Experience the Serverless application programming model in cloud native scenarios

background

Alibaba Cloud Serverless Kubernetes (ASK) is a serverless Kubernetes container service launched by Alibaba Cloud. It is based on ECI (Elastic Container Instance) and allows you to directly create secure and isolated container applications without purchasing ECS ​​nodes. ASK has passed the Kubernetes consistency test and provides you with a fully compatible experience with community Kubernetes.

Knative is an open source Serverless application orchestration framework based on Kubernetes. Its goal is to develop cloud-native, cross-platform Serverless application orchestration standards. Alibaba Cloud Knative is based on ASK. While being fully compatible with the community Knaitve, it provides unified application orchestration for FC and ECI workloads, supports event-driven, automatic elasticity, and provides you with a unified Serverless application programming model.

Architecture

Next, we will introduce it through a bullet screen service demo. The demo mainly includes three parts: HomePage, event drive, and message processing.

HomePage is mainly used to send and receive bullet comments. Event driving is used to receive events and filter and transfer them. Message processing is used to process bullet comment messages. HomePage and message processing are deployed to FC and ECI respectively through Knative Serving, and event driving is deployed to ECI through Knative Eventing.

The main process of the barrage service demo is shown in the figure. The user sends a barrage message to the HomePage through the front end. The HomePage then sends the barrage to Kafka. The event-driven system receives the barrage message and routes it to the message processing for processing. After the barrage is processed, the barrage result is sent to the table storage. Finally, the front end obtains the barrage result and displays it on the page.

Next, we start to deploy the barrage service demo. The operations include the following:

First deploy message processing, then deploy event-driven, then deploy HomePage, and after deployment is complete, access the bullet screen service

Step 1: Deploy message processing

This service is used to receive event-driven bullet message requests, automatically scale up and down based on the number of requests, and send the results to Table Storage after the bullet message processing is completed. Before deployment, we first confirm that there is no workload so that we can observe the results after deployment.

Select ask Cluster In the left navigation bar of the cluster management page, select Workload > Stateless . Select the default namespace and confirm that there is currently no workload.

Next, we deploy the bullet message processing to the ECI workload through Knative. Here we deploy it through YAML. The YAML content is as follows:

 apiVersion: serving.knative.dev/v1kind: Servicemetadata: name: test-barrage-processspec: template: metadata: annotations: autoscaling.knative.dev/maxScale: "100" autoscaling.knative.dev/minScale: "0" k8s.aliyun.com/eci-image-snapshot-id: imc-uf636kjjx8xr4e75npex labels: danmu.role: "manager" spec: containerConcurrency: 2 serviceAccountName: barrage-install-sa containers: - args: - /manager env: - name: OTS_ENDPOINT value: https://barrage.cn-hangzhou.tablestore.aliyuncs.com - name: TABLE_NAME value: barrage - name: OTS_INSTANCENAME value: barrage - name: OTS_KEYID value: xxx - name: OTS_SECRET value: xxx - name: POD_NAME valueFrom: fieldRef: fieldPath: metadata.name - name: ROLE value: manager - name: POD_NAMESPACE valueFrom: fieldRef: fieldPath: metadata.namespace - name: TRACE_NAME value: "process" - name: PARENT_SPAN value: "barrage-sender" - name: SUB_SPAN value: "process" - name: TRACING value: "http://tracing-analysis-dc-sh.aliyuncs.com/adapt_g2it2kg78n@5cf06035aec2eb9_g2it2kg78n@53df7ad2afe8301/api/traces" image: registry-vpc.cn-shanghai.aliyuncs.com/knative-sample/barrage-manager:forrester-yuanyi_4cd77c84-20210618215458 name: user-container ports: - containerPort: 8000 name: http1

Main parameter description:

minScale and maxScale: Indicates the minimum and maximum number of Pods configured for the service
containerConcurrency: indicates the maximum number of concurrent requests for the configured Pod
OTS_ENDPOINT: indicates the configured table storage access address
TRACING: indicates the configured call link address. Next, we deploy the service.

In the left navigation bar of the cluster management page, select Applications > Knative .
In the upper right corner of the Service Management tab, click Create using template . Select the default namespace, paste the above YAML content into the template, and click Create.

