HDFS-9498. Move code that tracks blocks with future generation stamps to BlockManagerSafeMode. (Contributed by Mingliang Liu)
This commit is contained in:
parent
1cc001db4c
commit
2e6990cdf2
|
@ -1,4 +1,4 @@
|
|||
Hadoop HDFS Change Log
|
||||
Hadoop HDFS Change Log
|
||||
|
||||
Release 2.9.0 - UNRELEASED
|
||||
|
||||
|
@ -913,6 +913,9 @@ Release 2.8.0 - UNRELEASED
|
|||
HDFS-7779. Support changing ownership, group and replication in HDFS Web
|
||||
UI. (Ravi Prakash via wheat9)
|
||||
|
||||
HDFS-9498. Move code that tracks blocks with future generation stamps
|
||||
to BlockManagerSafeMode. (Mingliang Liu via Arpit Agarwal)
|
||||
|
||||
OPTIMIZATIONS
|
||||
|
||||
HDFS-8026. Trace FSOutputSummer#writeChecksumChunks rather than
|
||||
|
|
|
@ -70,7 +70,6 @@ import org.apache.hadoop.hdfs.security.token.block.ExportedBlockKeys;
|
|||
import org.apache.hadoop.hdfs.server.blockmanagement.CorruptReplicasMap.Reason;
|
||||
import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeStorageInfo.AddBlockResult;
|
||||
import org.apache.hadoop.hdfs.server.blockmanagement.PendingDataNodeMessages.ReportedBlockInfo;
|
||||
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants;
|
||||
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.BlockUCState;
|
||||
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.ReplicaState;
|
||||
import org.apache.hadoop.hdfs.server.namenode.CachedBlock;
|
||||
|
@ -293,12 +292,6 @@ public class BlockManager implements BlockStatsMXBean {
|
|||
/** Check whether name system is running before terminating */
|
||||
private boolean checkNSRunning = true;
|
||||
|
||||
/** Keeps track of how many bytes are in Future Generation blocks. */
|
||||
private AtomicLong numberOfBytesInFutureBlocks;
|
||||
|
||||
/** Reports if Name node was started with Rollback option. */
|
||||
private boolean inRollBack = false;
|
||||
|
||||
public BlockManager(final Namesystem namesystem, final Configuration conf)
|
||||
throws IOException {
|
||||
this.namesystem = namesystem;
|
||||
|
@ -377,8 +370,6 @@ public class BlockManager implements BlockStatsMXBean {
|
|||
DFSConfigKeys.DFS_BLOCK_MISREPLICATION_PROCESSING_LIMIT,
|
||||
DFSConfigKeys.DFS_BLOCK_MISREPLICATION_PROCESSING_LIMIT_DEFAULT);
|
||||
this.blockReportLeaseManager = new BlockReportLeaseManager(conf);
|
||||
this.numberOfBytesInFutureBlocks = new AtomicLong();
|
||||
this.inRollBack = isInRollBackMode(NameNode.getStartupOption(conf));
|
||||
|
||||
bmSafeMode = new BlockManagerSafeMode(this, namesystem, conf);
|
||||
|
||||
|
@ -1807,14 +1798,18 @@ public class BlockManager implements BlockStatsMXBean {
|
|||
return bmSafeMode.getSafeModeTip();
|
||||
}
|
||||
|
||||
public void leaveSafeMode(boolean force) {
|
||||
bmSafeMode.leaveSafeMode(force);
|
||||
public boolean leaveSafeMode(boolean force) {
|
||||
return bmSafeMode.leaveSafeMode(force);
|
||||
}
|
||||
|
||||
void checkSafeMode() {
|
||||
bmSafeMode.checkSafeMode();
|
||||
}
|
||||
|
||||
public long getBytesInFuture() {
|
||||
return bmSafeMode.getBytesInFuture();
|
||||
}
|
||||
|
||||
/**
|
||||
* Removes the blocks from blocksmap and updates the safemode blocks total
|
||||
*
|
||||
|
@ -2200,12 +2195,7 @@ public class BlockManager implements BlockStatsMXBean {
|
|||
// If block does not belong to any file, we check if it violates
|
||||
// an integrity assumption of Name node
|
||||
if (storedBlock == null) {
|
||||
if (namesystem.isInStartupSafeMode()
|
||||
&& !shouldPostponeBlocksFromFuture
|
||||
&& !inRollBack
|
||||
&& namesystem.isGenStampInFuture(iblk)) {
|
||||
numberOfBytesInFutureBlocks.addAndGet(iblk.getBytesOnDisk());
|
||||
}
|
||||
bmSafeMode.checkBlocksWithFutureGS(iblk);
|
||||
continue;
|
||||
}
|
||||
|
||||
|
@ -3867,39 +3857,8 @@ public class BlockManager implements BlockStatsMXBean {
|
|||
return haContext.getState().shouldPopulateReplQueues();
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns the number of bytes that reside in blocks with Generation Stamps
|
||||
* greater than generation stamp known to Namenode.
