Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HDFS-6978. Directory scanner should correctly reconcile blocks on RAM disk. (Arpit Agarwal) 

Conflicts:
	hadoop-hdfs-project/hadoop-hdfs/CHANGES-HDFS-6581.txt
This commit is contained in:
arp 2014-09-12 22:13:18 -07:00 committed by Jitendra Pandey
parent 6906eccec4
commit 14edbc9419
7 changed files with 241 additions and 51 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/DirectoryScanner.java
View File

@ -83,6 +83,7 @@ public class DirectoryScanner implements Runnable {
long missingBlockFile = 0;
long missingMemoryBlocks = 0;
long mismatchBlocks = 0;
long duplicateBlocks = 0;
public Stats(String bpid) {
this.bpid = bpid;
@ -440,7 +441,7 @@ public class DirectoryScanner implements Runnable {
int d = 0; // index for blockpoolReport
int m = 0; // index for memReprot
while (m < memReport.length && d < blockpoolReport.length) {
Block memBlock = memReport[Math.min(m, memReport.length - 1)];
FinalizedReplica memBlock = memReport[Math.min(m, memReport.length - 1)];
ScanInfo info = blockpoolReport[Math.min(
d, blockpoolReport.length - 1)];
if (info.getBlockId() < memBlock.getBlockId()) {
@ -468,9 +469,23 @@ public class DirectoryScanner implements Runnable {
// or block file length is different than expected
statsRecord.mismatchBlocks++;
addDifference(diffRecord, statsRecord, info);
} else if (info.getBlockFile().compareTo(memBlock.getBlockFile()) != 0) {
// volumeMap record and on-disk files don't match.
statsRecord.duplicateBlocks++;
addDifference(diffRecord, statsRecord, info);
}
d++;
m++;
if (d < blockpoolReport.length) {
// There may be multiple on-disk records for the same block, don't increment
// the memory record pointer if so.
ScanInfo nextInfo = blockpoolReport[Math.min(d, blockpoolReport.length - 1)];
if (nextInfo.getBlockId() != info.blockId) {
++m;
}
} else {
++m;
}
}
while (m < memReport.length) {
FinalizedReplica current = memReport[m++];

9
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/BlockPoolSlice.java
View File

@ -269,15 +269,16 @@ class BlockPoolSlice {
/**
* Save the given replica to persistent storage.
*
* @param replicaInfo
* @return The saved meta and block files, in that order.
* @throws IOException
*/
File[] lazyPersistReplica(ReplicaInfo replicaInfo) throws IOException {
File[] lazyPersistReplica(long blockId, long genStamp,
File srcMeta, File srcFile) throws IOException {
if (!lazypersistDir.exists() && !lazypersistDir.mkdirs()) {
FsDatasetImpl.LOG.warn("Failed to create " + lazypersistDir);
}
File targetFiles[] = FsDatasetImpl.copyBlockFiles(replicaInfo, lazypersistDir);
File targetFiles[] = FsDatasetImpl.copyBlockFiles(
blockId, genStamp, srcMeta, srcFile, lazypersistDir);
dfsUsage.incDfsUsed(targetFiles[0].length() + targetFiles[1].length());
return targetFiles;
}
@ -510,7 +511,7 @@ class BlockPoolSlice {
* @return the replica that is retained.
* @throws IOException
*/
private ReplicaInfo resolveDuplicateReplicas(
ReplicaInfo resolveDuplicateReplicas(
final ReplicaInfo replica1, final ReplicaInfo replica2,
final ReplicaMap volumeMap) throws IOException {

