HDFS-6955. DN should reserve disk space for a full block when creating tmp files (Contributed by Kanaka Kumar Avvaru)

This commit is contained in:
Vinayakumar B 2015-09-18 16:37:10 +05:30
parent a7201d635f
commit 92c1af1646
8 changed files with 198 additions and 56 deletions

View File

@ -117,7 +117,7 @@ class BlockReceiver implements Closeable {
/** the block to receive */
private final ExtendedBlock block;
/** the replica to write */
private final ReplicaInPipelineInterface replicaInfo;
private ReplicaInPipelineInterface replicaInfo;
/** pipeline stage */
private final BlockConstructionStage stage;
private final boolean isTransfer;
@ -259,6 +259,9 @@ class BlockReceiver implements Closeable {
} catch (ReplicaNotFoundException bne) {
throw bne;
} catch(IOException ioe) {
if (replicaInfo != null) {
replicaInfo.releaseAllBytesReserved();
}
IOUtils.closeStream(this);
cleanupBlock();

View File

@ -62,13 +62,13 @@ public interface FsVolumeSpi {
boolean isTransientStorage();
/**
* Reserve disk space for an RBW block so a writer does not run out of
* space before the block is full.
* Reserve disk space for a block (RBW or Re-replicating)
* so a writer does not run out of space before the block is full.
*/
void reserveSpaceForRbw(long bytesToReserve);
void reserveSpaceForReplica(long bytesToReserve);
/**
* Release disk space previously reserved for RBW block.
* Release disk space previously reserved for block opened for write.
*/
void releaseReservedSpace(long bytesToRelease);

View File

@ -1157,7 +1157,7 @@ private synchronized ReplicaBeingWritten append(String bpid,
// Replace finalized replica by a RBW replica in replicas map
volumeMap.add(bpid, newReplicaInfo);
v.reserveSpaceForRbw(estimateBlockLen - replicaInfo.getNumBytes());
v.reserveSpaceForReplica(estimateBlockLen - replicaInfo.getNumBytes());
return newReplicaInfo;
}
@ -1487,7 +1487,7 @@ public ReplicaHandler createTemporary(
}
ReplicaInPipeline newReplicaInfo =
new ReplicaInPipeline(b.getBlockId(), b.getGenerationStamp(), v,
f.getParentFile(), 0);
f.getParentFile(), b.getLocalBlock().getNumBytes());
volumeMap.add(b.getBlockPoolId(), newReplicaInfo);
return new ReplicaHandler(newReplicaInfo, ref);
} else {
@ -2555,14 +2555,15 @@ private static class VolumeInfo {
final long usedSpace; // size of space used by HDFS
final long freeSpace; // size of free space excluding reserved space
final long reservedSpace; // size of space reserved for non-HDFS
final long reservedSpaceForRBW; // size of space reserved RBW
final long reservedSpaceForReplicas; // size of space reserved RBW or
// re-replication
VolumeInfo(FsVolumeImpl v, long usedSpace, long freeSpace) {
this.directory = v.toString();
this.usedSpace = usedSpace;
this.freeSpace = freeSpace;
this.reservedSpace = v.getReserved();
this.reservedSpaceForRBW = v.getReservedForRbw();
this.reservedSpaceForReplicas = v.getReservedForReplicas();
}
}
@ -2596,7 +2597,7 @@ public Map<String, Object> getVolumeInfoMap() {
innerInfo.put("usedSpace", v.usedSpace);
innerInfo.put("freeSpace", v.freeSpace);
innerInfo.put("reservedSpace", v.reservedSpace);
innerInfo.put("reservedSpaceForRBW", v.reservedSpaceForRBW);
innerInfo.put("reservedSpaceForReplicas", v.reservedSpaceForReplicas);
info.put(v.directory, innerInfo);
}
return info;

