How to implement microservices in PHP/Golang projects in the cloud native era

Original: https://wenda.swoole.com/detail/108785

foreword

Under the traditional architecture, we need to use technologies such as service discovery and service registration to implement a microservice architecture.通常我们需要将服务提供方（ Service Provider ）的节点（ IP:PORT ） ZooKeeper 、 ETCD 、 Consul , Nacos and other service management components, the service caller ( Service Customer ) can read the node list, initiate RPC call, establish a connection and send Receive the response.

Usually we need to rely on the implementation of specific frameworks, such as Spring Cloud , Dubble , Hyperf and other frameworks. It is no longer necessary to implement microservices in the cloud-native era. We can directly use the Service provided by K8s to implement the microservice architecture , and sink service registration and discovery to the bottom layer of the system. Higher reliability, better stability, and natural cross-language, Java, PHP, and Golang can be used, and the development framework will become simpler. You don't need to do anything, just implement the logic of the service, monitor this machine port.

Introduction to K8s Service

Service is a mechanism provided by K8s to discover back-end Pods. It provides a unified entry address for a group of all Pods belonging to the same Deployment, and distributes requests to each Pod.

In K8s, Pod is equivalent to a Linux system process, a container group that executes applications. How to access the service in the Pod? It is definitely not possible to directly connect to the Pod IP:Port , because it is unstable and may be scheduled and restarted at any time. The life cycle of the Pod is very short. The IP port will change after the Pod restarts. A service or application usually has 1-N Pods. When requesting Service IP:PORT , it will automatically load balance and forward it to one of the Pods. Service generally uses kube-proxy and the Linux kernel provides The IPVS technical implementation. It can also be understood that Service is actually a 4-layer proxy, and the back end is the Pod of the application. Before the Service, we can set up the Ingress access gateway to allow access from outside the cluster. Generally, the HTTP interface of external services is this way. It can also be used directly, only allowing internal access, which is the microservice pattern.

Using Service on Code-Galaxy Platform

K8s 为每个Service Name ，这个Name CoreDNS中，可gethostbyname其他DNS 解析method to resolve Server Name into the IP address of the Service. One Service needs to set the external port, that is, the port exposed by the Service, and the other is the internal port, that is, the port of the Pod. Pod related information will be saved in K8s ETCD distributed storage.

For example, using the Hyperf framework, it will listen on port 9501 by default, and expect the service caller to access directly through port 80, then the inside and outside of the Service are 9501:80 . At this time, in other services, you can use HTTP client access http://ServiceName:80/ to request this service.

When different namespaces access each other, you need to add ---2e0077e36c81b5124270095a5acf61c7 Server Name after .{$namespace}

refer to:

1. https://blog.csdn.net/huahua1999/article/details/124237065
2. https://blog.51cto.com/u_15049785/4174726