Develop Reference

How InetAddress.getLocalHost() works?

I'm trying to get how InetAddress.getLocalHost() works. Javadoc says that it retrieves the name of the host from the system and then resolves it into an InetAddress. What exactly does "resolve into InetAddess" mean? Does it simply ask DNS to resolve a hostname?

From InetAddress.java source:
private static InetAddress[] getAddressesFromNameService(String host, InetAddress reqAddr)
throws UnknownHostException
{
InetAddress[] addresses = null;
boolean success = false;
UnknownHostException ex = null;
// Check whether the host is in the lookupTable.
// 1) If the host isn't in the lookupTable when
// checkLookupTable() is called, checkLookupTable()
// would add the host in the lookupTable and
// return null. So we will do the lookup.
// 2) If the host is in the lookupTable when
// checkLookupTable() is called, the current thread
// would be blocked until the host is removed
// from the lookupTable. Then this thread
// should try to look up the addressCache.
// i) if it found the addresses in the
// addressCache, checkLookupTable() would
// return the addresses.
// ii) if it didn't find the addresses in the
// addressCache for any reason,
// it should add the host in the
// lookupTable and return null so the
// following code would do a lookup itself.
...
if (host.equalsIgnoreCase("localhost")) {
InetAddress[] local = new InetAddress[] { impl.loopbackAddress() }; // {0x7f,0x00,0x00,0x01}
addresses = local;
success = true;
break;
}
To recap:
both InetAddress.getAllByName() and InetAddress.getLocalHost() resolve the address by calling getAddressesFromNameService()
The JVM maintains its own cache of hostname -> IP address mapping.
If the address is not in the cache (lookupTable or addressCache) it will call the operating system's DNS (exact behavior may vary per JVM implementation).
For localhost sepcificly - there is a specific case inside getAddressesFromNameService()

How to learn system (Windows) non-proxy hosts in Java

I'm writting a Java (1.7) application to be running on Windows. The application is accessing additional services running on the same host and other ones running in the Internet. The application can be run in two environments where in one, proxy settings must be specified (there is proxy when accessing the Internet); while in the other environment, the proxy settings must not be specified (there is no proxy).
I want the application to be simple and don't want its users bother with specification of the proxy settings on cmd-line (-Dhttp.proxyHost, etc.) - the application should learn the proxy settings from Windows system settings (IE / Tools / Internet Properties / Connections / LAN Settings).
I have written a piece of code that is supposed to learn that settings, see below. The trouble is that this piece of code does not identify localhost, 127.0.0.1 and my-computer-name (where my-computer-name is the name of my computer) as URLs where proxy should be by-passed when being accessed (yes, I do have 'Bypass proxy server for local addresses' checked). As a result, the application tries to access local services through the proxy which is wrong.
So far I've found out that one way to teach JVM not to use proxy for 'local addresses' is to list the strings (localhost, 127.0.0.1, my-computer-name) in Proxy Settings / Exceptions (Do not use proxy server for addresses beginning with). Obviously, this is not a good solution as usually no one is listing these strings there (the first check-box is enough for non-Java applications).
Second (trivial) solution would be just to count with these strings in my piece of code and do not use proxy settings for them even when JVM thinks otherwise. I don't think this is a good solution and if this is the only solution, IMHO, there is a defect in JVM.
I've found many resources in the Internet how to learn System proxy settings. But how to learn the non-proxy settings?
Thanks,
PP
public static final String HTTP_PROXY_HOST_KEY = "http.proxyHost";
public static final String HTTPS_PROXY_HOST_KEY = "https.proxyHost";
public static final String HTTP_PROXY_PORT_KEY = "http.proxyPort";
public static final String HTTPS_PROXY_PORT_KEY = "https.proxyPort";
public static final String NO_PROXY_HOSTS_KEY = "http.nonProxyHosts";
// provide list of urls which are to be accessed by this application and return proxy and non-proxy settings
private Properties getSystemProxyConfiguration(String[] urls) {
log.debug("Getting system proxy");
Properties properties = new Properties();
SortedSet<String> nonProxyHosts = new TreeSet<>();
for (String url : urls) {
URI uri;
try {
uri = new URI(url);
} catch (URISyntaxException e) {
throw new IllegalStateException(e);
}
InetSocketAddress address = getSystemProxy(uri);
if (address != null) {
if (url.toLowerCase().startsWith("https")) {
properties.put(HTTPS_PROXY_HOST_KEY, address.getHostString());
properties.put(HTTPS_PROXY_PORT_KEY, ""+address.getPort());
//todo verify that all previous URLs in this array are using the same proxy
log.debug("HTTPS proxy: " + address.getHostString() + ":" + address.getPort());
} else {
properties.put(HTTP_PROXY_HOST_KEY, address.getHostString());
properties.put(HTTP_PROXY_PORT_KEY, ""+address.getPort());
//todo verify that all previous URLs in this array are using the same proxy
log.debug("HTTP proxy: " + address.getHostString() + ":" + address.getPort());
}
} else { //todo DEFECT -> this does not find the non-proxy hosts (even though specified in IE Internet settings)
nonProxyHosts.add(uri.getHost());
}
}
if (nonProxyHosts.size() > 0) {
String nonProxyHostsString = nonProxyHosts.first();
nonProxyHosts.remove(nonProxyHostsString);
for (String nonProxyHost : nonProxyHosts) {
nonProxyHostsString = nonProxyHostsString + "|" + nonProxyHost;
}
properties.put(NO_PROXY_HOSTS_KEY, nonProxyHostsString);
log.debug("Non HTTP(S) proxy hosts: "+nonProxyHostsString);
} else {
log.debug("No non HTTP(S) proxy hosts set");
}
return properties;
}
private InetSocketAddress getSystemProxy(URI uri) {
List<Proxy> proxyList;
proxyList = ProxySelector.getDefault().select(uri);
if (proxyList != null && proxyList.size() > 0) { //todo DEFECT - this never returns DIRECT proxy for localhost, 127.0.0.1, my-computer-name strings
Proxy proxy = proxyList.get(0);
if (proxyList.size() > 1) {
log.warn("There is more " + proxy.type() + " proxies available. Use "+PROXY_PROPERTIES_FILE_NAME+" to set the right one.");
}
InetSocketAddress address = (InetSocketAddress) proxy.address();
return address;
}
return null;
}