57
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetImpl.java
View File

@ -703,17 +703,16 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
}
/**
* Copy the block and meta files for the given block from the given
* Copy the block and meta files for the given block to the given destination.
* @return the new meta and block files.
* @throws IOException
*/
static File[] copyBlockFiles(ReplicaInfo replicaInfo, File destRoot)
static File[] copyBlockFiles(long blockId, long genStamp,
File srcMeta, File srcFile, File destRoot)
throws IOException {
final File destDir = DatanodeUtil.idToBlockDir(destRoot, replicaInfo.getBlockId());
final File dstFile = new File(destDir, replicaInfo.getBlockName());
final File dstMeta = FsDatasetUtil.getMetaFile(dstFile, replicaInfo.getGenerationStamp());
final File srcMeta = replicaInfo.getMetaFile();
final File srcFile = replicaInfo.getBlockFile();
final File destDir = DatanodeUtil.idToBlockDir(destRoot, blockId);
final File dstFile = new File(destDir, srcFile.getName());
final File dstMeta = FsDatasetUtil.getMetaFile(dstFile, genStamp);
try {
FileUtils.copyFile(srcMeta, dstMeta);
} catch (IOException e) {
@ -1847,10 +1846,20 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
File memFile = memBlockInfo.getBlockFile();
if (memFile.exists()) {
if (memFile.compareTo(diskFile) != 0) {
LOG.warn("Block file " + memFile.getAbsolutePath()
+ " does not match file found by scan "
+ diskFile.getAbsolutePath());
// TODO: Should the diskFile be deleted?
if (diskMetaFile.exists()) {
if (memBlockInfo.getMetaFile().exists()) {
// We have two sets of block+meta files. Decide which one to
// keep.
ReplicaInfo diskBlockInfo = new FinalizedReplica(
blockId, diskFile.length(), diskGS, vol, diskFile.getParentFile());
((FsVolumeImpl) vol).getBlockPoolSlice(bpid).resolveDuplicateReplicas(
memBlockInfo, diskBlockInfo, volumeMap);
}
} else {
if (!diskFile.delete()) {
LOG.warn("Failed to delete " + diskFile + ". Will retry on next scan");
}
}
}
} else {
// Block refers to a block file that does not exist.
@ -2315,6 +2324,9 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
FsVolumeImpl targetVolume;
ReplicaInfo replicaInfo;
BlockPoolSlice bpSlice;
File srcFile, srcMeta;
long genStamp;
synchronized (FsDatasetImpl.this) {
replicaInfo = volumeMap.get(bpid, blockId);
@ -2336,10 +2348,18 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
}
lazyWriteReplicaTracker.recordStartLazyPersist(bpid, blockId, targetVolume);
File[] savedFiles = targetVolume.getBlockPoolSlice(bpid)
.lazyPersistReplica(replicaInfo);
lazyWriteReplicaTracker.recordEndLazyPersist(
bpid, blockId, savedFiles[0], savedFiles[1]);
bpSlice = targetVolume.getBlockPoolSlice(bpid);
srcMeta = replicaInfo.getMetaFile();
srcFile = replicaInfo.getBlockFile();
genStamp = replicaInfo.getGenerationStamp();
}
// Drop the FsDatasetImpl lock for the file copy.
File[] savedFiles =
bpSlice.lazyPersistReplica(blockId, genStamp, srcMeta, srcFile);
synchronized (FsDatasetImpl.this) {
lazyWriteReplicaTracker.recordEndLazyPersist(bpid, blockId, savedFiles);
if (LOG.isDebugEnabled()) {
LOG.debug("LazyWriter finished saving blockId=" + blockId + "; bpid=" + bpid +
@ -2360,7 +2380,6 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
try {
replicaState = lazyWriteReplicaTracker.dequeueNextReplicaToPersist();
if (replicaState != null) {
// Move the replica outside the lock.
moveReplicaToNewVolume(replicaState.bpid, replicaState.blockId);
}
succeeded = true;
@ -2459,9 +2478,9 @@ class FsDatasetImpl implements FsDatasetSpi<FsVolumeImpl> {
// Remove the old replicas from transient storage.
if (blockFile.delete() || !blockFile.exists()) {
((FsVolumeImpl) replicaInfo.getVolume()).decDfsUsed(replicaState.bpid, blockFileUsed);
}
if (metaFile.delete() || !metaFile.exists()) {
((FsVolumeImpl) replicaInfo.getVolume()).decDfsUsed(replicaState.bpid, metaFileUsed);
if (metaFile.delete() || !metaFile.exists()) {
((FsVolumeImpl) replicaInfo.getVolume()).decDfsUsed(replicaState.bpid, metaFileUsed);
}
}
// If deletion failed then the directory scanner will cleanup the blocks

