Prevent Jenkins hang with unresponsive web servers

pom.xml

@@ -42,10 +42,10 @@
             <version>3.1</version>
         </dependency>
         <dependency>
-   <groupId>commons-codec</groupId>
-   <artifactId>commons-codec</artifactId>
-   <version>1.4</version>
-</dependency>
+   			<groupId>commons-codec</groupId>
+   			<artifactId>commons-codec</artifactId>
+   			<version>1.4</version>
+		</dependency>
     </dependencies>
 
     <repositories>

src/main/java/hudson/plugins/sitemonitor/SiteMonitorDescriptor.java

@@ -84,6 +84,18 @@ public SiteMonitorDescriptor() {
     }
 
     /**
+     * For testing
+     */
+    protected SiteMonitorDescriptor(
+			List<Integer> mSuccessResponseCodes, 
+			Integer mTimeout) {
+		super(SiteMonitorRecorder.class);
+		this.mValidator = new SiteMonitorValidator();
+		this.mSuccessResponseCodes = mSuccessResponseCodes;
+		this.mTimeout = mTimeout;
+	}
+
+	/**
      * @return the plugin's display name, used in the job's build drop down list
      */
     @Override

src/main/java/hudson/plugins/sitemonitor/SiteMonitorRecorder.java

@@ -51,6 +51,15 @@
 import javax.net.ssl.X509TrustManager;
 import java.util.ArrayList;
 import java.util.List;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutionException;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.TimeoutException;
+import java.util.logging.Level;
+import java.util.logging.Logger;
 import org.apache.commons.codec.binary.Base64;
 
 /**
@@ -59,6 +68,9 @@
  */
 public class SiteMonitorRecorder extends Recorder {
 
+	private static final Logger LOGGER = Logger
+	            .getLogger(SiteMonitorRecorder.class.getName());
+
     /**
      * 1 sec = 1000 msecs .
      */
@@ -131,6 +143,51 @@ public boolean verify(String arg0, SSLSession arg1) {
         }
     }
 
+    public class StatusResponse {
+    	public Status status = null;
+    	public Integer responseCode = null;
+    }
+
+    private StatusResponse checkURL(Site site, SiteMonitorDescriptor descriptor) throws Exception {
+    	StatusResponse response = new StatusResponse();
+    	HttpURLConnection connection = null;
+
+        LOGGER.log(Level.FINE,"Checking URL {}",site.getUrl());
+
+    	connection = getConnection(site.getUrl());
+    	connection.setConnectTimeout(descriptor.getTimeout()
+    			* MILLISECS_IN_SECS);
+    	connection.setReadTimeout(descriptor.getTimeout()
+    			* MILLISECS_IN_SECS);
+    	response.responseCode = connection.getResponseCode();
+
+    	List<Integer> successResponseCodes = descriptor
+    			.getSuccessResponseCodes();
+    	if (successResponseCodes.contains(response.responseCode)) {
+    		response.status = Status.UP;
+    	} else {
+    		response.status = Status.ERROR;
+    	}
+
+    	return response;
+    }
+
+    private ExecutorService exec = null;
+
+    public synchronized Future<StatusResponse> launchCheck(final Site site, final SiteMonitorDescriptor descriptor) throws IOException {
+    	if (exec == null) {
+    		exec = Executors.newCachedThreadPool();
+    		// exec = Executors.newSingleThreadExecutor();
+    	}
+    	return exec.submit(new Callable<StatusResponse>() {
+
+			public StatusResponse call() throws Exception {
+				return checkURL(site, descriptor);
+			}
+
+        });
+    }
+
     /**
      * Performs the web site monitoring by checking the response code of the
      * site's URL.
@@ -157,32 +214,47 @@ public final boolean perform(final AbstractBuild<?, ?> build,
         boolean hasFailure = false;
         for (Site site : mSites) {
 
+        	LOGGER.log(Level.FINE, "Checking site {}", site.getUrl());
+
             Integer responseCode = null;
             Status status;
             String note = "";
             HttpURLConnection connection = null;
-
+            StatusResponse response;
+
+            Future<StatusResponse> future = null;
+
             try {
-                connection = getConnection(site.getUrl());
-                connection.setConnectTimeout(descriptor.getTimeout()
-                        * MILLISECS_IN_SECS);
-                responseCode = connection.getResponseCode();
-
-                List<Integer> successResponseCodes = descriptor
-                        .getSuccessResponseCodes();
-                if (successResponseCodes.contains(responseCode)) {
-                    status = Status.UP;
-                } else {
-                    status = Status.ERROR;
+            	try {
+                	future = launchCheck(site, descriptor);
+                	LOGGER.log(Level.FINE, "Waiting for future for {}", site.getUrl());
+                	response = future.get(descriptor.getTimeout()+1, TimeUnit.SECONDS);
+                	status = response.status;
+                	responseCode = response.responseCode;
+                } catch (ExecutionException ee) {
+                	LOGGER.log(Level.WARNING, "ExecutionException in checking", ee);
+                    status = Status.DOWN;
+                	if (ee.getCause() instanceof Exception) {
+                		throw (Exception)ee.getCause();
+                	}
+                	throw ee;
                 }
+            } catch (TimeoutException toe) {
+                listener.getLogger().println("Timeout: " + toe);
+                status = Status.DOWN;            	
             } catch (SocketTimeoutException ste) {
-                listener.getLogger().println(ste + " - " + ste.getMessage());
+                listener.getLogger().println("Socket Timeout: " + ste + " - " + ste.getMessage());
                 status = Status.DOWN;
             } catch (Exception e) {
+            	LOGGER.log(Level.WARNING, "Exception in checking", e);
                 note = e + " - " + e.getMessage();
                 listener.getLogger().println(note);
                 status = Status.EXCEPTION;
             } finally {
+            	if (future != null) {
+            		LOGGER.log(Level.FINER, "Cancelling future {}", future);
+            		future.cancel(true);
+            	}
                 if (connection != null) {
                     connection.disconnect();
                 }