View File

@ -22,8 +22,8 @@
import java.io.FileOutputStream;
import java.io.FilenameFilter;
import java.io.IOException;
import java.nio.channels.ClosedChannelException;
import java.io.OutputStreamWriter;
import java.nio.channels.ClosedChannelException;
import java.nio.file.Files;
import java.nio.file.Paths;
import java.nio.file.StandardCopyOption;
@ -40,9 +40,6 @@
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicLong;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.base.Joiner;
import com.google.common.base.Preconditions;
import org.apache.hadoop.classification.InterfaceAudience;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.DF;
@ -54,21 +51,24 @@
import org.apache.hadoop.hdfs.protocol.ExtendedBlock;
import org.apache.hadoop.hdfs.server.datanode.DataStorage;
import org.apache.hadoop.hdfs.server.datanode.DatanodeUtil;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsVolumeReference;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsDatasetSpi;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsVolumeReference;
import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsVolumeSpi;
import org.apache.hadoop.hdfs.server.protocol.DatanodeStorage;
import org.apache.hadoop.util.CloseableReferenceCount;
import org.apache.hadoop.io.IOUtils;
import org.apache.hadoop.util.CloseableReferenceCount;
import org.apache.hadoop.util.DiskChecker.DiskErrorException;
import com.google.common.util.concurrent.ThreadFactoryBuilder;
import org.apache.hadoop.util.Time;
import org.codehaus.jackson.annotate.JsonProperty;
import org.codehaus.jackson.map.ObjectMapper;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.base.Joiner;
import com.google.common.base.Preconditions;
import com.google.common.util.concurrent.ThreadFactoryBuilder;
/**
* The underlying volume used to store replica.
*
@ -90,8 +90,9 @@ public class FsVolumeImpl implements FsVolumeSpi {
private final long reserved;
private CloseableReferenceCount reference = new CloseableReferenceCount();
// Disk space reserved for open blocks.
private AtomicLong reservedForRbw;
// Disk space reserved for blocks (RBW or Re-replicating) open for write.
private AtomicLong reservedForReplicas;
private long recentReserved = 0;
// Capacity configured. This is useful when we want to
// limit the visible capacity for tests. If negative, then we just
@ -113,7 +114,7 @@ public class FsVolumeImpl implements FsVolumeSpi {
this.reserved = conf.getLong(
DFSConfigKeys.DFS_DATANODE_DU_RESERVED_KEY,
DFSConfigKeys.DFS_DATANODE_DU_RESERVED_DEFAULT);
this.reservedForRbw = new AtomicLong(0L);
this.reservedForReplicas = new AtomicLong(0L);
this.currentDir = currentDir;
File parent = currentDir.getParentFile();
this.usage = new DF(parent, conf);
@ -353,8 +354,9 @@ public void setCapacityForTesting(long capacity) {
*/
@Override
public long getAvailable() throws IOException {
long remaining = getCapacity() - getDfsUsed() - reservedForRbw.get();
long available = usage.getAvailable() - reserved - reservedForRbw.get();
long remaining = getCapacity() - getDfsUsed() - reservedForReplicas.get();
long available = usage.getAvailable() - reserved
- reservedForReplicas.get();
if (remaining > available) {
remaining = available;
}
@ -362,8 +364,13 @@ public long getAvailable() throws IOException {
}
@VisibleForTesting
public long getReservedForRbw() {
return reservedForRbw.get();
public long getReservedForReplicas() {
return reservedForReplicas.get();
}
@VisibleForTesting
long getRecentReserved() {
return recentReserved;
}
long getReserved(){
@ -412,13 +419,20 @@ public String[] getBlockPoolList() {
*/
File createTmpFile(String bpid, Block b) throws IOException {
checkReference();
reserveSpaceForReplica(b.getNumBytes());
try {
return getBlockPoolSlice(bpid).createTmpFile(b);
} catch (IOException exception) {
releaseReservedSpace(b.getNumBytes());
throw exception;
}
}
@Override
public void reserveSpaceForRbw(long bytesToReserve) {
public void reserveSpaceForReplica(long bytesToReserve) {
if (bytesToReserve != 0) {
reservedForRbw.addAndGet(bytesToReserve);
reservedForReplicas.addAndGet(bytesToReserve);
recentReserved = bytesToReserve;
}
}
@ -428,14 +442,15 @@ public void releaseReservedSpace(long bytesToRelease) {
long oldReservation, newReservation;
do {
oldReservation = reservedForRbw.get();
oldReservation = reservedForReplicas.get();
newReservation = oldReservation - bytesToRelease;
if (newReservation < 0) {
// Failsafe, this should never occur in practice, but if it does we don't
// want to start advertising more space than we have available.
// Failsafe, this should never occur in practice, but if it does we
// don't want to start advertising more space than we have available.
newReservation = 0;
}
} while (!reservedForRbw.compareAndSet(oldReservation, newReservation));
} while (!reservedForReplicas.compareAndSet(oldReservation,
newReservation));
}
}
@ -779,7 +794,7 @@ public FsDatasetSpi getDataset() {
*/
File createRbwFile(String bpid, Block b) throws IOException {
checkReference();
reserveSpaceForRbw(b.getNumBytes());
reserveSpaceForReplica(b.getNumBytes());
try {
return getBlockPoolSlice(bpid).createRbwFile(b);
} catch (IOException exception) {
@ -790,16 +805,15 @@ File createRbwFile(String bpid, Block b) throws IOException {
/**
*
* @param bytesReservedForRbw Space that was reserved during
* @param bytesReserved Space that was reserved during
* block creation. Now that the block is being finalized we
* can free up this space.
* @return
* @throws IOException
*/
File addFinalizedBlock(String bpid, Block b,
File f, long bytesReservedForRbw)
File addFinalizedBlock(String bpid, Block b, File f, long bytesReserved)
throws IOException {
releaseReservedSpace(bytesReservedForRbw);
releaseReservedSpace(bytesReserved);
return getBlockPoolSlice(bpid).addFinalizedBlock(b, f);
}