|
||||
*
|
||||
* @return Bytes in future
|
||||
*/
|
||||
public long getBytesInFuture() {
|
||||
return numberOfBytesInFutureBlocks.get();
|
||||
}
|
||||
|
||||
/**
|
||||
* Clears the bytes in future counter.
|
||||
*/
|
||||
public void clearBytesInFuture() {
|
||||
numberOfBytesInFutureBlocks.set(0);
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns true if Namenode was started with a RollBack option.
|
||||
*
|
||||
* @param option - StartupOption
|
||||
* @return boolean
|
||||
*/
|
||||
private boolean isInRollBackMode(HdfsServerConstants.StartupOption option) {
|
||||
if (option == HdfsServerConstants.StartupOption.ROLLBACK) {
|
||||
return true;
|
||||
}
|
||||
if ((option == HdfsServerConstants.StartupOption.ROLLINGUPGRADE) &&
|
||||
(option.getRollingUpgradeStartupOption() ==
|
||||
HdfsServerConstants.RollingUpgradeStartupOption.ROLLBACK)) {
|
||||
return true;
|
||||
}
|
||||
return false;
|
||||
boolean getShouldPostponeBlocksFromFuture() {
|
||||
return shouldPostponeBlocksFromFuture;
|
||||
}
|
||||
|
||||
// async processing of an action, used for IBRs.
|
||||
|
|
|
@ -21,6 +21,9 @@ import org.apache.hadoop.classification.InterfaceAudience;
|
|||
import org.apache.hadoop.classification.InterfaceStability;
|
||||
import org.apache.hadoop.conf.Configuration;
|
||||
import org.apache.hadoop.hdfs.DFSConfigKeys;
|
||||
import org.apache.hadoop.hdfs.protocol.BlockListAsLongs.BlockReportReplica;
|
||||
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.RollingUpgradeStartupOption;
|
||||
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.StartupOption;
|
||||
import org.apache.hadoop.hdfs.server.namenode.NameNode;
|
||||
import org.apache.hadoop.hdfs.server.namenode.Namesystem;
|
||||
import org.apache.hadoop.hdfs.server.namenode.startupprogress.Phase;
|
||||
|
@ -106,6 +109,11 @@ class BlockManagerSafeMode {
|
|||
/** Counter for tracking startup progress of reported blocks. */
|
||||
private Counter awaitingReportedBlocksCounter;
|
||||
|
||||
/** Keeps track of how many bytes are in Future Generation blocks. */
|
||||
private final AtomicLong numberOfBytesInFutureBlocks = new AtomicLong();
|
||||
/** Reports if Name node was started with Rollback option. */
|
||||
private final boolean inRollBack;
|
||||
|
||||
BlockManagerSafeMode(BlockManager blockManager, Namesystem namesystem,
|
||||
Configuration conf) {
|
||||
this.blockManager = blockManager;
|
||||
|
@ -135,9 +143,10 @@ class BlockManagerSafeMode {
|
|||
this.replQueueThreshold =
|
||||
conf.getFloat(DFS_NAMENODE_REPL_QUEUE_THRESHOLD_PCT_KEY,
|
||||
(float) threshold);
|
||||
|
||||
this.extension = conf.getInt(DFS_NAMENODE_SAFEMODE_EXTENSION_KEY, 0);
|
||||
|
||||
this.inRollBack = isInRollBackMode(NameNode.getStartupOption(conf));
|
||||
|
||||
LOG.info("{} = {}", DFS_NAMENODE_SAFEMODE_THRESHOLD_PCT_KEY, threshold);
|
||||
LOG.info("{} = {}", DFS_NAMENODE_SAFEMODE_MIN_DATANODES_KEY,
|
||||
datanodeThreshold);
|
||||
|
@ -300,15 +309,15 @@ class BlockManagerSafeMode {
|
|||
numLive, datanodeThreshold);
|
||||
}
|
||||
|
||||
if (blockManager.getBytesInFuture() > 0) {
|
||||
if (getBytesInFuture() > 0) {
|
||||
msg += "Name node detected blocks with generation stamps " +
|
||||
"in future. This means that Name node metadata is inconsistent." +
|
||||