src/main/resources/hudson/plugins/sitemonitor/SiteMonitorRecorder/config_ja.properties

@@ -1,3 +1,3 @@
-URL=URL
-Delete=\u524a\u9664
-Add\ URL...=URL\u3092\u8ffd\u52a0...
+URL=URL
+Delete=\u524a\u9664
+Add\ URL...=URL\u3092\u8ffd\u52a0...

src/main/resources/hudson/plugins/sitemonitor/SiteMonitorRecorder/global_ja.properties

@@ -1,3 +1,3 @@
-Site\ Monitor=\u30b5\u30a4\u30c8\u30e2\u30cb\u30bf\u30fc
-Success\ Response\ Codes=\u6210\u529f\u6642\u30b9\u30c6\u30fc\u30bf\u30b9\u30b3\u30fc\u30c9
-Timeout\ in\ seconds=\u30bf\u30a4\u30e0\u30a2\u30a6\u30c8(\u79d2)
+Site\ Monitor=\u30b5\u30a4\u30c8\u30e2\u30cb\u30bf\u30fc
+Success\ Response\ Codes=\u6210\u529f\u6642\u30b9\u30c6\u30fc\u30bf\u30b9\u30b3\u30fc\u30c9
+Timeout\ in\ seconds=\u30bf\u30a4\u30e0\u30a2\u30a6\u30c8(\u79d2)

src/main/resources/hudson/plugins/sitemonitor/SiteMonitorRootAction/index_ja.properties

@@ -1,4 +1,4 @@
-Site\ Monitor\ Report=\u30b5\u30a4\u30c8\u30e2\u30cb\u30bf\u30fc\u30ec\u30dd\u30fc\u30c8
-URL=URL
-Response\ Code=\u30b9\u30c6\u30fc\u30bf\u30b9\u30b3\u30fc\u30c9
-Status=\u30b9\u30c6\u30fc\u30bf\u30b9
+Site\ Monitor\ Report=\u30b5\u30a4\u30c8\u30e2\u30cb\u30bf\u30fc\u30ec\u30dd\u30fc\u30c8
+URL=URL
+Response\ Code=\u30b9\u30c6\u30fc\u30bf\u30b9\u30b3\u30fc\u30c9
+Status=\u30b9\u30c6\u30fc\u30bf\u30b9

src/main/resources/hudson/plugins/sitemonitor/SiteMonitorRootAction/summary_ja.properties

@@ -1 +1 @@
-Site\ Monitor\ Report=\u30b5\u30a4\u30c8\u30e2\u30cb\u30bf\u30fc\u30ec\u30dd\u30fc\u30c8
+Site\ Monitor\ Report=\u30b5\u30a4\u30c8\u30e2\u30cb\u30bf\u30fc\u30ec\u30dd\u30fc\u30c8

src/main/webapp/successresponsecodes_ja.html

@@ -1,2 +1,2 @@
-Webサイトが、アップ状態 かつ エラーなしの状態であることを示すHTTPステータスコードの値を、カンマ区切りで記述します。
-詳細情報は、<a href="http://ja.wikipedia.org/wiki/HTTP%E3%82%B9%E3%83%86%E3%83%BC%E3%82%BF%E3%82%B9%E3%82%B3%E3%83%BC%E3%83%89">HTTPステータスコード一覧</a>で確認できます。
+Webサイトが、アップ状態 かつ エラーなしの状態であることを示すHTTPステータスコードの値を、カンマ区切りで記述します。
+詳細情報は、<a href="http://ja.wikipedia.org/wiki/HTTP%E3%82%B9%E3%83%86%E3%83%BC%E3%82%BF%E3%82%B9%E3%82%B3%E3%83%BC%E3%83%89">HTTPステータスコード一覧</a>で確認できます。

src/main/webapp/url_ja.html

@@ -1 +1 @@
-監視対象のWebサイトのURLアドレスを記述します。 URLは http:// または https:// から記述してください。
+監視対象のWebサイトのURLアドレスを記述します。 URLは http:// または https:// から記述してください。

src/test/java/hudson/plugins/sitemonitor/SiteMonitorRecorderTest.java

@@ -1,10 +1,18 @@
 package hudson.plugins.sitemonitor;
 
 import hudson.plugins.sitemonitor.SiteMonitorRecorder;
+import hudson.plugins.sitemonitor.SiteMonitorRecorder.StatusResponse;
 import hudson.plugins.sitemonitor.model.Site;
+import hudson.plugins.sitemonitor.model.Status;
 
