HDFS-9072. Fix random failures in TestJMXGet. Contributed by J.Andreina.

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

@@ -40,6 +40,8 @@ Release 2.7.3 - UNRELEASED
     HDFS-9383. TestByteArrayManager#testByteArrayManager fails.
     (szetszwo via kihwal)
 
+    HDFS-9072. Fix random failures in TestJMXGet (J.Andreina via kihwal)
+
     HDFS-6481. DatanodeManager#getDatanodeStorageInfos() should check the
     length of storageIDs. (szetszwo via Arpit Agarwal)
 

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

@@ -27,6 +27,7 @@ import com.google.common.collect.Lists;
 import com.google.common.collect.Maps;
 
 import org.apache.commons.io.FileUtils;
+import org.apache.commons.lang.UnhandledException;
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.conf.Configuration;
@@ -86,6 +87,7 @@ import org.apache.hadoop.hdfs.server.protocol.ReceivedDeletedBlockInfo;
 import org.apache.hadoop.hdfs.server.protocol.ReceivedDeletedBlockInfo.BlockStatus;
 import org.apache.hadoop.hdfs.server.protocol.StorageReceivedDeletedBlocks;
 import org.apache.hadoop.hdfs.tools.DFSAdmin;
+import org.apache.hadoop.hdfs.tools.JMXGet;
 import org.apache.hadoop.io.IOUtils;
 import org.apache.hadoop.io.nativeio.NativeIO;
 import org.apache.hadoop.net.NetUtils;
@@ -1811,4 +1813,22 @@ public class DFSTestUtil {
     lastBlock.setNumBytes(len);
     return lastBlock;
   }
+
+  public static void waitForMetric(final JMXGet jmx, final String metricName, final int expectedValue)
+      throws TimeoutException, InterruptedException {
+    GenericTestUtils.waitFor(new Supplier<Boolean>() {
+      @Override
+      public Boolean get() {
+        try {
+          final int currentValue = Integer.parseInt(jmx.getValue(metricName));
+          LOG.info("Waiting for " + metricName +
+                       " to reach value " + expectedValue +
+                       ", current value = " + currentValue);
+          return currentValue == expectedValue;
+        } catch (Exception e) {
+          throw new UnhandledException("Test failed due to unexpected exception", e);
+        }
+      }
+    }, 1000, Integer.MAX_VALUE);
+  }
 }

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/tools/TestJMXGet.java

@@ -41,6 +41,7 @@ import org.apache.hadoop.fs.FSDataOutputStream;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.FileUtil;
 import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
 import org.apache.hadoop.hdfs.tools.JMXGet;
@@ -114,10 +115,12 @@ public class TestJMXGet {
     assertTrue("error printAllValues", checkPrintAllValues(jmx));
 
     //get some data from different source
+    DFSTestUtil.waitForMetric(jmx, "NumLiveDataNodes", numDatanodes);
     assertEquals(numDatanodes, Integer.parseInt(
         jmx.getValue("NumLiveDataNodes")));
     assertGauge("CorruptBlocks", Long.parseLong(jmx.getValue("CorruptBlocks")),
                 getMetrics("FSNamesystem"));
+    DFSTestUtil.waitForMetric(jmx, "NumOpenConnections", numDatanodes);
     assertEquals(numDatanodes, Integer.parseInt(
         jmx.getValue("NumOpenConnections")));
 
@@ -161,6 +164,7 @@ public class TestJMXGet {
     String serviceName = "DataNode";
     jmx.setService(serviceName);
     jmx.init();
+    DFSTestUtil.waitForMetric(jmx, "BytesWritten", fileSize);
     assertEquals(fileSize, Integer.parseInt(jmx.getValue("BytesWritten")));
 
     cluster.shutdown();