"in future. This means that Name node metadata is inconsistent. " +
|
||||
"This can happen if Name node metadata files have been manually " +
|
||||
"replaced. Exiting safe mode will cause loss of " + blockManager
|
||||
.getBytesInFuture() + " byte(s). Please restart name node with " +
|
||||
"right metadata or use \"hdfs dfsadmin -safemode forceExit" +
|
||||
"if you are certain that the NameNode was started with the" +
|
||||
"correct FsImage and edit logs. If you encountered this during" +
|
||||
"replaced. Exiting safe mode will cause loss of " +
|
||||
getBytesInFuture() + " byte(s). Please restart name node with " +
|
||||
"right metadata or use \"hdfs dfsadmin -safemode forceExit\" " +
|
||||
"if you are certain that the NameNode was started with the " +
|
||||
"correct FsImage and edit logs. If you encountered this during " +
|
||||
"a rollback, it is safe to exit with -safemode forceExit.";
|
||||
return msg;
|
||||
}
|
||||
|
@ -333,11 +342,31 @@ class BlockManagerSafeMode {
|
|||
|
||||
/**
|
||||
* Leave start up safe mode.
|
||||
*
|
||||
* @param force - true to force exit
|
||||
* @return true if it leaves safe mode successfully else false
|
||||
*/
|
||||
void leaveSafeMode(boolean force) {
|
||||
boolean leaveSafeMode(boolean force) {
|
||||
assert namesystem.hasWriteLock() : "Leaving safe mode needs write lock!";
|
||||
|
||||
final long bytesInFuture = numberOfBytesInFutureBlocks.get();
|
||||
if (bytesInFuture > 0) {
|
||||
if (force) {
|
||||
LOG.warn("Leaving safe mode due to forceExit. This will cause a data "
|
||||
+ "loss of {} byte(s).", bytesInFuture);
|
||||
numberOfBytesInFutureBlocks.set(0);
|
||||
} else {
|
||||
LOG.error("Refusing to leave safe mode without a force flag. " +
|
||||
"Exiting safe mode will cause a deletion of {} byte(s). Please " +
|
||||
"use -forceExit flag to exit safe mode forcefully if data loss is" +
|
||||
" acceptable.", bytesInFuture);
|
||||
return false;
|
||||
}
|
||||
} else if (force) {
|
||||
LOG.warn("forceExit used when normal exist would suffice. Treating " +
|
||||
"force exit as normal safe mode exit.");
|
||||
}
|
||||
|
||||
// if not done yet, initialize replication queues.
|
||||
// In the standby, do not populate repl queues
|
||||
if (!blockManager.isPopulatingReplQueues() &&
|
||||
|
@ -345,14 +374,6 @@ class BlockManagerSafeMode {
|
|||
blockManager.initializeReplQueues();
|
||||
}
|
||||
|
||||
if (!force && blockManager.getBytesInFuture() > 0) {
|
||||
LOG.error("Refusing to leave safe mode without a force flag. " +
|
||||
"Exiting safe mode will cause a deletion of {} byte(s). Please use " +
|
||||
"-forceExit flag to exit safe mode forcefully if data loss is " +
|
||||
"acceptable.", blockManager.getBytesInFuture());
|
||||
return;
|
||||
}
|
||||
|
||||
if (status != BMSafeModeStatus.OFF) {
|
||||
NameNode.stateChangeLog.info("STATE* Safe mode is OFF");
|
||||
}
|
||||
|
@ -379,6 +400,8 @@ class BlockManagerSafeMode {
|
|||
BlockManagerSafeMode.STEP_AWAITING_REPORTED_BLOCKS);
|
||||
prog.endPhase(Phase.SAFEMODE);
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
/**
|
||||
|
@ -431,6 +454,35 @@ class BlockManagerSafeMode {
|
|||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Check if the block report replica has a generation stamp (GS) in future.
