HDFS-9429. Tests in TestDFSAdminWithHA intermittently fail with EOFException (Xiao Chen via Colin P. McCabe)

(cherry picked from commit 53e3bf7e70)
(cherry picked from commit 9b516a2a05)

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

@@ -1553,6 +1553,9 @@ Release 2.8.0 - UNRELEASED
     HDFS-6533. TestBPOfferService#testBasicFunctionalitytest fails
     intermittently. (Wei-Chiu Chuang via Arpit Agarwal)
 
+    HDFS-9429. Tests in TestDFSAdminWithHA intermittently fail with
+    EOFException (Xiao Chen via Colin P. McCabe)
+
 Release 2.7.3 - UNRELEASED
 
   INCOMPATIBLE CHANGES

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestRollingUpgrade.java

@@ -168,6 +168,7 @@ public class TestRollingUpgrade {
 
     final Configuration conf = new HdfsConfiguration();
     final MiniJournalCluster mjc = new MiniJournalCluster.Builder(conf).build();
+    mjc.waitActive();
     setConf(conf, nn1Dir, mjc);
 
     {

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestRollingUpgradeRollback.java

@@ -153,6 +153,7 @@ public class TestRollingUpgradeRollback {
     try {
       mjc = new MiniJournalCluster.Builder(conf).numJournalNodes(
           NUM_JOURNAL_NODES).build();
+      mjc.waitActive();
       conf.set(DFSConfigKeys.DFS_NAMENODE_EDITS_DIR_KEY, mjc
           .getQuorumJournalURI(JOURNAL_ID).toString());
       cluster = new MiniDFSCluster.Builder(conf).numDataNodes(0).build();

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

@@ -17,26 +17,34 @@
  */
 package org.apache.hadoop.hdfs.qjournal;
 
+import static org.apache.hadoop.hdfs.qjournal.QJMTestUtil.FAKE_NSINFO;
+import static org.junit.Assert.fail;
+
 import java.io.File;
 import java.io.IOException;
 import java.net.InetSocketAddress;
 import java.net.URI;
 import java.net.URISyntaxException;
 import java.util.List;
+import java.util.concurrent.TimeoutException;
 
+import com.google.common.base.Supplier;
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.FileUtil;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
+import org.apache.hadoop.hdfs.qjournal.client.QuorumJournalManager;
 import org.apache.hadoop.hdfs.qjournal.server.JournalNode;
 import org.apache.hadoop.net.NetUtils;
 
 import com.google.common.base.Joiner;
 import com.google.common.collect.Lists;
+import org.apache.hadoop.test.GenericTestUtils;
 
 public class MiniJournalCluster {
+  public static final String CLUSTER_WAITACTIVE_URI = "waitactive";
   public static class Builder {
     private String baseDir;
     private int numJournalNodes = 3;
@@ -217,4 +225,34 @@ public class MiniJournalCluster {
     return nodes.length;
   }
 
+  /**
+   * Wait until all the journalnodes start.
+   */
+  public void waitActive() throws IOException {
+    for (int i = 0; i < nodes.length; i++) {
+      final int index = i;
+      try {
+        GenericTestUtils.waitFor(new Supplier<Boolean>() {
+          // wait until all JN's IPC server is running
+          @Override public Boolean get() {
+            try {
+              QuorumJournalManager qjm =
+                  new QuorumJournalManager(nodes[index].node.getConf(),
+                      getQuorumJournalURI(CLUSTER_WAITACTIVE_URI), FAKE_NSINFO);
+              qjm.hasSomeData();
+              qjm.close();
+            } catch (IOException e) {
+              // Exception from IPC call, likely due to server not ready yet.
+              return false;
+            }
+            return true;
+          }
+        }, 50, 3000);
+      } catch (TimeoutException e) {
+        fail("Time out while waiting for journal node " + index + " to start.");
+      } catch (InterruptedException ite) {
+        LOG.warn("Thread interrupted when waiting for node start", ite);
+      }
+    }
+  }
 }

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

@@ -92,6 +92,7 @@ public class MiniQJMHACluster {
         // start 3 journal nodes
         journalCluster = new MiniJournalCluster.Builder(conf).format(true)
             .build();
+        journalCluster.waitActive();
         URI journalURI = journalCluster.getQuorumJournalURI(NAMESERVICE);
 
         // start cluster with 2 NameNodes

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

@@ -36,6 +36,7 @@ public class TestMiniJournalCluster {
     Configuration conf = new Configuration();
     MiniJournalCluster c = new MiniJournalCluster.Builder(conf)
       .build();
+    c.waitActive();
     try {
       URI uri = c.getQuorumJournalURI("myjournal");
       String[] addrs = uri.getAuthority().split(";");