12
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/LazyWriteReplicaTracker.java
View File

@ -161,9 +161,13 @@ class LazyWriteReplicaTracker {
replicaState.lazyPersistVolume = checkpointVolume;
}
/**
* @param bpid
* @param blockId
* @param savedFiles The saved meta and block files, in that order.
*/
synchronized void recordEndLazyPersist(
final String bpid, final long blockId,
final File savedMetaFile, final File savedBlockFile) {
final String bpid, final long blockId, final File[] savedFiles) {
Map<Long, ReplicaState> map = replicaMaps.get(bpid);
ReplicaState replicaState = map.get(blockId);
@ -172,8 +176,8 @@ class LazyWriteReplicaTracker {
bpid + "; blockId=" + blockId);
}
replicaState.state = State.LAZY_PERSIST_COMPLETE;
replicaState.savedMetaFile = savedMetaFile;
replicaState.savedBlockFile = savedBlockFile;
replicaState.savedMetaFile = savedFiles[0];
replicaState.savedBlockFile = savedFiles[1];
if (replicasNotPersisted.peek() == replicaState) {
// Common case.

171
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/datanode/TestDirectoryScanner.java
View File

@ -17,6 +17,8 @@
*/
package org.apache.hadoop.hdfs.server.datanode;
import static org.hamcrest.MatcherAssert.assertThat;
import static org.hamcrest.core.Is.is;
import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertFalse;
import static org.junit.Assert.assertNotNull;
@ -31,22 +33,21 @@ import java.util.LinkedList;
import java.util.List;
import java.util.Random;
import org.apache.commons.io.FileUtils;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hdfs.DFSConfigKeys;
import org.apache.hadoop.hdfs.DFSTestUtil;
import org.apache.hadoop.hdfs.HdfsConfiguration;
import org.apache.hadoop.hdfs.MiniDFSCluster;
import org.apache.hadoop.hdfs.StorageType;
import org.apache.hadoop.hdfs.*;
import org.apache.hadoop.hdfs.protocol.Block;
import org.apache.hadoop.hdfs.protocol.LocatedBlock;
import org.apache.hadoop.hdfs.server.common.GenerationStamp;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsDatasetSpi;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsVolumeSpi;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.FsDatasetTestUtil;
import org.apache.hadoop.io.IOUtils;
import org.apache.hadoop.test.GenericTestUtils;
import org.junit.Test;
/**
@ -60,22 +61,29 @@ public class TestDirectoryScanner {
private MiniDFSCluster cluster;
private String bpid;
private DFSClient client;
private FsDatasetSpi<? extends FsVolumeSpi> fds = null;
private DirectoryScanner scanner = null;
private final Random rand = new Random();
private final Random r = new Random();
private static final int BLOCK_LENGTH = 100;
static {
CONF.setLong(DFSConfigKeys.DFS_BLOCK_SIZE_KEY, 100);
CONF.setLong(DFSConfigKeys.DFS_BLOCK_SIZE_KEY, BLOCK_LENGTH);
CONF.setInt(DFSConfigKeys.DFS_BYTES_PER_CHECKSUM_KEY, 1);
CONF.setLong(DFSConfigKeys.DFS_HEARTBEAT_INTERVAL_KEY, 1L);
}
/** create a file with a length of <code>fileLen</code> */
private void createFile(String fileName, long fileLen) throws IOException {
private List<LocatedBlock> createFile(String fileNamePrefix,
long fileLen,
boolean isLazyPersist) throws IOException {
FileSystem fs = cluster.getFileSystem();
Path filePath = new Path(fileName);
DFSTestUtil.createFile(fs, filePath, fileLen, (short) 1, r.nextLong());