|
||||
* If safe mode is not currently on, this is a no-op.
|
||||
*
|
||||
* @param brr block report replica which belongs to no file in BlockManager
|
||||
*/
|
||||
void checkBlocksWithFutureGS(BlockReportReplica brr) {
|
||||
assert namesystem.hasWriteLock();
|
||||
if (status == BMSafeModeStatus.OFF) {
|
||||
return;
|
||||
}
|
||||
|
||||
if (!blockManager.getShouldPostponeBlocksFromFuture() &&
|
||||
!inRollBack &&
|
||||
namesystem.isGenStampInFuture(brr)) {
|
||||
numberOfBytesInFutureBlocks.addAndGet(brr.getBytesOnDisk());
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns the number of bytes that reside in blocks with Generation Stamps
|
||||
* greater than generation stamp known to Namenode.
|
||||
*
|
||||
* @return Bytes in future
|
||||
*/
|
||||
long getBytesInFuture() {
|
||||
return numberOfBytesInFutureBlocks.get();
|
||||
}
|
||||
|
||||
void close() {
|
||||
assert namesystem.hasWriteLock() : "Closing bmSafeMode needs write lock!";
|
||||
try {
|
||||
|
@ -449,6 +501,19 @@ class BlockManagerSafeMode {
|
|||
return reachedTime.get() + extension - monotonicNow();
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns true if Namenode was started with a RollBack option.
|
||||
*
|
||||
* @param option - StartupOption
|
||||
* @return boolean
|
||||
*/
|
||||
private static boolean isInRollBackMode(StartupOption option) {
|
||||
return (option == StartupOption.ROLLBACK) ||
|
||||
(option == StartupOption.ROLLINGUPGRADE &&
|
||||
option.getRollingUpgradeStartupOption() ==
|
||||
RollingUpgradeStartupOption.ROLLBACK);
|
||||
}
|
||||
|
||||
/** Check if we are ready to initialize replication queues. */
|
||||
private void initializeReplQueuesIfNecessary() {
|
||||
assert namesystem.hasWriteLock();
|
||||
|
|
|
@ -973,6 +973,7 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
|
|||
|
||||
@Override
|
||||
public void startSecretManagerIfNecessary() {
|
||||
assert hasWriteLock() : "Starting secret manager needs write lock";
|
||||
boolean shouldRun = shouldUseDelegationTokens() &&
|
||||
!isInSafeMode() && getEditLog().isOpenForWrite();
|
||||
boolean running = dtSecretManager.isRunning();
|
||||
|
@ -3967,29 +3968,13 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
|
|||
checkSuperuserPrivilege();
|
||||
switch(action) {
|
||||
case SAFEMODE_LEAVE: // leave safe mode
|
||||
if (blockManager.getBytesInFuture() > 0) {
|
||||
LOG.error("Refusing to leave safe mode without a force flag. " +
|
||||
"Exiting safe mode will cause a deletion of " + blockManager
|
||||
.getBytesInFuture() + " byte(s). Please use " +
|
||||
"-forceExit flag to exit safe mode forcefully and data loss is " +
|
||||
"acceptable.");
|
||||
} else {
|
||||
leaveSafeMode();
|
||||
}
|
||||
leaveSafeMode(false);
|
||||
break;
|
||||
case SAFEMODE_ENTER: // enter safe mode
|
||||
enterSafeMode(false);
|
||||
break;
|
||||
case SAFEMODE_FORCE_EXIT:
|
||||
if (blockManager.getBytesInFuture() > 0) {
|
||||
LOG.warn("Leaving safe mode due to forceExit. This will cause a data "
|
||||
+ "loss of " + blockManager.getBytesInFuture() + " byte(s).");
|
||||
blockManager.clearBytesInFuture();
|
||||
} else {
|
||||
LOG.warn("forceExit used when normal exist would suffice. Treating " +
|
||||
"force exit as normal safe mode exit.");
|
||||
}
|
||||
leaveSafeMode();
|
||||
leaveSafeMode(true);
|
||||
break;
|
||||
default:
|
||||
LOG.error("Unexpected safe mode action");
|
||||
|
@ -4086,16 +4071,19 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
|
|||
|
||||
/**
|
||||
* Leave safe mode.
