HDFS-2434. TestNameNodeMetrics.testCorruptBlock fails intermittently. Contributed by Jing Zhao.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1401423 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -482,7 +482,11 @@ Release 2.0.3-alpha - Unreleased
     HDFS-4072. On file deletion remove corresponding blocks pending
     replications. (Jing Zhao via suresh)
 
-    HDFS-4022. Replication not happening for appended block. (Vinay via umamahesh)
+    HDFS-4022. Replication not happening for appended block.
+    (Vinay via umamahesh)
+
+    HDFS-2434. TestNameNodeMetrics.testCorruptBlock fails intermittently.
+    (Jing Zhao via suresh)
 
 Release 2.0.2-alpha - 2012-09-07 
 

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/metrics/TestNameNodeMetrics.java

@@ -205,6 +205,12 @@ public class TestNameNodeMetrics {
     final Path file = getTestPath("testCorruptBlock");
     createFile(file, 100, (short)2);
     
+    // Disable the heartbeats, so that no corrupted replica
+    // can be fixed
+    for (DataNode dn : cluster.getDataNodes()) {
+      DataNodeTestUtils.setHeartbeatsDisabledForTests(dn, true);
+    }
+    
     // Corrupt first replica of the block
     LocatedBlock block = NameNodeAdapter.getBlockLocations(
         cluster.getNameNode(), file.toString(), 0, 1).get(0);
@@ -215,12 +221,23 @@ public class TestNameNodeMetrics {
     } finally {
       cluster.getNamesystem().writeUnlock();
     }
-    Thread.sleep(1000); // Wait for block to be marked corrupt
+    BlockManagerTestUtil.getComputedDatanodeWork(bm);
+    BlockManagerTestUtil.updateState(bm);
     MetricsRecordBuilder rb = getMetrics(NS_METRICS);
     assertGauge("CorruptBlocks", 1L, rb);
     assertGauge("PendingReplicationBlocks", 1L, rb);
     assertGauge("ScheduledReplicationBlocks", 1L, rb);
+    
     fs.delete(file, true);
+    // During the file deletion, both BlockManager#corruptReplicas and
+    // BlockManager#pendingReplications will be updated, i.e., the records
+    // for the blocks of the deleted file will be removed from both
+    // corruptReplicas and pendingReplications. The corresponding
+    // metrics (CorruptBlocks and PendingReplicationBlocks) will only be updated
+    // when BlockManager#computeDatanodeWork is run where the
+    // BlockManager#udpateState is called. And in
+    // BlockManager#computeDatanodeWork the metric ScheduledReplicationBlocks
+    // will also be updated.
     rb = waitForDnMetricValue(NS_METRICS, "CorruptBlocks", 0L);
     assertGauge("PendingReplicationBlocks", 0L, rb);
     assertGauge("ScheduledReplicationBlocks", 0L, rb);