playground-serverless

Playing around with Kubernetes serverless frameworks (and more)

⚠️ This playground is not fully functional. It contains unresolved issues.

The main idea was to create a cron job triggered function sending multiple messages over NATS. Another function should receive these messages and scale down to zero when finished.
First I tried to play around with Kubeless, but it turned out it does not support scale-to-zero feature at the moment.
After this I tried OpenFaas which supports this, but it looks here (and in Kubeless) everything should be written as a Http handler, and there is no direct NATS triggers working with autoscale (the gateway translates NATS messages to Http requests).

cron-function

A Camel Quarkus based function using Kotlin and Gradle to generate and send multiple messages to a given NATS topic.

• run with ./gradlew quarkusDev (using quarkusPluginVersion=1.1.1.Final)
• build runnable jar with ./gradlew quarkusBuild
• build native GraalVM image with ./gradlew buildNative

Install GraalVM using SDKMAN!:

sdk install java 19.2.1-grl  
export GRAALVM_HOME=/home/zoli/.sdkman/candidates/java/19.2.1-grl   
$GRAALVM_HOME/bin/gu install native-image  

The native Docker image was used with the help of this repo https://github.com/pmlopes/openfaas-quarkus-native-template

The project runs well in standalone mode, but it throws the exception below whe it is deployed to OpenFaas:

2020/01/12 15:08:14 stderr: Caused by: org.apache.camel.ResolveEndpointFailedException: Failed to resolve endpoint: nats://nats.openfaas:4222?topic=nats-test due to: There are 1 parameters that couldn't be set on the endpoint. Check the uri if the parameters are spelt correctly and that they are properties of the endpoint. Unknown parameters=[{topic=nats-test}]

async-function

A Go based function receiving messages from a given NATS topic.

• download dependencies wit go get
• run with go run handler.go

Kubernetes

Create a Kubernetes cluster using k3d https://github.com/rancher/k3d

• install kubectl (1.17.0)

• install k3d (it takes a couple of minutes)

  curl https://raw.githubusercontent.com/rancher/k3d/master/install.sh | TAG=v1.4.0 bash

• create Kubernetes cluster

  k3d create
  export KUBECONFIG="$(k3d get-kubeconfig --name='k3s-default')"

  or create with published port k3d create --publish 8080:8080

• install Helm

  curl https://raw.githubusercontent.com/helm/helm/master/scripts/get-helm-3 | bash

OpenFaas

• install OpenFaas with Helm https://github.com/openfaas/faas-netes/tree/master/chart/openfaas

  helm repo add openfaas https://openfaas.github.io/faas-netes/
  helm repo update
  kubectl apply -f https://raw.githubusercontent.com/openfaas/faas-netes/master/namespaces.yml
  helm upgrade openfaas --install openfaas/openfaas --namespace openfaas --set basic_auth=false --set functionNamespace=openfaas-fn

• install CLI

  curl -sL https://cli.openfaas.com | sudo sh
  export OPENFAAS_URL=http://127.0.0.1:31112

  • if installed with auth

      kubectl port-forward svc/gateway -n openfaas 31112:8080 &
      faas-cli login -g $OPENFAAS_URL -u admin -p pass

• add Docker registry

  kubectl run registry --image=registry:latest --port=5000 --namespace openfaas
  kubectl expose deployment registry --namespace openfaas --type=LoadBalancer --port=5000 --target-port=5000

• list pods with kubectl get pods -A and forward ports for gateway and registry with kubectl -n openfaas port-forward <pod_name> <port_number> &

• build and deploy functions

  faas-cli build -f <function>.yml
  docker push localhost:5000/<function>:latest
  faas-cli deploy -f <function>.yml

• see logs with kubectl logs -f -n openfaas-fn <pod_name>

• remove function with faas-cli remove -f <function>.yml

Test NATS

see https://github.com/openfaas-incubator/nats-connector

• install NATS publisher

  faas-cli deploy --name publish-message --image openfaas/publish-message:latest --fprocess='./handler' --env nats_url=nats://nats.openfaas:4222
  faas-cli invoke publish-message <<< "test message"

• or portforward the NATS pod and publish a message using telnet

  telnet localhost 4222
  pub nats-test 5
  hello

Test auto-scaling

see https://github.com/openfaas/workshop

• install OpenFass with faas-idler dry-run disabled

• or set dry-run=false in faas-idler deployment

  kubectl edit deployments.apps -n openfaas faas-idler

• create and deploy any function (in /openfaas-test) with the labels below

  faas-cli new --lang=go hello-openfaas
  faas-cli build -f hello-openfaas.yml
  docker tag hello-openfaas:latest localhost:5000/hello-openfaas:latest
  docker push localhost:5000/hello-openfaas:latest
  faas-cli deploy -f hello-openfaas.yml --label "com.openfaas.scale.min=0" --label "com.openfaas.scale.zero=true" --label "faasIdler.inactivityDuration=1m"

  update the image in hello-openfaas.yml to localhost:5000/hello-openfaas:latest

• call function using dashboard on http://localhost:8080/ui/

• or by using faas-cli invoke hello-openfaas, type message and press Ctrl+d

The function should scale down after 1 minute of inactivity, and should scale back to 1 when it is invoked.

Kubeless

see https://kubeless.io/docs/quick-start/

in /kubeless

• create Kubernetes cluster and with NGINX ingress

  k3d create --publish 8080:8080
  export KUBECONFIG="$(k3d get-kubeconfig --name='k3s-default')"
  ...
  kubectl create deployment nginx --image=nginx
  kubectl apply -f ingress.yml

• install Kubeless

  export RELEASE=v1.0.5
  kubectl create ns kubeless
  kubectl create -f https://github.com/kubeless/kubeless/releases/download/$RELEASE/kubeless-$RELEASE.yaml

• install CLI

  export OS=$(uname -s| tr '[:upper:]' '[:lower:]')
  curl -OL https://github.com/kubeless/kubeless/releases/download/$RELEASE/kubeless_$OS-amd64.zip && \
    unzip kubeless_$OS-amd64.zip && \
    sudo mv bundles/kubeless_$OS-amd64/kubeless /usr/local/bin/

• install NATS operator https://github.com/nats-io/nats-operator

  kubectl apply -f https://raw.githubusercontent.com/nats-io/nats-operator/v0.6.0/deploy/00-prereqs.yaml
  kubectl apply -f https://raw.githubusercontent.com/nats-io/nats-operator/v0.6.0/deploy/10-deployment.yaml

• create NATS cluster kubectl apply -f nats-cluster.yml

• install NATS trigger controller kubectl apply -f nats-trigger-controller-v1.0.0.yaml

Test Kubeless

• create function

  kubeless function deploy hello --runtime python2.7 --from-file hello-kubeless.py --handler test.hello

• list functions using kubectl get function or kubeless function ls

• call function using kubeless function call hello --data 'Hello world!'

• delete function using kubeless delete function hello

Test NATS

• create NATS trigger

  kubeless trigger nats create hello-nats-trigger --function-selector created-by=kubeless,function=hello --trigger-topic test

• list triggers kubeless trigger nats list

• publish NATS message

  kubeless trigger nats publish --url nats://localhost:4222 --topic test --message "Hello World!"

• delete trigger kubeless trigger nats delete hello-nats-trigger

It is possible to create auto-scale with the command below, but it does not scale down to zero

kubeless autoscale create hello --min 0 --max 2 --metric qps --value 1

Clean-up

• delete Kubernetes cluster with k3d delete