Path filePath = new Path("/" + fileNamePrefix + ".dat");
DFSTestUtil.createFile(
fs, filePath, isLazyPersist, 1024, fileLen,
BLOCK_LENGTH, (short) 1, r.nextLong(), false);
return client.getLocatedBlocks(filePath.toString(), 0, fileLen).getLocatedBlocks();
}
/** Truncate a block file */
@ -134,6 +142,43 @@ public class TestDirectoryScanner {
return 0;
}
/**
* Duplicate the given block on all volumes.
* @param blockId
* @throws IOException
*/
private void duplicateBlock(long blockId) throws IOException {
synchronized (fds) {
ReplicaInfo b = FsDatasetTestUtil.fetchReplicaInfo(fds, bpid, blockId);
for (FsVolumeSpi v : fds.getVolumes()) {
if (v.getStorageID().equals(b.getVolume().getStorageID())) {
continue;
}
// Volume without a copy of the block. Make a copy now.
File sourceBlock = b.getBlockFile();
File sourceMeta = b.getMetaFile();
String sourceRoot = b.getVolume().getBasePath();
String destRoot = v.getBasePath();
String relativeBlockPath = new File(sourceRoot).toURI().relativize(sourceBlock.toURI()).getPath();
String relativeMetaPath = new File(sourceRoot).toURI().relativize(sourceMeta.toURI()).getPath();
File destBlock = new File(destRoot, relativeBlockPath);
File destMeta = new File(destRoot, relativeMetaPath);
destBlock.getParentFile().mkdirs();
FileUtils.copyFile(sourceBlock, destBlock);
FileUtils.copyFile(sourceMeta, destMeta);
if (destBlock.exists() && destMeta.exists()) {
LOG.info("Copied " + sourceBlock + " ==> " + destBlock);
LOG.info("Copied " + sourceMeta + " ==> " + destMeta);
}
}
}
}
/** Get a random blockId that is not used already */
private long getFreeBlockId() {
long id = rand.nextLong();
@ -216,6 +261,12 @@ public class TestDirectoryScanner {
private void scan(long totalBlocks, int diffsize, long missingMetaFile, long missingBlockFile,
long missingMemoryBlocks, long mismatchBlocks) throws IOException {
scan(totalBlocks, diffsize, missingMetaFile, missingBlockFile,
missingMemoryBlocks, mismatchBlocks, 0);
}
private void scan(long totalBlocks, int diffsize, long missingMetaFile, long missingBlockFile,
long missingMemoryBlocks, long mismatchBlocks, long duplicateBlocks) throws IOException {
scanner.reconcile();
assertTrue(scanner.diffs.containsKey(bpid));
@ -229,9 +280,92 @@ public class TestDirectoryScanner {
assertEquals(missingBlockFile, stats.missingBlockFile);
assertEquals(missingMemoryBlocks, stats.missingMemoryBlocks);
assertEquals(mismatchBlocks, stats.mismatchBlocks);
assertEquals(duplicateBlocks, stats.duplicateBlocks);
}
@Test
@Test (timeout=300000)
public void testRetainBlockOnPersistentStorage() throws Exception {
cluster = new MiniDFSCluster
.Builder(CONF)
.storageTypes(new StorageType[] { StorageType.RAM_DISK, StorageType.DEFAULT })
.numDataNodes(1)
.build();
try {
cluster.waitActive();
bpid = cluster.getNamesystem().getBlockPoolId();
fds = DataNodeTestUtils.getFSDataset(cluster.getDataNodes().get(0));
client = cluster.getFileSystem().getClient();
scanner = new DirectoryScanner(fds, CONF);
scanner.setRetainDiffs(true);
FsDatasetTestUtil.stopLazyWriter(cluster.getDataNodes().get(0));
// Add a file with 1 block
List<LocatedBlock> blocks =
createFile(GenericTestUtils.getMethodName(), BLOCK_LENGTH, false);
