tencent cloud


Hosting Dubbo to TKE

Last updated: 2022-06-10 16:48:46


    This document describes how to host a Dubbo application to TKE.

    Strengths of hosting Dubbo applications to TKE

    • Improve the resource utilization.
    • Kubernetes is a natural fit for microservice architectures.
    • Improve the Ops efficiency and facilitate DevOps implementation.
    • Highly scalable Kubernetes makes it easy to dynamically scale applications.
    • TKE provides Kubernetes master management to ease Kubernetes cluster Ops and management.
    • TKE is integrated with other cloud-native products of Tencent Cloud to help you better use Tencent Cloud products.

    Best Practices

    The following describes how to host a Dubbo application to TKE by using the Q Cloud Book Mall (QCBM) project as an example.

    QCBM overview

    QCBM is an online bookstore demo project developed by using the microservice architecture and the Dubbo 2.7.8 framework. It is deployed and hosted on CODING. For more information, see here. QCBM contains the following microservices:

    Microservice Description
    QCBM-Front Frontend project developed through React, built and deployed based on the Nginx 1.19.8 Docker image.
    QCBM-Gateway API gateway that accepts HTTP requests from the frontend and converts them into Dubbo requests at the backend.
    User-Service Dubbo-based microservice, providing user registration, login, and authentication features.
    Favorites-Service Dubbo-based microservice, providing book favorites.
    Order-Service Dubbo-based microservice, providing order generation and query features.
    Store-Service Dubbo-based microservice, providing the book information storage feature.

    QCBM architecture and add-ons

    In the following best practice, applications deployed in CVM are containerized and hosted to TKE. In this use case, one VPC is used and divided into two subnets:

    • Subnet-Basic: Deployed with stateful basic services, including Dubbo's service registry Nacos, MySQL, and Redis.
    • Subnet-K8S: Deployed with QCBM application services, all of which are containerized and run in TKE.

    The VPC is divided as shown below:

    The network planning for the QCBM instance is as shown below:

    Network Planning Description
    Region/AZ Nanjing/Nanjing Zone 1
    Subnet-Basic Nanjing Zone 1, CIDR block:
    Subnet-K8S Nanjing Zone 1, CIDR block:
    Nacos cluster Nacos cluster built with three 1-core 2 GB MEM Standard SA2 CVM instances, with IP addresses of,, and

    The add-ons used in the QCBM instance are as shown below:

    Add-on Version Source Remarks
    k8s 1.8.4 Tencent Cloud TKE management mode
    MySQL 5.7 Tencent Cloud TencentDB for MySQL with two nodes
    Redis 5.0 Tencent Cloud TencentDB for Redis Standard Edition
    CLS N/A Tencent Cloud Log service
    TSW N/A Tencent Cloud Accessed with SkyWalking 8.4.0 Agent, which can be downloaded here
    Java 1.8 Open-source community Docker image of Java 8 JRE
    Nacos 2.0.0 Open-source community Download here
    Dubbo 2.7.8 Open-source community GitHub address



    Tencent Cloud Tencent Container Registry (TCR) are available in Personal Edition and Enterprise Edition as differentiated below:

    QCBM is a Dubbo containerized demo project, so TCR Personal Edition is perfectly suited to its needs. However, for enterprise users, TCR Enterprise Edition is recommended. To use an image repository, see Basic Image Repository Operations.


    Tencent Service Watcher (TSW) provides cloud-native service observability solutions that can trace upstream and downstream dependencies in distributed architectures, draw topologies, and provide multidimensional call observation by service, API, instance, and middleware. It is further described as shown below:

    TSW is architecturally divided into four modules:

    Data collection (client)


    You can use an open-source probe or SDK to collect data. If you are migrating to the cloud, you can change the reporting address and authentication information only and keep most of the configurations on the client.

    Data processing (server)


    Data is reported to the server via the Pulsar message queue, converted by the adapter into an OpenTracing-compatible format, and assigned to real-time and offline computing as needed.

    • Real-time computing provides real-time monitoring, statistical data display, and fast response to the connected alarming platform.
    • Offline computing aggregates the statistical data in large amounts over long periods of time and leverages big data analytics to provide business value.



    The storage layer can adapt to use cases with different data types, writing at the server layer, and query and reading requests at the data usage layer.

