

To generate load on the Kubernetes API server, create CRDs with:

CRD_COUNT=100 make crds

To remove the generated CRDs, run:

make clean-crds

To remove the benchmark resources, run:

make clean