How to get remote address from request of Netty HTTP Server

获取 remoteAddress 的逻辑

通过HTTP请求获取 remote address 时，最终是调用 ConnectionInfo 的 getRemoteAddress() 方法。

reactor.netty.http.server.HttpServerOperations#remoteAddress

@Override
public InetSocketAddress remoteAddress() {
	return this.connectionInfo.getRemoteAddress();
}

ConnectionInfo 实例的创建过程如下：

从 “Forwarded” 或 "X-Forwarded-" HTTP请求头中获取连接信息，如果不存在则直接从当前连接获取连接信息。*

请注意：如果 Nginx 和 Netty HTTP Server 位于同一服务器内，且未启动 use-forward-headers，则获取的 remoteAddress 为 127.0.0.1 。

reactor.netty.http.server.ConnectionInfo#from

static ConnectionInfo from(Channel channel, boolean headers, HttpRequest request, boolean secured) {
	if (headers) {
		return ConnectionInfo.newForwardedConnectionInfo(request, channel, secured);
	}
	else {
		return ConnectionInfo.newConnectionInfo(channel, secured);
	}
}

static ConnectionInfo newForwardedConnectionInfo(HttpRequest request, Channel channel, boolean secured) {
	if (request.headers().contains(FORWARDED_HEADER)) {
		return parseForwardedInfo(request, (SocketChannel)channel, secured);
	}
	else {
		return parseXForwardedInfo(request, (SocketChannel)channel, secured);
	}
}

static ConnectionInfo parseXForwardedInfo(HttpRequest request, SocketChannel channel, boolean secured) {
	InetSocketAddress hostAddress = channel.localAddress();
	InetSocketAddress remoteAddress = getRemoteAddress(channel);
	String scheme =  secured ? "https" : "http";
	if (request.headers().contains(XFORWARDED_IP_HEADER)) {
		String remoteIpValue = request.headers().get(XFORWARDED_IP_HEADER).split(",")[0];
		remoteAddress = parseAddress(remoteIpValue, remoteAddress.getPort());
	}
	if(request.headers().contains(XFORWARDED_HOST_HEADER)) {
		if(request.headers().contains(XFORWARDED_PORT_HEADER)) {
			hostAddress = InetSocketAddressUtil.createUnresolved(
					request.headers().get(XFORWARDED_HOST_HEADER).split(",")[0].trim(),
					Integer.parseInt(request.headers().get(XFORWARDED_PORT_HEADER).split(",")[0].trim()));
		}
		else {
			hostAddress = InetSocketAddressUtil.createUnresolved(
					request.headers().get(XFORWARDED_HOST_HEADER).split(",")[0].trim(),
					channel.localAddress().getPort());
		}
	}
	if (request.headers().contains(XFORWARDED_PROTO_HEADER)) {
		scheme = request.headers().get(XFORWARDED_PROTO_HEADER).trim();
	}
	return new ConnectionInfo(hostAddress, remoteAddress, scheme);
}

通过配置文件启用 use-forward-headers

要想通过 request header 获取连接信息，需要配置 use-forward-headers = true。（默认为 false）

server:
  port: 9090
  use-forward-headers: true

通过 NettyWebServerFactoryCustomizer 配置

也可以通过NettyWebServerFactoryCustomizer启动，示例代码如下：

@Configuration
public class NettyWebServerConfiguration {
    @Bean
    public NettyWebServerFactoryCustomizer nettyServerWiretapCustomizer(
            Environment environment, ServerProperties serverProperties) {
        return new NettyWebServerFactoryCustomizer(environment, serverProperties) {
            @Override
            public void customize(NettyReactiveWebServerFactory factory) {
                factory.addServerCustomizers(httpServer -> httpServer.forwarded(true));
                super.customize(factory);
            }
        };
    }
}

set X-Forwarded-For header from nginx

1.通过“proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for” 把从真实客户端IP和反向代理IP通过逗号分隔，添加到请求头中；
2.可以在第一个反向代理上配置“proxy_set_header X-Real-IP $remote_addr” 获取真实客户端IP；
3.配合 realip 模块从 X-Forwarded-For 也可以获取到真实客户端IP。

在整个反向代理链条的第一个反向代理可以不配置“proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for”，否则客户端可以伪造 X-Forwarded-For 从而伪造客户端真实 IP，如果服务端使用 X-Forwarded-For 第一个IP作为真实客户端 IP，则就出问题了。如果通过配置 X-Real-IP 请求头或者配合 realip 模块也不会出现该问题。如果自己解析 X-Forwarded-For 的话，记得使用 realip 算法解析，而不是取第一个。
当我们进行限流时一定注意限制的是真实客户端IP，而不是反向代理IP，我遇到过很多同事在这块出问题的。

location / {
    proxy_next_upstream		http_500 http_502 http_503 http_504 error timeout invalid_header;
    proxy_set_header		Host  $host;
    proxy_set_header		X-Forwarded-For $proxy_add_x_forwarded_for;
    proxy_pass				http://tomcat_microservice;
    expires					0;
}

参考

反向代理与 Real-IP 和 X-Forwarded-For