Container Service for Kubernetes：ACK KEDA

Overview

For Kubernetes, Horizontal Pod Autoscaler (HPA) is the most commonly used solution for automatically scaling pods. HPA determines scaling strategies based on the differences between the resource usage and predefined thresholds. HPA is an easy-to-use tool that supports various resource metrics. However, it does not support real-time scaling. For example, HPA cannot scale resources when specific events are detected. To address this issue, you can install KEDA in the ACK console. ACK KEDA supports event-driven scaling that can be used in various event-driven scenarios, such as offline video and audio transcoding, event-driven jobs, and stream processing.

How KEDA works

Case study

A simple transcoding job is used in the following case. When a new job is received, data similar to the following example is inserted into MongoDB: {"type":"mp4","state":"waiting","createTimeStamp":"1610332940","fileName":"World and peace","endTimeStamp":"","uuid":"1fae72ff-3239-42f5-af97-04711d8007e8"}. ACK KEDA searches the database for data entries that meet the "state":"waiting" condition. Then, ACK KEDA creates pods to process the data entries. One pod is created for each data entry. After the transcoding is completed, the value of the state field changes from waiting to finished. After the job is completed, the metadata is automatically cleared to reduce the load on the API server. This allows you to manage jobs in a convenient manner.

Step 1: Deploy ACK KEDA

1. Log on to the ACK console.

2. In the left-side navigation pane of the ACK console, choose Marketplace > Marketplace.

3. On the Marketplace page, find and click ack-keda.

4. Click Deploy in the upper-right corner of the ack-keda page. In the Basic Information step, select a cluster, click Next, and then click OK.

   In the left-side navigation pane, click Clusters. On the Clusters page, select the cluster where ACK KEDA is deployed and click the cluster name or click Details in the Actions column. In the left-side navigation pane, choose Applications > Helm to view ACK KEDA.

Step 2: Add MongoDB as an event source

1. Deploy MongoDB.

   If you already deployed MongoDB, skip this step.

   Important

   The database is used only for testing purposes. Do not use the database in production environments.

   1. Create a YAML file named mongoDB.yaml and add the following code to the file:

      apiVersion: apps/v1
      kind: Deployment
      metadata:
        name: mongodb
      spec:
        replicas: 1
        selector:
          matchLabels:
            name: mongodb
        template:
          metadata:
            labels:
              name: mongodb
          spec:
            containers:
            - name: mongodb
              image: mongo:4.2.1
              imagePullPolicy: IfNotPresent
              ports:
              - containerPort: 27017
                name: mongodb
                protocol: TCP
      ---
      kind: Service
      apiVersion: v1
      metadata:
        name: mongodb-svc
      spec:
        type: ClusterIP
        ports:
        - name: mongodb
          port: 27017
          targetPort: 27017
          protocol: TCP
        selector:
          name: mongodb

   2. Deploy MongoDB to the mongodb namespace of the cluster.

       kubectl apply -f mongoDB.yaml -n mongodb

2. Log on to MongoDB and create a user account.

   1. Create a user.

      # Create a user
      kubectl exec -n mongodb mongodb-xxxxx -- mongo --eval 'db.createUser({ user:"test_user",pwd:"test_password",roles:[{ role:"readWrite", db: "test"}]})'

   2. Log on to MongoDB.

      # Complete authentication
      kubectl exec -n mongodb mongodb-xxxxx -- mongo --eval  'db.auth("test_user","test_password")'

   3. Create a collection.

      # Create a collection
      kubectl exec -n mongodb mongodb-xxxxx -- mongo --eval  'db.createCollection("test_collection")'

3. Deploy TriggerAuthentication and ScaledJob.

   ACK KEDA uses TriggerAuthentication to authenticate logon requests to event sources. In this case, MongoDB is added as an event source. ACK KEDA uses the secretTargetRef field in TriggerAuthentication to retrieve the connection information from the specified Secret and then uses the connection information to authenticate requests to MongDB.

