big-data/hadoop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

HDFS-15359. EC: Allow closing a file with committed blocks. Contributed by Ayush Saxena.

Browse Source
This commit is contained in:
Ayush Saxena 2020-06-05 19:15:05 +05:30
parent 8abff5151a
commit 2326123705
4 changed files with 69 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java
View File

@ -1158,12 +1158,19 @@ public boolean commitOrCompleteLastBlock(BlockCollection bc,
/**
* If IBR is not sent from expected locations yet, add the datanodes to
* pendingReconstruction in order to keep RedundancyMonitor from scheduling
* the block.
* the block. In case of erasure coding blocks, adds only in case there
* isn't any missing node.
*/
public void addExpectedReplicasToPending(BlockInfo blk) {
if (!blk.isStriped()) {
DatanodeStorageInfo[] expectedStorages =
blk.getUnderConstructionFeature().getExpectedStorageLocations();
boolean addForStriped = false;
DatanodeStorageInfo[] expectedStorages =
blk.getUnderConstructionFeature().getExpectedStorageLocations();
if (blk.isStriped()) {
BlockInfoStriped blkStriped = (BlockInfoStriped) blk;
addForStriped =
blkStriped.getRealTotalBlockNum() == expectedStorages.length;
}
if (!blk.isStriped() || addForStriped) {
if (expectedStorages.length - blk.numNodes() > 0) {
ArrayList<DatanodeStorageInfo> pendingNodes = new ArrayList<>();
for (DatanodeStorageInfo storage : expectedStorages) {

22
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/INodeFile.java
View File

@ -374,17 +374,27 @@ static String checkBlockComplete(BlockInfo[] blocks, int i,
if (state == BlockUCState.COMPLETE) {
return null;
}
if (b.isStriped() || i < blocks.length - numCommittedAllowed) {
if (i < blocks.length - numCommittedAllowed) {
return b + " is " + state + " but not COMPLETE";
}
if (state != BlockUCState.COMMITTED) {
return b + " is " + state + " but neither COMPLETE nor COMMITTED";
}
final int numExpectedLocations
= b.getUnderConstructionFeature().getNumExpectedLocations();
if (numExpectedLocations <= minReplication) {
return b + " is " + state + " but numExpectedLocations = "
+ numExpectedLocations + " <= minReplication = " + minReplication;
if (b.isStriped()) {
BlockInfoStriped blkStriped = (BlockInfoStriped) b;
if (b.getUnderConstructionFeature().getNumExpectedLocations()
!= blkStriped.getRealTotalBlockNum()) {
return b + " is a striped block in " + state + " with less then "
+ "required number of blocks.";
}
} else {
final int numExpectedLocations =
b.getUnderConstructionFeature().getNumExpectedLocations();
if (numExpectedLocations <= minReplication) {
return b + " is " + state + " but numExpectedLocations = "
+ numExpectedLocations + " <= minReplication = " + minReplication;
}
}
return null;
}

4
hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml
View File

@ -4903,7 +4903,9 @@
<description>
Normally a file can only be closed with all its blocks are committed.
When this value is set to a positive integer N, a file can be closed
when N blocks are committed and the rest complete.
when N blocks are committed and the rest complete. In case of Erasure Coded
blocks, the committed block shall be allowed only when the block group is
complete. i.e no missing/lost block in the blockgroup.
</description>
</property>

39
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestDistributedFileSystem.java
View File

@ -19,6 +19,7 @@
package org.apache.hadoop.hdfs;
import static org.apache.hadoop.fs.CommonConfigurationKeys.FS_CLIENT_TOPOLOGY_RESOLUTION_ENABLED;
import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_FILE_CLOSE_NUM_COMMITTED_ALLOWED_KEY;
import static org.apache.hadoop.hdfs.client.HdfsClientConfigKeys.DFS_CLIENT_CONTEXT;
import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertFalse;
@ -100,6 +101,7 @@
import org.apache.hadoop.hdfs.protocol.OpenFileEntry;
import org.apache.hadoop.hdfs.protocol.OpenFilesIterator;
import org.apache.hadoop.hdfs.server.datanode.DataNode;
import org.apache.hadoop.hdfs.server.datanode.DataNodeTestUtils;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsDatasetSpi;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsVolumeSpi;
import org.apache.hadoop.hdfs.server.namenode.ErasureCodingPolicyManager;
@ -2105,4 +2107,41 @@ public void testGetECTopologyResultForPolicies() throws Exception {
assertFalse(result.isSupported());
}
}
@Test
public void testECCloseCommittedBlock() throws Exception {
HdfsConfiguration conf = new HdfsConfiguration();
conf.setInt(DFS_NAMENODE_FILE_CLOSE_NUM_COMMITTED_ALLOWED_KEY, 1);
try (MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf)
.numDataNodes(3).build()) {
cluster.waitActive();
final DistributedFileSystem dfs = cluster.getFileSystem();
Path dir = new Path("/dir");
dfs.mkdirs(dir);
dfs.enableErasureCodingPolicy("XOR-2-1-1024k");
dfs.setErasureCodingPolicy(dir, "XOR-2-1-1024k");
try (FSDataOutputStream str = dfs.create(new Path("/dir/file"));) {
for (int i = 0; i < 1024 * 1024 * 4; i++) {
str.write(i);
}
DataNodeTestUtils.pauseIBR(cluster.getDataNodes().get(0));
DataNodeTestUtils.pauseIBR(cluster.getDataNodes().get(1));
}
DataNodeTestUtils.resumeIBR(cluster.getDataNodes().get(0));
DataNodeTestUtils.resumeIBR(cluster.getDataNodes().get(1));
// Check if the blockgroup isn't complete then file close shouldn't be
// success with block in committed state.
cluster.getDataNodes().get(0).shutdown();
FSDataOutputStream str = dfs.create(new Path("/dir/file1"));
for (int i = 0; i < 1024 * 1024 * 4; i++) {
str.write(i);
}
DataNodeTestUtils.pauseIBR(cluster.getDataNodes().get(1));
DataNodeTestUtils.pauseIBR(cluster.getDataNodes().get(2));
LambdaTestUtils.intercept(IOException.class, "", () -> str.close());
}
}
}