HDFS-9347. Invariant assumption in TestQuorumJournalManager.shutdown() is wrong. Contributed by Wei-Chiu Chuang.

Change-Id: Idde8522e11c77f82b376e0c90a9bf33d163d5cb9

hadoop-common-project/hadoop-common/src/test/java/org/apache/hadoop/test/GenericTestUtils.java

@@ -417,6 +417,26 @@ public static void assertValueWithinRange(long expectedMin, long expectedMax,
         + expectedMax + ")", expectedMin <= actual && actual <= expectedMax);
   }
 
+  /**
+   * Determine if there are any threads whose name matches the regex.
+   * @param pattern a Pattern object used to match thread names
+   * @return true if there is any thread that matches the pattern
+   */
+  public static boolean anyThreadMatching(Pattern pattern) {
+    ThreadMXBean threadBean = ManagementFactory.getThreadMXBean();
+
+    ThreadInfo[] infos =
+        threadBean.getThreadInfo(threadBean.getAllThreadIds(), 20);
+    for (ThreadInfo info : infos) {
+      if (info == null)
+        continue;
+      if (pattern.matcher(info.getThreadName()).matches()) {
+        return true;
+      }
+    }
+    return false;
+  }
+
   /**
    * Assert that there are no threads running whose name matches the
    * given regular expression.
@@ -424,18 +444,34 @@ public static void assertValueWithinRange(long expectedMin, long expectedMax,
    */
   public static void assertNoThreadsMatching(String regex) {
     Pattern pattern = Pattern.compile(regex);
-    ThreadMXBean threadBean = ManagementFactory.getThreadMXBean();
+    if (anyThreadMatching(pattern)) {
+      Assert.fail("Leaked thread matches " + regex);
+    }
+  }
 
-    ThreadInfo[] infos = threadBean.getThreadInfo(threadBean.getAllThreadIds(), 20);
-    for (ThreadInfo info : infos) {
-      if (info == null) continue;
-      if (pattern.matcher(info.getThreadName()).matches()) {
-        Assert.fail("Leaked thread: " + info + "\n" +
-            Joiner.on("\n").join(info.getStackTrace()));
-      }
+  /**
+   * Periodically check and wait for any threads whose name match the
+   * given regular expression.
+   *
+   * @param regex the regex to match against.
+   * @param checkEveryMillis time (in milliseconds) between checks.
+   * @param waitForMillis total time (in milliseconds) to wait before throwing
+   *                      a time out exception.
+   * @throws TimeoutException
+   * @throws InterruptedException
+   */
+  public static void waitForThreadTermination(String regex,
+      int checkEveryMillis, final int waitForMillis) throws TimeoutException,
+      InterruptedException {
+    final Pattern pattern = Pattern.compile(regex);
+    waitFor(new Supplier<Boolean>() {
+      @Override public Boolean get() {
+        return !anyThreadMatching(pattern);
       }
+    }, checkEveryMillis, waitForMillis);
   }
 
+
   /**
    * Skip test if native build profile of Maven is not activated.
    * Sub-project using this must set 'runningWithNative' property to true

hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt

@@ -2532,6 +2532,9 @@ Release 2.8.0 - UNRELEASED
     replicas should go through BlockPlacementPolicy
     (J.Andreina via vinayakumarb)
 
+    HDFS-9347. Invariant assumption in TestQuorumJournalManager.shutdown()
+    is wrong. (Wei-Chiu Chuang via zhz)
+
 Release 2.7.3 - UNRELEASED
 
   INCOMPATIBLE CHANGES

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/client/TestQuorumJournalManager.java

@@ -38,6 +38,7 @@
 import java.util.ArrayList;
 import java.util.List;
 import java.util.concurrent.ExecutorService;
+import java.util.concurrent.TimeoutException;
 
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
@@ -105,12 +106,14 @@ public void setup() throws Exception {
   }
 
   @After
-  public void shutdown() throws IOException {
+  public void shutdown() throws IOException, InterruptedException,
+      TimeoutException {
     IOUtils.cleanup(LOG, toClose.toArray(new Closeable[0]));
 
     // Should not leak clients between tests -- this can cause flaky tests.
     // (See HDFS-4643)
-    GenericTestUtils.assertNoThreadsMatching(".*IPC Client.*");
+    // Wait for IPC clients to terminate to avoid flaky tests
+    GenericTestUtils.waitForThreadTermination(".*IPC Client.*", 100, 1000);
     
     if (cluster != null) {
       cluster.shutdown();