Dynamic Scheduler

The dynamic scheduler targets an application class of pods with the following characteristics:

• no volumes attached
• stateless
• variable workloads
• long-lived containers

This repository is organized as follows:

customapp

This folders stores the assets responsible for generating the customapp application, which is responsible for play the customer container role in the Kubernetes environment.

experiments

In this folder we have the Dynamic Scheduler core, the testing scenarios and all the artifacts required to run the experiments related to this work.

pods

Here we have templates to create the pods into the Kubernetes cluster.

results

All the experiments results such as logs, diagrams and graphs are stored in this folder.

Cluster setup

Create Kubernetes cluster

minikube start --nodes 4 -p ppgcc --cpus 2 --memory 4096 --disk-size 20G --vm --kubernetes-version v1.25.3

Enable Metrics API

minikube addons enable metrics-server -p ppgcc

Create "lab" namespace

kubectl create namespace lab

Useful commands

Remove deployments from "lab" namespace

kubectl --namespace lab delete deployments kubectl get deployments --namespace lab --no-headers -o custom-columns=":metadata.name"

Remove pods from "lab" namespace

kubectl --namespace lab delete pod kubectl get pods --namespace lab --no-headers -o custom-columns=":metadata.name"

Purge "ppgcc" minikube profile

minikube delete -p ppgcc

List minikube services

minikube service list -p ppgcc

List pods info at "lab" namespace

kubectl get pods --namespace lab -o wide

Show pods resources at "lab" namespace

kubectl top node --namespace lab

Show nodes through Metrics API

kubectl get --raw /apis/metrics.k8s.io/v1beta1/nodes

./results_X.py > results/results.csv; ./graphics_generator.py; git add results/.txt; git add results/.csv; git add results/.html; git add results/.svg; git commit -m "update"; git push

Access tesla machine

ssh [email protected]

for i in .svg;do rsvg-convert -f pdf -o ${i%.}.pdf $i;done