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

@@ -52,6 +52,7 @@ public class TestNNWithQJM {
   @Before
   public void startJNs() throws Exception {
     mjc = new MiniJournalCluster.Builder(conf).build();
+    mjc.waitActive();
   }
   
   @After

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

@@ -212,6 +212,7 @@ public class TestSecureNNWithQJM {
   private void startCluster() throws IOException {
     mjc = new MiniJournalCluster.Builder(conf)
       .build();
+    mjc.waitActive();
     conf.set(DFS_NAMENODE_EDITS_DIR_KEY,
       mjc.getQuorumJournalURI("myjournal").toString());
     cluster = new MiniDFSCluster.Builder(conf)

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

@@ -51,6 +51,7 @@ public class TestEpochsAreUnique {
   public void testSingleThreaded() throws IOException {
     Configuration conf = new Configuration();
     MiniJournalCluster cluster = new MiniJournalCluster.Builder(conf).build();
+    cluster.waitActive();
     URI uri = cluster.getQuorumJournalURI(JID);
     QuorumJournalManager qjm = new QuorumJournalManager(
         conf, uri, FAKE_NSINFO);

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

@@ -98,6 +98,7 @@ public class TestQJMWithFaults {
   private static long determineMaxIpcNumber() throws Exception {
     Configuration conf = new Configuration();
     MiniJournalCluster cluster = new MiniJournalCluster.Builder(conf).build();
+    cluster.waitActive();
     QuorumJournalManager qjm = null;
     long ret;
     try {
@@ -146,6 +147,7 @@ public class TestQJMWithFaults {
         
         MiniJournalCluster cluster = new MiniJournalCluster.Builder(conf)
           .build();
+        cluster.waitActive();
         QuorumJournalManager qjm = null;
         try {
           qjm = createInjectableQJM(cluster);
@@ -218,6 +220,7 @@ public class TestQJMWithFaults {
     
     MiniJournalCluster cluster = new MiniJournalCluster.Builder(conf)
       .build();
+    cluster.waitActive();
     
     // Format the cluster using a non-faulty QJM.
     QuorumJournalManager qjmForInitialFormat =

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

@@ -94,6 +94,7 @@ public class TestQuorumJournalManager {
     
     cluster = new MiniJournalCluster.Builder(conf)
       .build();
+    cluster.waitActive();
     
     qjm = createSpyingQJM();
     spies = qjm.getLoggerSetForTests().getLoggersForTests();

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/qjournal/server/TestJournalNodeMXBean.java

@@ -19,6 +19,7 @@ package org.apache.hadoop.hdfs.qjournal.server;
 
 import static org.junit.Assert.assertEquals;
 import static org.junit.Assert.assertFalse;
+import static org.junit.Assert.assertTrue;
 
 import java.io.IOException;
 import java.lang.management.ManagementFactory;
@@ -52,6 +53,7 @@ public class TestJournalNodeMXBean {
     // start 1 journal node
     jCluster = new MiniJournalCluster.Builder(new Configuration()).format(true)
         .numJournalNodes(NUM_JN).build();
+    jCluster.waitActive();
     jn = jCluster.getJournalNode(0);
   }
   
@@ -89,19 +91,19 @@ public class TestJournalNodeMXBean {
     Map<String, String> infoMap = new HashMap<String, String>();
     infoMap.put("Formatted", "true");
     jMap.put(NAMESERVICE, infoMap);
+    Map<String, String> infoMap1 = new HashMap<>();
+    infoMap1.put("Formatted", "false");
+    jMap.put(MiniJournalCluster.CLUSTER_WAITACTIVE_URI, infoMap1);
     assertEquals(JSON.toString(jMap), journalStatus);
     
     // restart journal node without formatting
     jCluster = new MiniJournalCluster.Builder(new Configuration()).format(false)
         .numJournalNodes(NUM_JN).build();
+    jCluster.waitActive();
     jn = jCluster.getJournalNode(0);
     // re-check 
     journalStatus = (String) mbs.getAttribute(mxbeanName, "JournalsStatus");
     assertEquals(jn.getJournalsStatus(), journalStatus);
-    jMap = new HashMap<String, Map<String, String>>();
-    infoMap = new HashMap<String, String>();
-    infoMap.put("Formatted", "true");
-    jMap.put(NAMESERVICE, infoMap);
     assertEquals(JSON.toString(jMap), journalStatus);
   }
 }