    Data usage


    The data usage layer provides underlying support for console operations, data display, and alarming.

    The architecture is as shown below:


    Building basic service cluster

    Building Docker image

    Writing Dockerfile

    The following uses user-service as an example to describe how to write a Dockerfile. The project directory structure of user-service is displayed, Dockerfile is in the root directory of the project, and user-service-1.0.0.zip is the packaged file that needs to be added to the image.

    ➜  user-service tree
    ├── Dockerfile
    ├── assembly
    │    ....
    ├── bin
    │    ....
    ├── pom.xml
    ├── src
    │    ....
    ├── target
    │    .....
    │   └── user-service-1.0.0.zip
    └── user-service.iml

    The Dockerfile of user-service is as shown below:

    FROM java:8-jre
    ARG APP_NAME=user-service
    # The working directory of the container is `/app`.
    WORKDIR /app
    # Add the locally packaged application to the image.
    COPY ./target/${FULL_APP_NAME}.zip .
    # Create the `logs` directory. Decompress and delete the original files and directory after the decompression.
    RUN mkdir logs \
        && unzip ${FULL_APP_NAME}.zip \
        && mv ${FULL_APP_NAME}/** . \
        && rm -rf ${FULL_APP_NAME}*
    # Start script and parameters of `user-service`
    ENTRYPOINT ["/app/bin/user-service.sh"] CMD ["start", "-t"]
    # Dubbo port number
    EXPOSE 20880

    • Java applications in the production environment have a lot of configuration parameters, making the start script complex. It's a heavy workload to write all the content of the start script to the Dockerfile, which is far less flexible than shell scripts and can't implement fast troubleshooting. We recommend you not enable the start script.
    • In general, nohup is used at the end of the start script to start the Java application, but the deamon process that comes along will cause the container to exit directly after execution. Therefore, you need to change nohup java ${OPTIONS} -jar user-service.jar > ${LOG_PATH} 2>&1 & to java ${OPTIONS} -jar user-service.jar > ${LOG_PATH} 2>&1.
    • As each Run command in the Dockerfile will generate an image layer, we recommend you combine these commands into one.

    Building image

    TCR provides both automatic and manual methods to build an image. For more information, see Image Building Overview. To demonstrate the build process, the manual method is used.

    The image name needs to be in line with the convention of ccr.ccs.tencentyun.com/[namespace]/[ImageName]:[image tag]:

    • Here, namespace can be the project name to facilitate image management and use. In this document, QCBM represents all the images under the QCBM project.
    • ImageName can contain the subpath, generally used for multi-project use cases of enterprise users. In addition, if a local image is already built, you can run the docker tag command to rename the image in line with the naming convention.
    1. Run the following command to build an image as shown below:

      # Recommended build method, which eliminates the need for secondary tagging operations
      sudo docker build -t ccr.ccs.tencentyun.com/[namespace]/[ImageName]:[image tag]
      # Build a local `user-service` image. The last `.` indicates that the Dockerfile is stored in the current directory (`user-service`).
      ➜  user-service docker build -t ccr.ccs.tencentyun.com/qcbm/user-service:1.0.0 .
      # Rename existing images in line with the naming convention
      sudo docker tag [ImageId] ccr.ccs.tencentyun.com/[namespace]/[ImageName]:[image tag]
    2. After the build is complete, you can run the following command to view all the images in your local repository.

      docker images

      A sample is as shown below:

    Uploading image to TCR

    Creating namespace

    The QCBM project uses TCR Personal Edition (TCR Enterprise Edition is recommended for enterprise users).

    1. Log in to the TKE console.
    2. Click TCR > Personal > Namespace to enter the Namespace page.
    3. Click Create and create the qcbm namespace in the pop-up window. All the images of the QCBM project are stored under this namespace as shown below:

    Uploading image

    Log in to TCR and upload an image.

