Overview
An endpoint group refers to a set of endpoints located in a specific region, used to receive and handle client requests forwarded via Tencent Cloud's Global Accelerator network. Its core purpose is to efficiently distribute traffic from the acceleration entry (acceleration IP address) to the backend services (such as ECS and CLB), achieving cross-region access with low latency. The main features of the endpoint group are as follows:
Traffic distribution: A Global Accelerator listener forwards client requests to the associated endpoint group based on the routing rules (currently, only intelligent routing is supported). Then, the endpoints in the endpoint group deliver the requests to the backend services.
Region association: Each endpoint group is bound to a specific region (such as Beijing or Shanghai) to ensure nearby traffic access to the backend services, reducing network latency.
Multi-node disaster recovery: Usually, up to 4 endpoints can be added to an endpoint group. High availability is achieved through health checks, and an endpoint can be automatically switched in case of a fault.
Note:
China Unicom (CUCC) provides agency operations for the cross-border segment of Global Accelerator. If there are cross-border elements in the acceleration region or the endpoint group, your account needs to pass the Cross-Border Qualification Review first. Endpoint Group Types
Endpoint groups are divided into default endpoint groups and custom endpoint groups. The node group created when a listener is first created is the default endpoint group.
Default endpoint group:
TCP/UDP listening: Only 1 default endpoint group can be created. After creation, the business traffic will be forwarded to healthy endpoints based on the configured health check policies.
HTTP/HTTPS listening: Only 1 default group can be created.
Custom endpoint group:
Suitable for HTTP/HTTPS listening. Up to 10 custom node groups can be created. Some traffic can be directed to specific custom node groups through forwarding policies (such as URL-based forwarding).
Note:
After a default endpoint group is created, it is bound to the default forwarding policy.
Only HTTP and HTTPS listeners support configuring the forwarding policy and binding to a custom endpoint group. For more information, see Forwarding Policy. Endpoint Types
Endpoint refers to the backend service instance where client requests eventually reach and are processed. As a proxy node in the Global Accelerator network, it forwards the acceleration traffic to the actual business server. After the configuration is completed, Global Accelerator will forward the business traffic to endpoints via the public network. Endpoints can be of the following types:
Custom public IP address
Custom domain name
Origin-Pull Protocol
The origin-pull protocol refers to the application layer protocol used when Global Accelerator forwards client requests to backend servers (endpoints). For example, when a client accesses a load balancer via HTTPS, the load balancer might use HTTP for origin-pull. The origin-pull protocol can only be selected when an endpoint group is created for a layer-7 listener.
Port Mapping
Port mapping refers to the association and forwarding of the frontend listening ports of client requests with the actual service ports of the backend servers by the Global Accelerator service at the application layer (HTTP/HTTPS).
Frontend port: The port for a client to access Global Accelerator (such as ports 80 and 443).
Backend port: The port to which Global Accelerator forwards requests to the backend endpoints (such as ports 8080 and 8443).
Mapping relationship: When creating an endpoint group, configure the port mapping to direct requests from the frontend port to the specified backend port, achieving protocol conversion or port decoupling.
Note:
When port mapping is configured for the endpoint group, the listening port needs to be consistent with the listener port, and this cannot be modified.
Health Check
Health check is a mechanism used to detect whether backend endpoints are available in the Global Accelerator service. By sending probe requests regularly, it determines the running status of backend services and ensures that traffic is only distributed to healthy endpoints, thereby guaranteeing the high availability of the business. Health checks can be configured for endpoint groups. For more information, see Configuring HTTP and HTTPS Listeners. Documentation