View File

@ -492,7 +492,7 @@ public boolean isTransientStorage() {
}
@Override
public void reserveSpaceForRbw(long bytesToReserve) {
public void reserveSpaceForReplica(long bytesToReserve) {
}
@Override

View File

@ -612,7 +612,7 @@ public boolean isTransientStorage() {
}
@Override
public void reserveSpaceForRbw(long bytesToReserve) {
public void reserveSpaceForReplica(long bytesToReserve) {
}
@Override

View File

@ -74,7 +74,7 @@ public boolean isTransientStorage() {
}
@Override
public void reserveSpaceForRbw(long bytesToReserve) {
public void reserveSpaceForReplica(long bytesToReserve) {
}
@Override

View File

@ -28,8 +28,10 @@
import static org.hamcrest.core.Is.is;
import static org.junit.Assert.assertThat;
import static org.junit.Assert.assertTrue;
import static org.junit.Assert.assertEquals;
import static org.junit.Assert.fail;
import org.apache.hadoop.fs.BlockLocation;
import org.apache.hadoop.fs.FSDataOutputStream;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hdfs.*;
@ -60,10 +62,10 @@
/**
* Ensure that the DN reserves disk space equivalent to a full block for
* replica being written (RBW).
* replica being written (RBW) & Replica being copied from another DN.
*/
public class TestRbwSpaceReservation {
static final Log LOG = LogFactory.getLog(TestRbwSpaceReservation.class);
public class TestSpaceReservation {
static final Log LOG = LogFactory.getLog(TestSpaceReservation.class);
private static final int DU_REFRESH_INTERVAL_MSEC = 500;
private static final int STORAGES_PER_DATANODE = 1;
@ -165,14 +167,14 @@ private void createFileAndTestSpaceReservation(
int bytesWritten = buffer.length;
// Check that space was reserved for a full block minus the bytesWritten.
assertThat(singletonVolume.getReservedForRbw(),
assertThat(singletonVolume.getReservedForReplicas(),
is((long) fileBlockSize - bytesWritten));
out.close();
out = null;
// Check that the reserved space has been released since we closed the
// file.
assertThat(singletonVolume.getReservedForRbw(), is(0L));
assertThat(singletonVolume.getReservedForReplicas(), is(0L));
// Reopen the file for appends and write 1 more byte.
out = fs.append(path);
@ -182,7 +184,7 @@ private void createFileAndTestSpaceReservation(
// Check that space was again reserved for a full block minus the
// bytesWritten so far.
assertThat(singletonVolume.getReservedForRbw(),
assertThat(singletonVolume.getReservedForReplicas(),
is((long) fileBlockSize - bytesWritten));
// Write once again and again verify the available space. This ensures
@ -191,7 +193,7 @@ private void createFileAndTestSpaceReservation(
out.write(buffer);
out.hsync();
bytesWritten += buffer.length;
assertThat(singletonVolume.getReservedForRbw(),
assertThat(singletonVolume.getReservedForReplicas(),
is((long) fileBlockSize - bytesWritten));
} finally {
if (out != null) {
@ -282,7 +284,7 @@ public void testSpaceReleasedOnUnexpectedEof()
GenericTestUtils.waitFor(new Supplier<Boolean>() {
@Override
public Boolean get() {
return (volume.getReservedForRbw() == 0);
return (volume.getReservedForReplicas() == 0);
}
}, 500, Integer.MAX_VALUE); // Wait until the test times out.
}
@ -324,12 +326,30 @@ public void testRBWFileCreationError() throws Exception {
}
// Ensure RBW space reserved is released
assertTrue("Expected ZERO but got " + fsVolumeImpl.getReservedForRbw(),
fsVolumeImpl.getReservedForRbw() == 0);
assertTrue(
"Expected ZERO but got " + fsVolumeImpl.getReservedForReplicas(),
fsVolumeImpl.getReservedForReplicas() == 0);
// Reserve some bytes to verify double clearing space should't happen
fsVolumeImpl.reserveSpaceForReplica(1000);
try {
// Write 1 byte to the file
FSDataOutputStream os = fs.create(new Path("/" + methodName + ".02.dat"),
replication);
os.write(new byte[1]);
os.hsync();
os.close();
fail("Expecting IOException file creation failure");
} catch (IOException e) {
// Exception can be ignored (expected)
}
// Ensure RBW space reserved is released only once