// Ensure no difference between volumeMap and disk.
scan(1, 0, 0, 0, 0, 0);
// Make a copy of the block on RAM_DISK and ensure that it is
// picked up by the scanner.
duplicateBlock(blocks.get(0).getBlock().getBlockId());
scan(2, 1, 0, 0, 0, 0, 1);
verifyStorageType(blocks.get(0).getBlock().getBlockId(), false);
scan(1, 0, 0, 0, 0, 0);
} finally {
if (scanner != null) {
scanner.shutdown();
scanner = null;
}
cluster.shutdown();
cluster = null;
}
}
@Test (timeout=300000)
public void testDeleteBlockOnTransientStorage() throws Exception {
cluster = new MiniDFSCluster
.Builder(CONF)
.storageTypes(new StorageType[] { StorageType.RAM_DISK, StorageType.DEFAULT })
.numDataNodes(1)
.build();
try {
cluster.waitActive();
bpid = cluster.getNamesystem().getBlockPoolId();
fds = DataNodeTestUtils.getFSDataset(cluster.getDataNodes().get(0));
client = cluster.getFileSystem().getClient();
scanner = new DirectoryScanner(fds, CONF);
scanner.setRetainDiffs(true);
FsDatasetTestUtil.stopLazyWriter(cluster.getDataNodes().get(0));
// Create a file file on RAM_DISK
List<LocatedBlock> blocks =
createFile(GenericTestUtils.getMethodName(), BLOCK_LENGTH, true);
// Ensure no difference between volumeMap and disk.
scan(1, 0, 0, 0, 0, 0);
// Make a copy of the block on DEFAULT storage and ensure that it is
// picked up by the scanner.
duplicateBlock(blocks.get(0).getBlock().getBlockId());
scan(2, 1, 0, 0, 0, 0, 1);
// Ensure that the copy on RAM_DISK was deleted.
verifyStorageType(blocks.get(0).getBlock().getBlockId(), false);
scan(1, 0, 0, 0, 0, 0);
} finally {
if (scanner != null) {
scanner.shutdown();
scanner = null;
}
cluster.shutdown();
cluster = null;
}
}
@Test (timeout=600000)
public void testDirectoryScanner() throws Exception {
// Run the test with and without parallel scanning
for (int parallelism = 1; parallelism < 3; parallelism++) {
@ -245,16 +379,17 @@ public class TestDirectoryScanner {
cluster.waitActive();
bpid = cluster.getNamesystem().getBlockPoolId();
fds = DataNodeTestUtils.getFSDataset(cluster.getDataNodes().get(0));
client = cluster.getFileSystem().getClient();
CONF.setInt(DFSConfigKeys.DFS_DATANODE_DIRECTORYSCAN_THREADS_KEY,
parallelism);
scanner = new DirectoryScanner(fds, CONF);
scanner.setRetainDiffs(true);
// Add files with 100 blocks
createFile("/tmp/t1", 10000);
createFile(GenericTestUtils.getMethodName(), BLOCK_LENGTH * 100, false);
long totalBlocks = 100;
// Test1: No difference between in-memory and disk
// Test1: No difference between volumeMap and disk
scan(100, 0, 0, 0, 0, 0);
// Test2: block metafile is missing
@ -355,7 +490,10 @@ public class TestDirectoryScanner {
assertFalse(scanner.getRunStatus());
} finally {
scanner.shutdown();
if (scanner != null) {
scanner.shutdown();
scanner = null;
}
cluster.shutdown();
}
}
@ -389,6 +527,13 @@ public class TestDirectoryScanner {
assertEquals(genStamp, memBlock.getGenerationStamp());
}
private void verifyStorageType(long blockId, boolean expectTransient) {
final ReplicaInfo memBlock;
memBlock = FsDatasetTestUtil.fetchReplicaInfo(fds, bpid, blockId);
assertNotNull(memBlock);
assertThat(memBlock.getVolume().isTransientStorage(), is(expectTransient));
}
private static class TestFsVolumeSpi implements FsVolumeSpi {
@Override
public String[] getBlockPoolList() {