Identify server on Tomcat (HttpServletRequest.getLocalAddr() fails)

With Tomcat setup behind Apache, how can an id (IP address ideally) of the server be easily determined?
The specific situation is that multiple servers are setup behind a load balancer, thus the incoming request host name is non-unique and insufficient to identify a particular server for logging purposes. Using HttpServletRequest.getLocalAddr() is unfortunately returning the same hostname instead of the IP address as would be expected (I am assuming this is related to this very old issue here: https://issues.apache.org/bugzilla/show_bug.cgi?id=46082).
Is there a way to make getLocalAddr() perform as documented, or are other methods required to query the IP address of the server?

On our project, we use JMX to get all the config information.
It takes a few steps, because it is like navigating down the server.xml file
This link has some info: http://oss.wxnet.org/mbeans.html
It is probably overkill if all you want is the IP, but I thought I'd throw it out there.
MBeanServer mbeanServer = ManagementFactory.getPlatformMBeanServer();
Set<ObjectName> theConnectors = mbeanServer.queryNames(
new ObjectName("Catalina:type=Connector,*"),
null);
if (theConnectors != null)
{
for (ObjectName nextConnectorName : theConnectors)
{
InetAddress theInetAddress = (InetAddress) mbeanServer.getAttribute(
nextConnectorName,
"address");
if (theInetAddress != null)
{
ipAddress = theInetAddress.getHostAddress();
}
if (!StringUtil.isEmpty(ipAddress))
{
// found the IP address
break;
}
}
}

For my situation, the solution was to get the IP address of the server directly instead of attempting to get the local address via HttpServleRequest.
I cached the IP for use in my filter via:
private static final String serverIp;
static {
String addressString = null;
try
{
InetAddress address = InetAddress.getLocalHost();
addressString = address.getHostAddress();
} catch (Exception e)
{
logger.error("Exception while attempting to determine local ip address",e);
}
if (addressString != null) serverIp = addressString;
else serverIp = "unknown";
}

I had a similar issue recently (a few years after the original question) and found this question and answers. The issue in my case was that the ServletRequest#getLocalAddr() implementation was returning the remote address instead of the local address. The issue was caused by a regression in Tomcat v9.0.22. It was fixed in v9.0.23. See the question and answer here:
https://stackoverflow.com/a/57725039/9602527

How to identify the URL of an Java web application from within?