assertTrue(fsVolumeImpl.getReservedForReplicas() == 1000);
}
@Test(timeout = 30000)
public void testRBWInJMXBean() throws Exception {
public void testReservedSpaceInJMXBean() throws Exception {
final short replication = 1;
startCluster(BLOCK_SIZE, replication, -1);
@ -348,7 +368,111 @@ public void testRBWInJMXBean() throws Exception {
final String volumeInfo = (String) mbs.getAttribute(mxbeanName,
"VolumeInfo");
assertTrue(volumeInfo.contains("reservedSpaceForRBW"));
// verify reserved space for Replicas in JMX bean volume info
assertTrue(volumeInfo.contains("reservedSpaceForReplicas"));
}
}
@Test(timeout = 300000)
public void testTmpSpaceReserve() throws Exception {
final short replication = 2;
startCluster(BLOCK_SIZE, replication, -1);
final int byteCount1 = 100;
final int byteCount2 = 200;
final String methodName = GenericTestUtils.getMethodName();
// Test positive scenario
{
final Path file = new Path("/" + methodName + ".01.dat");
try (FSDataOutputStream os = fs.create(file, (short) 1)) {
// Write test data to the file
os.write(new byte[byteCount1]);
os.hsync();
}
BlockLocation[] blockLocations = fs.getFileBlockLocations(file, 0, 10);
String firstReplicaNode = blockLocations[0].getNames()[0];
int newReplicaDNIndex = 0;
if (firstReplicaNode.equals(cluster.getDataNodes().get(0)
.getDisplayName())) {
newReplicaDNIndex = 1;
}
FsVolumeImpl fsVolumeImpl = (FsVolumeImpl) cluster.getDataNodes()
.get(newReplicaDNIndex).getFSDataset().getFsVolumeReferences().get(0);
performReReplication(file, true);
assertEquals("Wrong reserve space for Tmp ", byteCount1,
fsVolumeImpl.getRecentReserved());
assertEquals("Reserved Tmp space is not released", 0,
fsVolumeImpl.getReservedForReplicas());
}
// Test when file creation fails
{
final Path file = new Path("/" + methodName + ".01.dat");
try (FSDataOutputStream os = fs.create(file, (short) 1)) {
// Write test data to the file
os.write(new byte[byteCount2]);
os.hsync();
}
BlockLocation[] blockLocations = fs.getFileBlockLocations(file, 0, 10);
String firstReplicaNode = blockLocations[0].getNames()[0];
int newReplicaDNIndex = 0;
if (firstReplicaNode.equals(cluster.getDataNodes().get(0)
.getDisplayName())) {
newReplicaDNIndex = 1;
}
BlockPoolSlice blockPoolSlice = Mockito.mock(BlockPoolSlice.class);
Mockito.when(blockPoolSlice.createTmpFile((Block) Mockito.any()))
.thenThrow(new IOException("Synthetic IO Exception Throgh MOCK"));
final FsVolumeImpl fsVolumeImpl = (FsVolumeImpl) cluster.getDataNodes()
.get(newReplicaDNIndex).getFSDataset().getFsVolumeReferences().get(0);
// Reserve some bytes to verify double clearing space should't happen
fsVolumeImpl.reserveSpaceForReplica(1000);
Field field = FsVolumeImpl.class.getDeclaredField("bpSlices");
field.setAccessible(true);
@SuppressWarnings("unchecked")
Map<String, BlockPoolSlice> bpSlices = (Map<String, BlockPoolSlice>) field
.get(fsVolumeImpl);
bpSlices.put(fsVolumeImpl.getBlockPoolList()[0], blockPoolSlice);
performReReplication(file, false);
assertEquals("Wrong reserve space for Tmp ", byteCount2,
fsVolumeImpl.getRecentReserved());
assertEquals("Tmp space is not released OR released twice", 1000,
fsVolumeImpl.getReservedForReplicas());
}
}
private void performReReplication(Path filePath, boolean waitForSuccess)
throws Exception {
fs.setReplication(filePath, (short) 2);
Thread.sleep(4000);
BlockLocation[] blockLocations = fs.getFileBlockLocations(filePath, 0, 10);
if (waitForSuccess) {
// Wait for the re replication
while (blockLocations[0].getNames().length < 2) {
Thread.sleep(2000);
blockLocations = fs.getFileBlockLocations(filePath, 0, 10);
}
}
}
@ -387,7 +511,7 @@ public void stressTest() throws IOException, InterruptedException {
" files and hit " + numFailures + " failures");
// Check no space was leaked.
assertThat(singletonVolume.getReservedForRbw(), is(0L));
assertThat(singletonVolume.getReservedForReplicas(), is(0L));
}
private static class Writer extends Daemon {