hadoop-common-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From weic...@apache.org
Subject hadoop git commit: HDFS-13486. Backport HDFS-11817 (A faulty node can cause a lease leak and NPE on accessing data) to branch-2.7. Contributed by Kihwal Lee.
Date Mon, 07 May 2018 23:24:48 GMT
Repository: hadoop
Updated Branches:
  refs/heads/branch-2.7 20c6b448a -> e8777342d


HDFS-13486. Backport HDFS-11817 (A faulty node can cause a lease leak and NPE on accessing
data) to branch-2.7. Contributed by Kihwal Lee.


Project: http://git-wip-us.apache.org/repos/asf/hadoop/repo
Commit: http://git-wip-us.apache.org/repos/asf/hadoop/commit/e8777342
Tree: http://git-wip-us.apache.org/repos/asf/hadoop/tree/e8777342
Diff: http://git-wip-us.apache.org/repos/asf/hadoop/diff/e8777342

Branch: refs/heads/branch-2.7
Commit: e8777342d3d916276a06bbd8cb2257159402bd58
Parents: 20c6b44
Author: Wei-Chiu Chuang <weichiu@apache.org>
Authored: Mon May 7 16:22:34 2018 -0700
Committer: Wei-Chiu Chuang <weichiu@apache.org>
Committed: Mon May 7 16:23:41 2018 -0700

----------------------------------------------------------------------
 .../BlockInfoContiguousUnderConstruction.java   | 66 +++++++++++++-------
 .../server/blockmanagement/DatanodeManager.java |  3 +-
 .../hdfs/server/namenode/FSNamesystem.java      |  5 +-
 .../hdfs/server/namenode/LeaseManager.java      | 18 +++++-
 .../TestBlockInfoUnderConstruction.java         |  8 +--
 .../namenode/TestBlockUnderConstruction.java    | 42 +++++++++++++
 .../TestCommitBlockSynchronization.java         |  2 +-
 7 files changed, 111 insertions(+), 33 deletions(-)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/hadoop/blob/e8777342/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockInfoContiguousUnderConstruction.java
----------------------------------------------------------------------
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockInfoContiguousUnderConstruction.java
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockInfoContiguousUnderConstruction.java
index e98ab7e..6b9b7c2 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockInfoContiguousUnderConstruction.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockInfoContiguousUnderConstruction.java
@@ -190,12 +190,23 @@ public class BlockInfoContiguousUnderConstruction extends BlockInfoContiguous
{
 
   /** Set expected locations */
   public void setExpectedLocations(DatanodeStorageInfo[] targets) {
-    int numLocations = targets == null ? 0 : targets.length;
+    if (targets == null) {
+      return;
+    }
+    int numLocations = 0;
+    for (DatanodeStorageInfo target : targets) {
+      if (target != null) {
+        numLocations++;
+      }
+    }
+
     this.replicas = new ArrayList<ReplicaUnderConstruction>(numLocations);
-    for(int i = 0; i < numLocations; i++)
+    for(int i = 0; i < targets.length; i++) {
+      // Only store non-null DatanodeStorageInfo.
       if (targets[i] != null) {
         replicas.add(new ReplicaUnderConstruction(this, targets[i], ReplicaState.RBW));
       }
+    }
   }
 