    1. Run the following command to log in to TCR.

      docker login --username=[Tencent Cloud account ID] ccr.ccs.tencentyun.com
      • You can get your Tencent Cloud account ID on the Account Info page.
      • If you forget your TCR login password, you can reset it in My Images of TCR Personal Edition.
      • If you are prompted that you have no permission to run the command, add sudo before the command and run it as shown below. In this case, you need to enter two passwords, the server admin password required for sudo and the TCR login password.
        sudo docker login --username=[Tencent Cloud account ID] ccr.ccs.tencentyun.com
        As shown below:
    2. Run the following command to push the locally generated image to TCR.

      docker push ccr.ccs.tencentyun.com/[namespace]/[ImageName]:[image tag]

      As shown below:

    3. In My Images, you can view all the uploaded images. The following figure shows the five QCBM images uploaded to TCR.

      The default image type is Private. If you want to let others use the image, you can set it to Public in Image Info as shown below:

    Deploying service in TKE

    Creating K8s cluster of QCBM

    1. Before the deployment, you need to create a K8s cluster as instructed in Quickly Creating a Standard Cluster.

      When a cluster is created, we recommend you enable Placement Group on the Select Model page. It helps distribute CVM instances across different hosts to increase the system reliability.

    2. After the cluster is created, you can view its information on the cluster management page in the TKE console. Here, the new cluster is named qcbm-k8s-demo as shown below:
    3. Click the Cluster Name to enter the Basic Info page to view the cluster configuration information as shown below:
    4. (Optional) If you want to use K8s management tools such as kubectl and Lens, you need to follow two steps:
      1. Enable public network access.
      2. Store the API authentication token in the local config file under user home/.kube (choose another if the config file has content) to ensure that the default cluster can be accessed each time. If you choose not to store the token in the config file under .kube, see the Instructions on Connecting to Kubernetes Cluster via kubectl under Cluster API Server Info in the console as shown below:

    Creating namespace

    A namespace is a logical environment in a Kubernetes cluster that allows you to divide teams or projects. You can create a namespace in the following three methods, and method 1 is recommended.

    Run the following command to create a namespace:

    kubectl create namespace qcbm

    Using ConfigMap to store configuration information

    ConfigMap allows you to decouple the configuration from the running image, making the application more portable. The QCBM backend service needs to get the Nacos, MySQL, and Redis host and port information from the environment variables and store them by using ConfigMap.
    You can use ConfigMap to store configuration information in the following two methods:

    The following is the ConfigMap YAML for QCBM, where values of pure digits require double quotation marks, for example, MYSQL_PORT in the sample YAML below:

    # Create a ConfigMap.

    apiVersion: v1
    kind: ConfigMap
    name: qcbm-env
    namespace: qcbm
    NACOS_PORT: "8848"
    MYSQL_PORT: "3306"
    REDIS_PORT: "6379"
    SW_AGENT_COLLECTOR_BACKEND_SERVICES: xxx # TSW access address as described below

    Using Secret to store sensitive information

    A Secret can be used to store sensitive information such as passwords, tokens, and keys to reduce exposure risks. QCBM uses it to store account and password information.
    You can use a Secret to store sensitive information in the following two methods:

    The following is the YAML for creating a Secret in QCBM, where the value of the Secret needs to be a Base64-encoded string.

    # Create a Secret.
    apiVersion: v1
    kind: Secret
    name: qcbm-keys
    namespace: qcbm
    qcloud-app: qcbm-keys
    # xxx is the Base64-encoded string, which can be generated by using the echo -n raw string | base64 shell command.
    SW_AGENT_AUTHENTICATION: xxx # TSW access token as described below
    type: Opaque

    Deploying Deployment

    A Deployment declares the Pod template and controls the Pod running policy, which is suitable for deploying stateless applications. Both front and Dubbo services of QCBM are stateless applications and can use the Deployment.

    YAML parameters for the user-service Deployment are as shown below:

    Parameter Description
    replicas Indicates the number of Pods to be created.
    image Image address
    imagePullSecrets The key to pull an image, which can be obtained from Cluster > Configuration Management > Secret. It is not required for public images.
  • Defines Pod environment variables and values.
  • The key-value defined in the ConfigMap can be referenced by using configMapKeyRef.
  • The key-value defined in the Secret can be referenced by using secretKeyRef.
  • ports Specifies the port number of the container. It is 20880 for Dubbo applications.