10
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetTestUtil.java
View File

@ -23,6 +23,7 @@ import java.util.Collection;
import org.apache.hadoop.hdfs.protocol.Block;
import org.apache.hadoop.hdfs.protocol.ExtendedBlock;
import org.apache.hadoop.hdfs.server.datanode.DataNode;
import org.apache.hadoop.hdfs.server.datanode.ReplicaInfo;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsDatasetSpi;
@ -62,4 +63,13 @@ public class FsDatasetTestUtil {
String bpid) {
return ((FsDatasetImpl)fsd).volumeMap.replicas(bpid);
}
/**
* Stop the lazy writer daemon that saves RAM disk files to persistent storage.
* @param dn
*/
public static void stopLazyWriter(DataNode dn) {
FsDatasetImpl fsDataset = ((FsDatasetImpl) dn.getFSDataset());
((FsDatasetImpl.LazyWriter) fsDataset.lazyWriter.getRunnable()).stop();
}
}

14
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/TestLazyPersistFiles.java
View File

@ -368,6 +368,8 @@ public class TestLazyPersistFiles {
// Found a persisted copy for this block!
boolean added = persistedBlockIds.add(lb.getBlock().getBlockId());
assertThat(added, is(true));
} else {
LOG.error(blockFile + " not found");
}
}
}
@ -423,7 +425,7 @@ public class TestLazyPersistFiles {
final int SEED = 0XFADED;
// Stop lazy writer to ensure block for path1 is not persisted to disk.
stopLazyWriter(cluster.getDataNodes().get(0));
FsDatasetTestUtil.stopLazyWriter(cluster.getDataNodes().get(0));
makeRandomTestFile(path1, BLOCK_SIZE, true, SEED);
ensureFileReplicasOnStorageType(path1, RAM_DISK);
@ -488,7 +490,7 @@ public class TestLazyPersistFiles {
throws IOException, InterruptedException {
startUpCluster(true, -1);
final String METHOD_NAME = GenericTestUtils.getMethodName();
stopLazyWriter(cluster.getDataNodes().get(0));
FsDatasetTestUtil.stopLazyWriter(cluster.getDataNodes().get(0));
Path path = new Path("/" + METHOD_NAME + ".dat");
makeTestFile(path, BLOCK_SIZE, true);
@ -682,7 +684,7 @@ public class TestLazyPersistFiles {
throws IOException, InterruptedException {
startUpCluster(true, 1);
stopLazyWriter(cluster.getDataNodes().get(0));
FsDatasetTestUtil.stopLazyWriter(cluster.getDataNodes().get(0));
final String METHOD_NAME = GenericTestUtils.getMethodName();
Path path1 = new Path("/" + METHOD_NAME + ".01.dat");
@ -794,12 +796,6 @@ public class TestLazyPersistFiles {
return locatedBlocks;
}
private void stopLazyWriter(DataNode dn) {
// Stop the lazyWriter daemon.
FsDatasetImpl fsDataset = ((FsDatasetImpl) dn.getFSDataset());
((FsDatasetImpl.LazyWriter) fsDataset.lazyWriter.getRunnable()).stop();
}
private void makeRandomTestFile(Path path, long length, final boolean isLazyPersist,
long seed) throws IOException {
DFSTestUtil.createFile(fs, path, isLazyPersist, BUFFER_LENGTH, length,