   /**
@@ -286,10 +297,15 @@ public class BlockInfoContiguousUnderConstruction extends BlockInfoContiguous
{
    * Initialize lease recovery for this block.
    * Find the first alive data-node starting from the previous primary and
    * make it primary.
+   * @param recoveryId Recovery ID (new gen stamp)
+   * @param startRecovery Issue recovery command to datanode if true.
    */
-  public void initializeBlockRecovery(long recoveryId) {
+  public void initializeBlockRecovery(long recoveryId, boolean startRecovery) {
     setBlockUCState(BlockUCState.UNDER_RECOVERY);
     blockRecoveryId = recoveryId;
+    if (!startRecovery) {
+      return;
+    }
     if (replicas.size() == 0) {
       NameNode.blockStateChangeLog.warn("BLOCK*"
         + " BlockInfoUnderConstruction.initLeaseRecovery:"
@@ -337,27 +353,33 @@ public class BlockInfoContiguousUnderConstruction extends BlockInfoContiguous
{
   void addReplicaIfNotPresent(DatanodeStorageInfo storage,
                      Block block,
                      ReplicaState rState) {
-    Iterator<ReplicaUnderConstruction> it = replicas.iterator();
-    while (it.hasNext()) {
-      ReplicaUnderConstruction r = it.next();
-      DatanodeStorageInfo expectedLocation = r.getExpectedStorageLocation();
-      if(expectedLocation == storage) {
-        // Record the gen stamp from the report
-        r.setGenerationStamp(block.getGenerationStamp());
-        return;
-      } else if (expectedLocation != null &&
-                 expectedLocation.getDatanodeDescriptor() ==
-                     storage.getDatanodeDescriptor()) {
-
-        // The Datanode reported that the block is on a different storage
-        // than the one chosen by BlockPlacementPolicy. This can occur as
-        // we allow Datanodes to choose the target storage. Update our
-        // state by removing the stale entry and adding a new one.
-        it.remove();
-        break;
+    if (replicas == null) {
+      replicas = new ArrayList<ReplicaUnderConstruction>(1);
+      replicas.add(new ReplicaUnderConstruction(block, storage,
+          rState));
+    } else {
+      Iterator<ReplicaUnderConstruction> it = replicas.iterator();
+      while (it.hasNext()) {
+        ReplicaUnderConstruction r = it.next();
+        DatanodeStorageInfo expectedLocation = r.getExpectedStorageLocation();
+        if (expectedLocation == storage) {
+          // Record the gen stamp from the report
+          r.setGenerationStamp(block.getGenerationStamp());
+          return;
+        } else if (expectedLocation != null
+            && expectedLocation.getDatanodeDescriptor() ==
+            storage.getDatanodeDescriptor()) {
+
+          // The Datanode reported that the block is on a different storage
+          // than the one chosen by BlockPlacementPolicy. This can occur as
+          // we allow Datanodes to choose the target storage. Update our
+          // state by removing the stale entry and adding a new one.
+          it.remove();
+          break;
+        }
       }
+      replicas.add(new ReplicaUnderConstruction(block, storage, rState));
     }
-    replicas.add(new ReplicaUnderConstruction(block, storage, rState));
   }
 
   @Override // BlockInfo

http://git-wip-us.apache.org/repos/asf/hadoop/blob/e8777342/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/DatanodeManager.java
----------------------------------------------------------------------
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/DatanodeManager.java
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/DatanodeManager.java
index 5d82186..e449585 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/DatanodeManager.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/DatanodeManager.java
@@ -505,10 +505,11 @@ public class DatanodeManager {
       DatanodeID[] datanodeID, String[] storageIDs,
       String format, Object... args) throws UnregisteredNodeException {
     if (datanodeID.length != storageIDs.length) {
+      // Error for pre-2.0.0-alpha clients.
       final String err = (storageIDs.length == 0?
           "Missing storageIDs: It is likely that the HDFS client,"
           + " who made this call, is running in an older version of Hadoop"
-          + " which does not support storageIDs."
+          + "(pre-2.0.0-alpha)  which does not support storageIDs."
           : "Length mismatched: storageIDs.length=" + storageIDs.length + " != "
           ) + " datanodeID.length=" + datanodeID.length;
       throw new HadoopIllegalArgumentException(

http://git-wip-us.apache.org/repos/asf/hadoop/blob/e8777342/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
----------------------------------------------------------------------
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
index b458900..f3d9484 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
@@ -2205,7 +2205,8 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
           truncatedBlockUC.getTruncateBlock().getNumBytes(), truncatedBlockUC);
     }
     if (shouldRecoverNow) {
-      truncatedBlockUC.initializeBlockRecovery(newBlock.getGenerationStamp());
+      truncatedBlockUC.initializeBlockRecovery(newBlock.getGenerationStamp(),
+          true);
     }
 
     return newBlock;
@@ -4105,7 +4106,7 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
       } else if(truncateRecovery) {
         recoveryBlock.setGenerationStamp(blockRecoveryId);
       }
-      uc.initializeBlockRecovery(blockRecoveryId);
+      uc.initializeBlockRecovery(blockRecoveryId, true);
       leaseManager.renewLease(lease);
       // Cannot close file right now, since the last block requires recovery.
       // This may potentially cause infinite loop in lease recovery