   1. Create a YAML file named auth.yaml and add the following code to the file:

      apiVersion: keda.sh/v1alpha1
      kind: TriggerAuthentication
      metadata:
        name: mongodb-trigger
      spec:
        secretTargetRef:
          - parameter: connectionString
            name: mongodb-secret
            key: connect
      ---
      apiVersion: v1
      kind: Secret
      metadata:
        name: mongodb-secret
      type: Opaque
      data:
        connect: bW9uZ29kYjovL3Rlc3RfdXNlcjp0ZXN0X3Bhc3N3b3JkQG1vbmdvZGItc3ZjLm1vbmdvZGIuc3ZjLmNsdXN0ZXIubG9jYWw6MjcwMTcvdGVzdA==

   2. Deploy TriggerAuthentication to the mongodb-test namespace of the cluster.

      kubectl apply -f auth.yaml -n mongodb-test

4. Deploy ScaledJob.

   ScaledJob is used to define the job template and specify the database to be queried and the query expression. In the following example, a job is created to query the test_collection collection in the test database and transcode data entries that meet the {"type":"mp4","state":"waiting"} condition.

   1. Create a YAML file named scaledJob.yaml and add the following code to the file:

      apiVersion: keda.sh/v1alpha1
      kind: ScaledJob
      metadata:
        name: mongodb-job
      spec:
        jobTargetRef:
          # Job template
          template:
            spec:
              containers:
                - name: mongo-update
                  image: registry.cn-hangzhou.aliyuncs.com/carsnow/mongo-update:v6
                  args:
                    - --connectStr=mongodb://test_user:test_password@mongodb-svc.mongodb.svc.cluster.local:27017/test
                    - --dataBase=test
                    - --collection=test_collection
                  imagePullPolicy: IfNotPresent
              restartPolicy: Never
          backoffLimit: 1
        pollingInterval: 15
        maxReplicaCount: 5
        successfulJobsHistoryLimit: 0
        failedJobsHistoryLimit: 10
        triggers:
          - type: mongodb
            metadata:
              dbName: test                               # The database to be queried.
              collection: test_collection                # The collection to be queried.
              query: '{"type":"mp4","state":"waiting"}'  # Create a job to process each data entry whose type is mp4 and state is waiting.
              queryValue: "1"
            authenticationRef:
              name: mongodb-trigger

      query: Set a condition. When ACK KEDA detects data entries that meet the specified condition, jobs are started.

   2. Deploy ScaledJob to the mongodb-test namespace of the cluster.

      kubectl apply -f scaledJob.yaml -n mongodb-test

5. Insert five data entries to MongoDB.

   # Insert five data entries to MongoDB
   kubectl exec -n mongodb mongodb-xxxxx -- mongo --eval 'db.test_collection.insert([
     {"type":"mp4","state":"waiting","createTimeStamp":"1610352740","fileName":"My Love","endTimeStamp":"","uuid":"1gae72ff-3239-42f5-af97-04711d8007e8"},
     {"type":"mp4","state":"waiting","createTimeStamp":"1610350740","fileName":"Harker","endTimeStamp":"","uuid":"1gae72ff-3239-42f5-af97-04711d8007e8"},
     {"type":"mp4","state":"waiting","createTimeStamp":"1610152940","fileName":"The World","endTimeStamp":"","uuid":"1gae72ff-3239-42f5-af97-04711d87767e8"},
     {"type":"mp4","state":"waiting","createTimeStamp":"1610390740","fileName":"Mother","endTimeStamp":"","uuid":"1gae72ff-3239-42f5-af97-04799d8007e8"},
     {"type":"mp4","state":"waiting","createTimeStamp":"1610344740","fileName":"Jagger","endTimeStamp":"","uuid":"1gae72ff-3239-42f5-af97-04711d80099e8"},
     ])'

Step 3: Check whether ACK KEDA works as expected

# watch job
watch -n 1 kubectl get job -n mongodb-test

References

• Horizontal pod autoscaling based on the metrics of ApsaraMQ for RocketMQ

• Horizontal pod autoscaling based on the metrics of ApsaraMQ for RabbitMQ