Step 2: Deploy event-driven

Event-driven is used to receive events and filter and transfer event streams. Here we use Kafka event source as event-driven to receive bullet message from Kafka and then route the bullet message to message processing. We deploy it in yaml mode. The yaml content is as follows:

Step 3: Deploy HomePage

This service is used to receive front-end bullet message, send bullet message to Kafka, and receive bullet results from table storage. After deployment through Knative function mode, services, functions, and custom domain names will be automatically created in FC. Before operation, we first confirm that there is no bullet service, function, or custom domain name in FC.

Log in to the FC console and select Region (Shanghai) from the top menu bar.

Open the service and function page and confirm that there is no bullet message service and function

In the left navigation bar, click Custom Domain Name and confirm that there is no domain name information.

Open the custom domain name page and confirm that there is no custom domain name

Next, we deploy HomePage to the FC workload through Knative. Here we deploy it through yaml. The yaml content is as follows:

 apiVersion: serving.knative.dev/v1kind: Servicemetadata: name: demo-barrage annotations: workload.serving.knative.aliyun.com/class: "fc"spec: template: metadata: annotations: fc.revision.serving.knative.aliyun.com/code-space: "image" fc.revision.serving.knative.aliyun.com/role-arm: "acs:ram::xxxx:role/knative-fc" fc.revision.serving.knative.aliyun.com/domain: '{"domain":"barrage.demo.knative.top","path":"/*"}' spec: containers: - image: registry.cn-shanghai.aliyuncs.com/knative-sample/barrage-main:forrester-yuanyi_4cd77c84-20210618214527 env: - name: OTS_ENDPOINT value: https://barrage.cn-hangzhou.ots.aliyuncs.com - name: TABLE_NAME value: barrage - name: OTS_INSTANCENAME value: barrage - name: OTS_KEYID value: xxx - name: OTS_SECRET value: xxx - name: KAFKA_SERVER value: "106.15.11.179:9093,47.100.131.71:9093,47.102.44.91:9093" - name: KAFKA_USER value: "alikafka_pre-cn-xxx" - name: KAFKA_PWD value: "xxx" - name: KAFKA_TOPIC value: "barrage-info" - name: TRACING value: "http://tracing-analysis-dc-sh.aliyuncs.com/adapt_g2it2kg78n@5cf06035aec2eb9_g2it2kg78n@53df7ad2afe8301/api/traces" - name: TRACE_NAME1 value: "sender" - name: TRACE_NAME2 value: "receiver" - name: TRACE_NAME3 value: "result" - name: PARENT_SPAN value: "barrage-sender" - name: SUB_SPAN1 value: "sender" - name: SUB_SPAN2 value: "result"

Main parameter description:

The fc-related parameter configuration includes: deploying fc-type workloads, deploying through images, and specifying the access domain name as barrage.demo.knative.top
Configure the table storage access address: OTS_ENDPOINT
Kafka related parameter configuration: Kafka service address, bullet message topic
Call link address configuration: TRACING
So let's deploy the service.

Log in to the Container Service management console.
In the left navigation bar of the cluster management page, select Applications > Knative .
In the upper right corner of the Service Management tab, click Create using template . Select the default namespace, paste the above YAML content into the template, and click Create.
After the deployment is complete, we verify it in the Function Compute console.

Log in to the Function Compute console.
In the top menu bar, select Region.
In the left navigation bar, click Services and Functions . Select Services and Functions and you can see that the bullet screen service has been deployed. Open the Services and Functions page and you can see that the HomePage has been deployed.

Step 4: Service Access

The above services have been deployed. Next, we will access the services through a custom domain name. http://barrage.demo.knative.top

Next, we send the bullet message. Here you can customize the bullet message to be sent, the number of concurrent messages, and the duration. Here we use the default configuration to send.

Set the message, concurrency and duration, and click [Send]

We can see that barrage messages are constantly displayed.

summary

Alibaba Cloud Knative provides a unified programming model for containers and functions on top of Serverless Kubernetes, bringing you a unified Serverless application programming model. Interested students are welcome to communicate with us.