Connection and connection request timeout

Question

I am using Http Apache Components to perform the http interactions. I need to adjust my http client. For this purpose I have two parameters: connection timeout and connectio

Answer 1

So, my question is what do [connection-timeout and connection-request-timeout] mean and how they distinct from each other.

Connection-timeout is the timeout in milliseconds until the server accepts the request. If you specify 3000, the http-client will wait 3 seconds for the server to accept the TCP connection before timing out. This typically is used to make sure you don't have a networking issue or that you are contacting the right hostname or address. This is equivalent to curl's --connect-timeout seconds option.

Connection-request-timeout is the input/output timeout after the connection has been established. If you specify this value as 10000 then after the http-client has connected to the server and sends a request, it will wait 10 seconds for the server to return a result. This typically is used to ensure that your job doesn't wait for a slow server forever. This is equivalent to curl's --max-time seconds option.

In HttpClient 4.X.X , the following is how you build a client that uses a particular connectTimeoutMillis and requestTimeoutMillis.

HttpClientBuilder clientBuilder = HttpClientBuilder.create();
RequestConfig.Builder requestBuilder = RequestConfig.custom();
requestBuilder = requestBuilder.setConnectTimeout(connectTimeoutMillis);
requestBuilder = requestBuilder.setConnectionRequestTimeout(requestTimeoutMillis);
clientBuilder.setDefaultRequestConfig(requestBuilder.build());
CloseableHttpClient httpClient = clientBuilder.build();
...

Btw, the javadocs for this code sucks. Try to figure out by hand how to use the config builder. Holy crap.

Answer 2

From the documentation:

/**
 * Returns the timeout in milliseconds used when requesting a connection
 * from the connection manager. A timeout value of zero is interpreted
 * as an infinite timeout.
 * <p>
 * A timeout value of zero is interpreted as an infinite timeout.
 * A negative value is interpreted as undefined (system default).
 * </p>
 * <p>
 * Default: {@code -1}
 * </p>
 */
public int getConnectionRequestTimeout() {
    return connectionRequestTimeout;
}

/**
 * Determines the timeout in milliseconds until a connection is established.
 * A timeout value of zero is interpreted as an infinite timeout.
 * <p>
 * A timeout value of zero is interpreted as an infinite timeout.
 * A negative value is interpreted as undefined (system default).
 * </p>
 * <p>
 * Default: {@code -1}
 * </p>
 */
public int getConnectTimeout() {
    return connectTimeout;
}

This is how the code should look:

HttpClientBuilder clientBuilder = HttpClientBuilder.create();
RequestConfig.Builder requestBuilder = RequestConfig.custom();
// Connection Timeout to establish a connection
requestBuilder = requestBuilder.setConnectTimeout(connectTimeoutMillis);
// Timeout to get a connection from the connection manager for the Http Client
requestBuilder = requestBuilder.setConnectionRequestTimeout(requestTimeoutMillis);
// Timeout between two data packets from the server
requestBuilder = requestBuilder.setSocketTimeout(requestTimeoutMillis);
clientBuilder.setDefaultRequestConfig(requestBuilder.build());
CloseableHttpClient httpClient = clientBuilder.build();

Answer 3

HttpClient has a way to set connection and socket timeout (setConnectionTimeout() and setTimeout()) according to the HttpClient javadocs.

Connection timeout is the timeout until a connection with the server is established.

Socket timeout is the timeout to receive data (socket timeout).

Example:

Let's say you point your browser to access a web page. If the server does not anwser in X seconds, a connection timeout will occur. But if it establishes the connection, then the server will start to process the result for the browser. If it does not ends this processing in Y seconds, a socket timeout will occur.