seifrajhi / k8s-random-deploys

Deploy example workloads with random requests and limits

Random Kubernetes Deployments

This is a set of scripts to create Kubernetes deployments that have random requests and limits. It allows you to provide a deployment template and then create workloads based on it. These scripts were designed to demo karpenter but you can use them by providing your own templates.

These tools were used to create demos like the one found here

Scripts

create-workloads.sh

create-workloads.sh creates deployments

example

# create 1000 pods using deployments 
# with 250 replicas each
# use fast.template.yaml

create-workloads.sh \
  -t 1000 \
  -b 250 \
  -f fast-template.yaml

delete-workloads.sh

delete-workloads.sh deletes deployments This will delete all deployments. Use with caution.

Optional arguments are passed to kubectl

examples

delete-workloads.sh

delete workloads in a namespace

delete-workloads.sh -n default

scale-deployments.sh

scale-deployments.sh can randomly scale deployments up and down between 1-10 percent of existing deployment replicas. It will scale all deployments with an optional sleep between scaling activities. You can also set a scaling direction (up or down) to scale randomly in a single direction.

example

scale-deployments.sh

roll-deployments.sh

roll-deployments.sh will roll new versions of all the pods by changing metadata from the deployment without changing the requests or limits. Rolls all deployments in the requested namespace. Use with caution.

example

./roll-deployments.sh -n my-namespace

Templates

The templates folder has various templates for deployments using host and AZ spread, GPU workloads, and other options.

For more Karpenter examples see the karpenter examples.