    A complete sample YAML file for the user-service Deployment is as follows:

    # user-service Deployment

    apiVersion: apps/v1
    kind: Deployment
    name: user-service
    namespace: qcbm
    app: user-service
    version: v1
    replicas: 1
    app: user-service
    version: v1
    app: user-service
    version: v1
    - name: user-service
    image: ccr.ccs.tencentyun.com/qcbm/user-service:1.1.4
    - name: NACOS_HOST # IP address of the Dubbo service registry Nacos
    key: NACOS_HOST
    name: qcbm-env
    optional: false
    - name: MYSQL_HOST # MySQL address
    key: MYSQL_HOST
    name: qcbm-env
    optional: false
    - name: REDIS_HOST # Redis IP address
    key: REDIS_HOST
    name: qcbm-env
    optional: false
    - name: MYSQL_ACCOUNT # MySQL account
    name: qcbm-keys
    optional: false
    - name: MYSQL_PASSWORD # MySQL password
    name: qcbm-keys
    optional: false
    - name: REDIS_PASSWORD # Redis password
    name: qcbm-keys
    optional: false
    - name: SW_AGENT_COLLECTOR_BACKEND_SERVICES # SkyWalking backend service address
    name: qcbm-env
    optional: false
    - name: SW_AGENT_AUTHENTICATION # Authentication token for SkyWalking Agent to connect to the backend service
    name: qcbm-keys
    optional: false
    - containerPort: 20880 # Dubbo port name
    protocol: TCP
    imagePullSecrets: # The key to pull the image. It is not required as the images of all QCBM services are public.
    - name: qcloudregistrykey

    Deploying Service

    You can specify the Service type with Kubernetes ServiceType, which defaults to ClusterIP. Valid values of ServiceType include the following:

    • LoadBalancer: Provides public network, VPC, and private network access.
    • NodePort: : Accesses services through the CVM IP and host port.
    • ClusterIP: Accesses services through the service name and port.

    For a production system, the gateway needs to be accessible within the VPC or private network, and the front needs to provide access to the private and public networks. Therefore, you need to set ServiceType to LoadBalancer for the QCBM gateway and front.
    TKE enriches the LoadBalancer mode by configuring the Service through annotations.

    If you use the service.kubernetes.io/qcloud-loadbalancer-internal-subnetid annotations, a private network CLB instance will be created when the Service is deployed. In general, we recommend you create the CLB instance in advance and use the service.kubernetes.io/loadbalance-id annotations in the deployment YAML to improve the efficiency.

    The deployment YAML for the qcbm-front Service is as follows:

    # Deploy the `qcbm-front` Service.
    apiVersion: v1
    kind: Service
    name: qcbm-front
    namespace: qcbm
    # ID of the CLB instance of `Subnet-K8S`
    service.kubernetes.io/loadbalance-id: lb-66pq34pk
    externalTrafficPolicy: Cluster
    - name: http
    port: 80
    targetPort: 80
    protocol: TCP
    selector: # Map the backend `qcbm-gateway` to the Service.
    app: qcbm-front
    version: v1
    type: LoadBalancer

    Deploying Ingress

    An Ingress is a collection of rules that allow external access to the cluster Service, thereby eliminating the need to expose the Service. For QCBM projects, you need to create an Ingress for qcbm-front, which corresponds to the following YAML:

    # Deploy the `qcbm-front` Ingress.

    apiVersion: networking.k8s.io/v1beta1
    kind: Ingress
    name: front
    namespace: qcbm
    ingress.cloud.tencent.com/direct-access: "false"
    kubernetes.io/ingress.class: qcloud
    kubernetes.io/ingress.extensiveParameters: '{"AddressIPVersion":"IPV4"}'
    kubernetes.io/ingress.http-rules: '[{"host":"qcbm.com","path":"/","backend":{"serviceName":"qcbm-front","servicePort":"80"}}]'
    - host: qcbm.com
    - path: /
    backend: # Associate with backend services.
    serviceName: qcbm-front
    servicePort: 80

    Viewing deployment result

    So far, you have completed the deployment of QCBM in TKE and can view the deployment result in the following steps:

    1. Log in to the TKE console and click the Cluster ID/Name to enter the cluster details page.
    2. Click Services and Routes > Ingress to enter the Ingress page, where you can see the created Ingress. You can access the QCBM page through the Ingress VIP.

    Integrating CLS

    Enabling container log collection

    The container log collection feature is disabled by default and needs to be enabled as instructed below:

    1. Log in to the TKE console and click Cluster Ops > Feature Management on the left sidebar.
    2. At the top of the Feature Management page, select the region. On the right of the target cluster, click Set.
    3. On the Configure Features page, click Edit for log collection and select Enable Log Collection as shown below:
    4. Click OK.