My Java web application contains a startup servlet. Its init() method is invoked, when the web application server (Tomcat) is started. Within this method I need the URL of my web application. Since there is no HttpServletRequest, how to get this information?

You can't. Because there is no "URL of an Java web application" as seen "from within". A servlet is not tied to an URL, that is done from the outside. (Perhaps you have a Apache server that connects to a Tomcat - Tomcat can't know about it)
It makes sense to ask a HttpServletRequest for its url, because we are speaking of the information of a event (the URL that was actually used to generate this request), it does not make sense to ask for a configuration URL.

A workaround could be to perform the initialization lazy when the first request arrives. You can implement a filter that do that once, e.g. by storing a boolean flag in a static variable and synchronizing access to the flag correctly. But it implies a little overhead because each subsequent request will go through the filter which then bypass the initialization. It was just a thought.

There is nothing in the servlet API that provides this information, plus any given resource may be bound to multiple URL's.
What you CAN do, is to inspect the servlet context when you receive an actual request and see what URL was used.

Here is how it works for me and probably for most configurations:
public static String getWebappUrl(ServletConfig servletConfig, boolean ssl) {
String protocol = ssl ? "https" : "http";
String host = getHostName();
String context = servletConfig.getServletContext().getServletContextName();
return protocol + "://" + host + "/" + context;
}
public static String getHostName() {
String[] hostnames = getHostNames();
if (hostnames.length == 0) return "localhost";
if (hostnames.length == 1) return hostnames[0];
for (int i = 0; i < hostnames.length; i++) {
if (!"localhost".equals(hostnames[i])) return hostnames[i];
}
return hostnames[0];
}
public static String[] getHostNames() {
String localhostName;
try {
localhostName = InetAddress.getLocalHost().getHostName();
} catch (UnknownHostException ex) {
return new String[] {"localhost"};
}
InetAddress ia[];
try {
ia = InetAddress.getAllByName(localhostName);
} catch (UnknownHostException ex) {
return new String[] {localhostName};
}
String[] sa = new String[ia.length];
for (int i = 0; i < ia.length; i++) {
sa[i] = ia[i].getHostName();
}
return sa;
}

Java: Common way to validate and convert "host:port" to InetSocketAddress?

What is the common way in Java to validate and convert a string of the form host:port into an instance of InetSocketAddress?
It would be nice if following criteria were met:
No address lookups;
Working for IPv4, IPv6, and "string" hostnames;
(For IPv4 it's ip:port, for IPv6 it's [ip]:port, right? Is there some RFC which defines all these schemes?)
Preferable without parsing the string by hand.
(I'm thinking about all those special cases, when someone think he knows all valid forms of socket addresses, but forgets about "that special case" which leads to unexpected results.)

I myself propose one possible workaround solution.
Convert a string into URI (this would validate it automatically) and then query the URI's host and port components.
Sadly, an URI with a host component MUST have a scheme. This is why this solution is "not perfect".
String string = ... // some string which has to be validated
try {
// WORKAROUND: add any scheme to make the resulting URI valid.
URI uri = new URI("my://" + string); // may throw URISyntaxException
String host = uri.getHost();
int port = uri.getPort();
if (uri.getHost() == null || uri.getPort() == -1) {
throw new URISyntaxException(uri.toString(),
"URI must have host and port parts");
}
// here, additional checks can be performed, such as
// presence of path, query, fragment, ...
// validation succeeded
return new InetSocketAddress (host, port);
} catch (URISyntaxException ex) {
// validation failed
}
This solution needs no custom string parsing, works with IPv4 (1.1.1.1:123), IPv6 ([::0]:123) and host names (my.host.com:123).
Accidentally, this solution is well suited for my scenario. I was going to use URI schemes anyway.

A regex will do this quite neatly:
Pattern p = Pattern.compile("^\\s*(.*?):(\\d+)\\s*$");
Matcher m = p.matcher("127.0.0.1:8080");
if (m.matches()) {
String host = m.group(1);
int port = Integer.parseInt(m.group(2));
}
You can this in many ways such as making the port optional or doing some validation on the host.