+import java.net.UnknownHostException;
 import java.util.ArrayList;
+import java.util.Arrays;
 import java.util.List;
+import java.util.concurrent.ExecutionException;
+import java.util.concurrent.Future;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.TimeoutException;
 
 import junit.framework.TestCase;
 
@@ -19,4 +27,103 @@ public void testGetSitesShouldGiveExpectedSites() {
         builder = new SiteMonitorRecorder(sites);
         assertEquals(sites, builder.getSites());
     }
+
+    private class MockSMD extends SiteMonitorDescriptor {
+    	public MockSMD(List<Integer> okCodes, Integer timeout) {
+    		super(okCodes, timeout);
+    	}
+    }
+
+    public void testLaunch() throws Exception {
+        List<Site> sites = new ArrayList<Site>();
+        Site site0 = new Site("http://hudson-ci.org");
+		sites.add(site0);
+        builder = new SiteMonitorRecorder(sites);
+
+        List<Integer> okCodes = Arrays.asList(200);
+        Integer timeout;
+        SiteMonitorDescriptor descriptor;
+
+        timeout = 1;
+        descriptor = new MockSMD(okCodes, timeout);
+
+        try {
+        	Future<StatusResponse> f = builder.launchCheck(site0, descriptor);
+        	StatusResponse s = f.get(1, TimeUnit.MICROSECONDS);
+        } catch (Exception e) {
+        	assertTrue("Expected exception", e instanceof TimeoutException);
+        }
+
+        timeout = 30;
+        descriptor = new MockSMD(okCodes, timeout);
+
+    	Future<StatusResponse> f = builder.launchCheck(site0, descriptor);
+    	StatusResponse s = f.get(5, TimeUnit.MINUTES);
+    	assertTrue("got result within 5 minutes", true);
+
+    }
+
+    public void testLaunch_badUrl() throws Exception {
+        List<Site> sites = new ArrayList<Site>();
+        Site site0 = new Site("http://hudson-ci.org/bogusURL");
+		sites.add(site0);
+        builder = new SiteMonitorRecorder(sites);
+
+        List<Integer> okCodes = Arrays.asList(200);
+        Integer timeout;
+        SiteMonitorDescriptor descriptor;
+
+        timeout = 1;
+        descriptor = new MockSMD(okCodes, timeout);
+
+        try {
+        	Future<StatusResponse> f = builder.launchCheck(site0, descriptor);
+        	StatusResponse s = f.get(1, TimeUnit.MICROSECONDS);
+        	assertFalse("got result from bad URL", true);
+        } catch (Exception e) {
+        	assertTrue("Expected exception", e instanceof TimeoutException);
+        }
+
+        timeout = 30;
+        descriptor = new MockSMD(okCodes, timeout);
+
+    	Future<StatusResponse> f = builder.launchCheck(site0, descriptor);
+    	StatusResponse s = f.get(5, TimeUnit.MINUTES);
+    	assertTrue("got result within 5 minutes", true);
+
+    }
+
+    public void testLaunch_badHost() throws Exception {
+        List<Site> sites = new ArrayList<Site>();
+        Site site0 = new Site("http://no.such.host.hudson-ci.org/");
+		sites.add(site0);
+        builder = new SiteMonitorRecorder(sites);
+
+        List<Integer> okCodes = Arrays.asList(200);
+        Integer timeout;
+        SiteMonitorDescriptor descriptor;
+
+        timeout = 1;
+        descriptor = new MockSMD(okCodes, timeout);
+
+        try {
+        	Future<StatusResponse> f = builder.launchCheck(site0, descriptor);
+        	StatusResponse s = f.get(1, TimeUnit.MICROSECONDS);
+        } catch (Exception e) {
+        	assertTrue("Expected exception", e instanceof TimeoutException);
+        }
+
+        timeout = 30;
+        descriptor = new MockSMD(okCodes, timeout);
+
+        try {
+        	Future<StatusResponse> f = builder.launchCheck(site0, descriptor);
+        	StatusResponse s = f.get(5, TimeUnit.MINUTES);
+        	assertFalse("got result from bad URL", true);
+        } catch (Exception e) {
+        	assertTrue("Expected exception", e instanceof ExecutionException);
+        	assertTrue("Expected cause", e.getCause() instanceof UnknownHostException);
+        }
+    }
+
 }