KnativeCloud Scheduler Source CRD.

Overview

This repository implements an Event Source for Knative Eventing defined with a CustomResourceDefinition (CRD). This Event Source represents Google Cloud Scheduler. Point is to demonstrate an Event Source that does not live in the Knative Eventing Sources that can be independently maintained, deployed and so forth.

This particular example demonstrates how to perform basic operations such as:

• Create a Cloud Scheduler Job when a Source is created
• Delete a Job when that Source is deleted
• Update a Job when the Source spec changes

Details

Actual implementation contacts the Cloud Scheduler API and creates a Job as specified in the CloudSechedulerSource CRD Spec. Upon success a Knative service is created to receive calls from the Cloud Scheduler and will then forward them to the Channel.

Purpose

Provide an Event Source that allows subscribing to Cloud Scheduler and processing them in Knative.

Another purpose is to serve as an example of how to build an Event Source using a [Warm Image[(https://github.com/mattmoor/warm-image) as a starting point.

Prerequisites

1. Create a Google Cloud project and install the gcloud CLI and run gcloud auth login. This sample will use a mix of gcloud and kubectl commands. The rest of the sample assumes that you've set the $PROJECT_ID environment variable to your Google Cloud project id, and also set your project ID as default using gcloud config set project $PROJECT_ID.

2. Setup Knative Serving

3. Configure static IP

4. Configure custom dns

5. Configure outbound network access

6. Setup Knative Eventing using the release.yaml file. This example does not require GCP.

Create a GCP Service Account and a corresponding secret in Kubernetes

1. Enable Google Cloud Scheduler API

   gcloud services enable cloudscheduler.googleapis.com

2. Create a GCP Service Account. This sample creates one service account for both registration and receiving messages, but you can also create a separate service account for receiving messages if you want additional privilege separation.

   1. Create a new service account named csr-source with the following command:

      gcloud iam service-accounts create csr-source

   2. Give that Service Account the Editor' role on your GCP project:

      gcloud projects add-iam-policy-binding $PROJECT_ID \
        --member=serviceAccount:csr-source@$PROJECT_ID.iam.gserviceaccount.com \
        --role roles/cloudscheduler.admin

   3. Download a new JSON private key for that Service Account. Be sure not to check this key into source control!

      gcloud iam service-accounts keys create csr-source.json \
        --iam-account=csr-source@$PROJECT_ID.iam.gserviceaccount.com

   4. Create a namespace for where the secret is created and where our controller will run

      kubectl create namespace cloudschedulersource-system

   5. Create a secret on the kubernetes cluster for the downloaded key. You need to store this key in key.json in a secret named gcppubsub-source-key

      kubectl -n cloudschedulersource-system create secret generic cloudschedulersource-key --from-file=key.json=csr-source.json

      The name cloudschedulersource-key and key.json are pre-configured values in the controller which manages your Cloud Scheduler sources.

Install Cloud Scheduler Source

kubectl apply -f https://raw.githubusercontent.com/vaikas-google/csr/master/release.yaml

Inspect the Cloud Scheduler Source

First list the available sources, you might have others available to you, but this is the one we'll be using in this example

 kubectl get crds -l "eventing.knative.dev/source=true"

You should see something like this:

vaikas@penguin:~/projects/go/src/github.com/vaikas-google/csr$
NAME                                      AGE
cloudschedulersources.sources.aikas.org   29d

you can then get more details about it, for example what are the available configuration options for it:

kubectl get crds cloudschedulersources.sources.aikas.org -oyaml

And in particular the Spec section is of interest:

 validation:
    openAPIV3Schema:
      properties:
        apiVersion:
          type: string
        kind:
          type: string
        metadata:
          type: object
        spec:
          properties:
            body:
              description: Optional body to send in the event
              type: string
            googleCloudProject:
              description: Google Cloud Project ID to create the scheduler job in.
              type: string
            httpMethod:
              description: Optional HTTP method to use when delivering the event.
                If omitted, uses POST
              type: string
            location:
              description: 'Google Cloud Platform region to create the scheduler job
                in. For example: us-central1.'
              type: string
            schedule:
              description: 'Schedule in cron format. For example: ''* * * * *'' (once
                a minute), or human readable: ''every 1 mins'''
              type: string
            serviceAccountName:
              description: Service Account to run Receive Adapter as. If omitted,
                uses 'default'.
              type: string
            sink:
              type: object
            timezone:
              description: Optional timezone of the schedule. If omitted, uses UTC.
              type: string
          required:
          - googleCloudProject
          - location
          - schedule

