http-server based on Java Sockets

I believe that every backender should write their own server

Main goal: understand how a connection pooling works with Java Sockets

Run server:

./gradlew clean build && java -jar build/libs/http-server.jar --port=8080 --threadsNumber=2 --keepAliveTime=30000

Configuration example:

HttpServer.create()
        .port(propertyResolver.getInt("port", 8080))
        .threadsNumber(propertyResolver.getInt("threadsNumber", 20))
        .keepAliveTime(propertyResolver.getInt("keepAliveTime", 30000))
        .endpoint(Endpoint
                .create()
                .method("GET")
                .path("/hello")
                .handler((request, response) -> response.writeAsString("Hello!"))
                .build())
        .endpoint(Endpoint
                .create()
                .method("POST")
                .path("/login")
                .handler((request, response) -> response.writeAsString("Hello, " + request.getRequestBody()))
                .build())
        .endpoint(Endpoint
                .create()
                .method("GET")
                .path("/favicon.ico")
                .handler((request, response) -> {
                    writeFileFromResource("static/favicon.ico", response.getOutputStream());
                    response.addHeader("Content-Type", "image/x-icon");
                    response.addHeader("Content-Disposition", "attachment; filename=favicon.ico");
                })
                .build())
        .endpoint(Endpoint
                .create()
                .method("GET")
                .path("/")
                .handler((request, response) -> {
                    writeFileFromResource("static/index.html", response.getOutputStream());
                    response.addHeader("Content-Type", "text/html;charset=utf-8");
                    response.addHeader("encoding", "utf-8");
                })
                .build())
        .build()
        .run();

Upload to S3:

./gradlew upload -Pbucket=your-bucket

Deploy to CloudFormation:

aws cloudformation create-stack \
  --stack-name <stack-name> \
  --template-body file://aws/ec2-java-server.yaml \
  --parameters \
    ParameterKey=KeyName,ParameterValue=<key-name> \
    ParameterKey=JarUrl,ParameterValue=<jar-url> \
    ParameterKey=JarParams,ParameterValue=<jar-params> \
  --capabilities CAPABILITY_NAMED_IAM

Useful links:

1. Sun http server implementation: sun/net/httpserver/ServerImpl.java:386
2. Java NIO: Non-blocking Server
3. Java NIO ServerSocketChannel
4. HTTP/1.0 specification
5. HTTP/1.1 specification

Tips to know:

1. Connection reuse only works since HTTP 1.1 (in curl at least). See Persistence section
2. It seems that it is better to build a connection pool based on Java NIO (ServerSocketChannel, SocketChannel), where we can register a selector that listens on channels in one thread. When using Java standard IO, we basically have a thread for each socket that is listening on it. This method may be ineffective with a large number of connections.