    Creating log topic and logset

    QCBM is deployed in Nanjing region, so you need to select Nanjing region when creating logsets:

    1. Log in to the CLS console and select Nanjing region on the Log Topic page.
    2. Click Create Log Topic and enter the relevant information in the pop-up window as prompted as shown below:

    • Log Topic Name: Enter qcbm.
    • Logset Operation: Select Create Logset.
    • Logset Name: Enter qcbm-logs.
    1. Click OK.

      As QCBM has multiple backend microservices, you can create a log topic for each microservice to facilitate log categorization.

      • A log topic is created for each QCBM service.
      • You need the log topic ID when creating log rules for containers.

    Configuring log collection rule

    You can configure container log collection rules in the console or with CRD.

    Log rules specify the location of a log in a container:

    1. Log in to the TKE console and click Cluster Ops > Log Rules on the left sidebar.
    2. On the Log Rules page, click Create to create a rule.
      • Log Source: Specify the location of a log in a container. All the QCBM logs are output to the /app/logs directory, so you can use the container file path to specify the workload and log location.
      • Consumer: Select the previously created logset and topic.
    3. Click Next to enter the Log Parsing Method. Here, single-line text is used for QCBM. For more information on the log formats supported by CLS, see Full Text in a Single Line.

    Viewing log

    1. Log in to the CLS console and enter the Search and Analysis page.
    2. On the Search and Analysis page, Create Index for the logs first and then click Search and Analysis to view the logs.

    You can't find logs if no indexes are created.

    Integrating TSW

    TSW is currently in beta test and can be deployed in Guangzhou and Shanghai. Here, Shanghai is used as an example (QCBM is deployed in Nanjing).

    Accessing TSW - getting access point information

    1. Log in to the TSW console and click Service Observation > Service List on the left sidebar.
    2. Click Access Service and select Java and the SkyWalking data collection method. The access method provides the Access Point and Token information.

    Accessing TSW - application and container configuration

    Enter the Access Point and Token of the TSW obtained in the previous step in collector.backend_service and agent.authentication respectively in the agent.config of SkyWalking. agent.service_name is the service name, and agent.namespace can be used to group microservices under the same domain. user-service configuration is as shown below:

    You can also configure SkyWalking Agent by using environment variables. QCBM uses the ConfigMap and Secret to configure environment variables:

    • Use the ConfigMap to configure SW_AGENT_COLLECTOR_BACKEND_SERVICES.
    • Use the Secret to configure SW_AGENT_AUTHENTICATION.

    As shown below:

    At this point, you have completed TSW access. After starting the container service, you can view the call chain, service topology, and SQL analysis in the TSW console.

    Using TSW

    Viewing call exception through service API or call chain

    1. Log in to the TSW console and click Service Observation > API Observation on the left sidebar.
    2. On the API Observation page, you can view the call status of all APIs under a service, including the number of requests, success rate, error rate, response time, and other metrics.
    3. In the above figure, two qcbm-gateway APIs (/api/favorites/query/{userId} for querying user favorites and /api/order/{userId} for querying user orders) encountered call exceptions. Click the /api/favorites/query/{userId} API to view all the call records, locate the abnormal call chain, and click it to view the cause.
      The analysis shows that favorites-service encountered a call exception due to time-out.

    Using TSW to analyze add-on (such as SQL and caching) call

    1. Log in to the TSW console and click Add-on Call Observation > SQL Call on the left sidebar.
    2. On the SQL Call page, you can view the call details of SQL, NoSQL, MQ, and other add-ons. For example, you can quickly locate frequent SQL requests and slow queries in your application with the number and durations of SQL requests.

    Viewing service topology

    1. Log in to the TSW console and click Chain Tracing > Distributed Dependency Topology on the left sidebar.
    2. On the Distributed Dependency Topology page, you can view the completed service dependencies as well as information such as the number of calls and average latency.
    Contact Us

    Contact our sales team or business advisors to help your business.

    Technical Support

    Open a ticket if you're looking for further assistance. Our Ticket is 7x24 avaliable.

    7x24 Phone Support