Create a Knative Service that will be invoked for each Scheduler job invocation

To verify the Cloud Scheduler is working, we will create a simple Knative Service that dumps incoming messages to its log. The service.yaml file defines this basic service. Image might be different if a new version has been released.

apiVersion: serving.knative.dev/v1alpha1
kind: Service
metadata:
  name: github-message-dumper
spec:
  runLatest:
    configuration:
      revisionTemplate:
        spec:
          container:
            image: us.gcr.io/probable-summer-223122/eventdumper-833f921e52f6ce76eb11f89bbfcea1df@sha256:7edb9fc190dcf350f4c49c48d3ff2bf71de836ff3dc32b1d5082fd13f90edee3

Enter the following command to create the service from service.yaml:

kubectl --namespace default apply -f https://raw.githubusercontent.com/vaikas-google/csr/master/service.yaml

Configure a Cloud Scheduler Source to send events directly to the service

The simplest way to consume events is to wire the Source directly into the consuming function. The logical picture looks like this:

Wire Cloud Scheduler Events to the function

Create a Cloud Scheduler instance targeting your function with the following:

curl https://raw.githubusercontent.com/vaikas-google/csr/master/one-to-one-csr.yaml | \
sed "s/MY_GCP_PROJECT/$PROJECT_ID/g" | kubectl apply -f -

Check that the Cloud Scheduler Source was created

kubectl get cloudschedulersources

And you should see something like this:

vaikas@penguin:~/projects/go/src/github.com/vaikas-google/csr$ kubectl get cloudschedulersources
NAME             AGE
scheduler-test   1m

Check that the Cloud Scheduler Job was created

gcloud beta scheduler jobs list

You should see something like this:

vaikas@penguin:~/projects/go/src/github.com/vaikas-google/csr$ gcloud beta scheduler jobs list
ID             LOCATION     SCHEDULE (TZ)       TARGET_TYPE  STATE
filter-source  us-central1  every 1 mins (UTC)  HTTP         ENABLED

Then wait a couple of minutes and you should see events in your message dumper.

Check that Cloud Scheduler invoked the function

Note this might take couple of minutes after the creation while the Cloud Scheduler gets going

kubectl -l 'serving.knative.dev/service=message-dumper' logs -c user-container

And you should see an entry like this there

2018/12/20 00:23:00 Received Cloud Event Context as: {CloudEventsVersion:0.1 EventID:2cd5d2ed-d2d1-94a1-bee7-d542d7ab834e EventTime:2018-12-20 00:23:00.498638175 +0000 UTC EventType:GoogleCloudScheduler EventTypeVersion: SchemaURL: ContentType:application/json Source:GCPCloudScheduler Extensions:map[]}
2018/12/20 00:23:00 Received event data as: {"data": "test does this work"}

Where the first line is displaying the Cloud Events Context and the second line is the actual data line.

Uninstall

kubectl delete cloudschedulersources scheduler-test
kubectl delete services.serving message-dumper

Check that the Cloud Scheduler Job was deleted

gcloud beta scheduler jobs list

More complex examples

• Multiple functions working together

Usage

Specification

The specification for a scheduler job looks like:

apiVersion: sources.aikas.org/v1alpha1
kind: CloudSchedulerSource
metadata:
  name: scheduler-test
spec:
  googleCloudProject: quantum-reducer-434
  location: us-central1
  schedule: "every 1 mins"
  body: "{test does this work}"
  sink:
    apiVersion: eventing.knative.dev/v1alpha1
    kind: Channel
    name: scheduler-demo

Creation

With the above in foo.yaml, you would create the Cloud Scheduler Job with:

kubectl create -f foo.yaml

Listing

You can see what Cloud Scheduler Jobs have been created:

$ kubectl get cloudschedulersources
NAME             AGE
scheduler-test   4m

Updating

You can upgrade foo.yaml jobs by updating the spec. For example, say you wanted to change the above job to send a different body, you'd update the foo.yaml from above like so:

apiVersion: sources.aikas.org/v1alpha1
kind: CloudSchedulerSource
metadata:
  name: scheduler-test
spec:
  googleCloudProject: quantum-reducer-434
  location: us-central1
  schedule: "every 1 mins"
  body: "{test does this work, hopefully this does too}"
  sink:
    apiVersion: eventing.knative.dev/v1alpha1
    kind: Channel
    name: scheduler-demo

And then update the spec.

kubectl replace -f foo.yaml

Of course you can also do this in place by using:

kubectl edit cloudschedulersources scheduler-test

And on the next run (or so) the body send to your function will by changed to '{test does this work, hopefully this does too}' instead of '{test does this work}' like before.

Removing

You can remove a Cloud Scheduler jobs via:

kubectl delete cloudschedulersources scheduler-test