http://git-wip-us.apache.org/repos/asf/hadoop/blob/e8777342/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/LeaseManager.java
----------------------------------------------------------------------
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/LeaseManager.java
b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/LeaseManager.java
index fd7c42e..5270d0b 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/LeaseManager.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/LeaseManager.java
@@ -491,8 +491,20 @@ public class LeaseManager {
         try {
           INodesInPath iip = fsnamesystem.getFSDirectory().getINodesInPath(p,
               true);
-          boolean completed = fsnamesystem.internalReleaseLease(leaseToCheck, p,
-              iip, HdfsServerConstants.NAMENODE_LEASE_HOLDER);
+          // Sanity check to make sure the path is correct
+          if (!p.startsWith("/")) {
+            throw new IOException("Invalid path in the lease " + p);
+          }
+          boolean completed = false;
+          try {
+            completed = fsnamesystem.internalReleaseLease(
+                leaseToCheck, p, iip,
+                HdfsServerConstants.NAMENODE_LEASE_HOLDER);
+          } catch (IOException e) {
+            LOG.warn("Cannot release the path " + p + " in the lease "
+                + leaseToCheck + ". It will be retried.", e);
+            continue;
+          }
           if (LOG.isDebugEnabled()) {
             if (completed) {
               LOG.debug("Lease recovery for " + p + " is complete. File closed.");
@@ -505,7 +517,7 @@ public class LeaseManager {
             needSync = true;
           }
         } catch (IOException e) {
-          LOG.error("Cannot release the path " + p + " in the lease "
+          LOG.warn("Removing lease with an invalid path: " + p + ","
               + leaseToCheck, e);
           removing.add(p);
         }

http://git-wip-us.apache.org/repos/asf/hadoop/blob/e8777342/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestBlockInfoUnderConstruction.java
----------------------------------------------------------------------
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestBlockInfoUnderConstruction.java
b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestBlockInfoUnderConstruction.java
index a7ba293..c72a356 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestBlockInfoUnderConstruction.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestBlockInfoUnderConstruction.java
@@ -50,7 +50,7 @@ public class TestBlockInfoUnderConstruction {
     DFSTestUtil.resetLastUpdatesWithOffset(dd1, -3 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd2, -1 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd3, -2 * 1000);
-    blockInfo.initializeBlockRecovery(1);
+    blockInfo.initializeBlockRecovery(1, true);
     BlockInfoContiguousUnderConstruction[] blockInfoRecovery = dd2.getLeaseRecoveryCommand(1);
     assertEquals(blockInfoRecovery[0], blockInfo);
 
@@ -58,7 +58,7 @@ public class TestBlockInfoUnderConstruction {
     DFSTestUtil.resetLastUpdatesWithOffset(dd1, -2 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd2, -1 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd3, -3 * 1000);
-    blockInfo.initializeBlockRecovery(2);
+    blockInfo.initializeBlockRecovery(2, true);
     blockInfoRecovery = dd1.getLeaseRecoveryCommand(1);
     assertEquals(blockInfoRecovery[0], blockInfo);
 
@@ -66,7 +66,7 @@ public class TestBlockInfoUnderConstruction {
     DFSTestUtil.resetLastUpdatesWithOffset(dd1, -2 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd2, -1 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd3, -3 * 1000);
-    blockInfo.initializeBlockRecovery(3);
+    blockInfo.initializeBlockRecovery(3, true);
     blockInfoRecovery = dd3.getLeaseRecoveryCommand(1);
     assertEquals(blockInfoRecovery[0], blockInfo);
 