It doesn't answer the question exactly, but this answer could still be useful others like me who just want to parse a host and port, but not necessarily a full InetAddress. Guava has a HostAndPort class with a parseString method.

Another person has given a regex answer which is what I was doing to do when originally asking the question about hosts. I will still do because it's an example of a regex that is slightly more advanced and can help determine what kind of address you are dealing with.
String ipPattern = "(\\d{1,3}\\.\\d{1,3}\\.\\d{1,3}\\.\\d{1,3}):(\\d+)";
String ipV6Pattern = "\\[([a-zA-Z0-9:]+)\\]:(\\d+)";
String hostPattern = "([\\w\\.\\-]+):(\\d+)"; // note will allow _ in host name
Pattern p = Pattern.compile( ipPattern + "|" + ipV6Pattern + "|" + hostPattern );
Matcher m = p.matcher( someString );
if( m.matches() ) {
if( m.group(1) != null ) {
// group(1) IP address, group(2) is port
} else if( m.group(3) != null ) {
// group(3) is IPv6 address, group(4) is port
} else if( m.group(5) != null ) {
// group(5) is hostname, group(6) is port
} else {
// Not a valid address
}
}
Modifying so that port is optional is pretty straight forward. Wrap the ":(\d+)" as "(?::(\d+))?" and then check for null for group(2), etc.
Edit: I'll note that there's no "common way" way that I'm aware of but the above is how I'd do it if I had to.
Also note: the IPv4 case can be removed if the host and IPv4 cases will actually be handled the same. I split them out because sometimes you can avoid an ultimate host look-up if you know you have the IP address.

new InetSocketAddress(
addressString.substring(0, addressString.lastIndexOf(":")),
Integer.parseInt(addressString.substring(addressString.lastIndexOf(":")+1, addressString.length));
? I probably made some little silly mistake. and I'm assuming you just wanted a new InetSocketAddress object out of the String in only that format. host:port

All kind of peculiar hackery, and elegant but unsafe solutions provided elsewhere. Sometimes the inelegant brute-force solution is the way.
public static InetSocketAddress parseInetSocketAddress(String addressAndPort) throws IllegalArgumentException {
int portPosition = addressAndPort.length();
int portNumber = 0;
while (portPosition > 1 && Character.isDigit(addressAndPort.charAt(portPosition-1)))
{
--portPosition;
}
String address;
if (portPosition > 1 && addressAndPort.charAt(portPosition-1) == ':')
{
try {
portNumber = Integer.parseInt(addressAndPort.substring(portPosition));
} catch (NumberFormatException ignored)
{
throw new IllegalArgumentException("Invalid port number.");
}
address = addressAndPort.substring(0,portPosition-1);
} else {
portNumber = 0;
address = addressAndPort;
}
return new InetSocketAddress(address,portNumber);
}

The open-source IPAddress Java library has a HostName class which will do the required parsing. Disclaimer: I am the project manager of the IPAddress library.
It will parse IPv4, IPv6 and string host names with or without ports. It will handle all the various formats of hosts and addresses. BTW, there is no single RFC for this, there are a number of RFCs that apply in different ways.
String hostName = "[a:b:c:d:e:f:a:b]:8080";
String hostName2 = "1.2.3.4:8080";
String hostName3 = "a.com:8080";
try {
HostName host = new HostName(hostName);
host.validate();
InetSocketAddress address = host.asInetSocketAddress();
HostName host2 = new HostName(hostName2);
host2.validate();
InetSocketAddress address2 = host2.asInetSocketAddress();
HostName host3 = new HostName(hostName3);
host3.validate();
InetSocketAddress address3 = host3.asInetSocketAddress();
// use socket address
} catch (HostNameException e) {
String msg = e.getMessage();
// handle improperly formatted host name or address string
}

URI can accomplish this:
URI uri = new URI(null, "example.com:80", null, null, null);
Unfortunately, there's a bug in current OpenJDK (or in the documentation) where the authority isn't properly validated. The documentation states:
The resulting URI string is then parsed as if by invoking the URI(String) constructor and then invoking the parseServerAuthority() method upon the result
That call to parseServerAuthority just doesn't happen unfortunately so the real solution here that properly validates is:
URI uri = new URI(null, "example.com:80", null, null, null).parseServerAuthority();
then
InetSocketAddress address = new InetSocketAddress(uri.getHost(), uri.getPort());