@@ -75,7 +75,7 @@ public class TestBlockInfoUnderConstruction {
     DFSTestUtil.resetLastUpdatesWithOffset(dd1, -2 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd2, -1 * 1000);
     DFSTestUtil.resetLastUpdatesWithOffset(dd3, 0);
-    blockInfo.initializeBlockRecovery(3);
+    blockInfo.initializeBlockRecovery(3, true);
     blockInfoRecovery = dd3.getLeaseRecoveryCommand(1);
     assertEquals(blockInfoRecovery[0], blockInfo);
   }

http://git-wip-us.apache.org/repos/asf/hadoop/blob/e8777342/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBlockUnderConstruction.java
----------------------------------------------------------------------
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBlockUnderConstruction.java
b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBlockUnderConstruction.java
index 1fbe160..18215d6 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBlockUnderConstruction.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBlockUnderConstruction.java
@@ -37,6 +37,7 @@ import org.apache.hadoop.hdfs.protocol.LocatedBlock;
 import org.apache.hadoop.hdfs.protocol.LocatedBlocks;
 import org.apache.hadoop.hdfs.server.blockmanagement.BlockInfoContiguous;
 import org.apache.hadoop.hdfs.server.blockmanagement.BlockInfoContiguousUnderConstruction;
+import org.apache.hadoop.hdfs.server.blockmanagement.BlockManager;
 import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.BlockUCState;
 import org.apache.hadoop.hdfs.server.protocol.NamenodeProtocols;
 import org.junit.AfterClass;
@@ -181,4 +182,45 @@ public class TestBlockUnderConstruction {
     // close file
     out.close();
   }
+
+  /**
+   * A storage ID can be invalid if the storage failed or the node
+   * reregisters. When the node heart-beats, the storage report in it
+   * causes storage volumes to be added back. An invalid storage ID
+   * should not cause an NPE.
+   */
+  @Test
+  public void testEmptyExpectedLocations() throws Exception {
+    final NamenodeProtocols namenode = cluster.getNameNodeRpc();
+    final FSNamesystem fsn = cluster.getNamesystem();
+    final BlockManager bm = fsn.getBlockManager();
+    final Path p = new Path(BASE_DIR, "file2.dat");
+    final String src = p.toString();
+    final FSDataOutputStream out = TestFileCreation.createFile(hdfs, p, 1);
+    writeFile(p, out, 256);
+    out.hflush();
+
+    // make sure the block is readable
+    LocatedBlocks lbs = namenode.getBlockLocations(src, 0, 256);
+    LocatedBlock lastLB = lbs.getLocatedBlocks().get(0);
+    final Block b = lastLB.getBlock().getLocalBlock();
+
+    // fake a block recovery
+    long blockRecoveryId = fsn.getBlockIdManager().nextGenerationStamp(false);
+    BlockInfoContiguousUnderConstruction uc = bm.getStoredBlock(b).
+        convertToBlockUnderConstruction(BlockUCState.UNDER_CONSTRUCTION, null);
+    uc.initializeBlockRecovery(blockRecoveryId, false);
+
+    try {
+      String[] storages = { "invalid-storage-id1" };
+      fsn.commitBlockSynchronization(lastLB.getBlock(), blockRecoveryId, 256L,
+          true, false, lastLB.getLocations(), storages);
+    } catch (java.lang.IllegalStateException ise) {
+       // Although a failure is expected as of now, future commit policy
+       // changes may make it not fail. This is not critical to the test.
+    }
+
+    // Invalid storage should not trigger an exception.
+    lbs = namenode.getBlockLocations(src, 0, 256);
+  }
 }

http://git-wip-us.apache.org/repos/asf/hadoop/blob/e8777342/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestCommitBlockSynchronization.java
----------------------------------------------------------------------
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestCommitBlockSynchronization.java
b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestCommitBlockSynchronization.java
index c29383a..aa8ed31 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestCommitBlockSynchronization.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestCommitBlockSynchronization.java
@@ -68,7 +68,7 @@ public class TestCommitBlockSynchronization {
         block, (short) 1, HdfsServerConstants.BlockUCState.UNDER_CONSTRUCTION, targets);
     blockInfo.setBlockCollection(file);
     blockInfo.setGenerationStamp(genStamp);
-    blockInfo.initializeBlockRecovery(genStamp);
+    blockInfo.initializeBlockRecovery(genStamp, true);
     doReturn(true).when(file).removeLastBlock(any(Block.class));
     doReturn(true).when(file).isUnderConstruction();
     doReturn(new BlockInfoContiguous[1]).when(file).getBlocks();


---------------------------------------------------------------------
To unsubscribe, e-mail: common-commits-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-commits-help@hadoop.apache.org


Mime
View raw message