|
||||
* @param force true if to leave safe mode forcefully with -forceExit option
|
||||
*/
|
||||
void leaveSafeMode() {
|
||||
void leaveSafeMode(boolean force) {
|
||||
writeLock();
|
||||
try {
|
||||
if (!isInSafeMode()) {
|
||||
NameNode.stateChangeLog.info("STATE* Safe mode is already OFF");
|
||||
return;
|
||||
}
|
||||
setManualAndResourceLowSafeMode(false, false);
|
||||
blockManager.leaveSafeMode(true);
|
||||
if (blockManager.leaveSafeMode(force)) {
|
||||
setManualAndResourceLowSafeMode(false, false);
|
||||
startSecretManagerIfNecessary();
|
||||
}
|
||||
} finally {
|
||||
writeUnlock();
|
||||
}
|
||||
|
|
|
@ -22,6 +22,7 @@ import org.apache.hadoop.conf.Configuration;
|
|||
import org.apache.hadoop.hdfs.DFSConfigKeys;
|
||||
import org.apache.hadoop.hdfs.HdfsConfiguration;
|
||||
import org.apache.hadoop.hdfs.protocol.Block;
|
||||
import org.apache.hadoop.hdfs.protocol.BlockListAsLongs.BlockReportReplica;
|
||||
import org.apache.hadoop.hdfs.server.blockmanagement.BlockManagerSafeMode.BMSafeModeStatus;
|
||||
import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.NamenodeRole;
|
||||
import org.apache.hadoop.hdfs.server.namenode.FSNamesystem;
|
||||
|
@ -31,7 +32,6 @@ import org.apache.hadoop.test.GenericTestUtils;
|
|||
import org.junit.Before;
|
||||
import org.junit.Test;
|
||||
|
||||
import org.mockito.Mockito;
|
||||
import org.mockito.internal.util.reflection.Whitebox;
|
||||
|
||||
import java.io.IOException;
|
||||
|
@ -42,7 +42,6 @@ import static org.junit.Assert.assertEquals;
|
|||
import static org.junit.Assert.assertFalse;
|
||||
import static org.junit.Assert.assertTrue;
|
||||
|
||||
import static org.junit.Assert.fail;
|
||||
import static org.mockito.Matchers.any;
|
||||
import static org.mockito.Mockito.doReturn;
|
||||
import static org.mockito.Mockito.mock;
|
||||
|
@ -91,9 +90,10 @@ public class TestBlockManagerSafeMode {
|
|||
DATANODE_NUM);
|
||||
|
||||
FSNamesystem fsn = mock(FSNamesystem.class);
|
||||
Mockito.doReturn(true).when(fsn).hasWriteLock();
|
||||
Mockito.doReturn(true).when(fsn).hasReadLock();
|
||||
Mockito.doReturn(true).when(fsn).isRunning();
|
||||
doReturn(true).when(fsn).hasWriteLock();
|
||||
doReturn(true).when(fsn).hasReadLock();
|
||||
doReturn(true).when(fsn).isRunning();
|
||||
doReturn(true).when(fsn).isGenStampInFuture(any(Block.class));
|
||||
NameNode.initMetrics(conf, NamenodeRole.NAMENODE);
|
||||
|
||||
bm = spy(new BlockManager(fsn, conf));
|
||||
|
@ -109,7 +109,7 @@ public class TestBlockManagerSafeMode {
|
|||
* Test set block total.
|
||||
*
|
||||
* The block total is set which will call checkSafeMode for the first time
|
||||
* and bmSafeMode transfers from INITIALIZED to PENDING_THRESHOLD status
|
||||
* and bmSafeMode transfers from OFF to PENDING_THRESHOLD status
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testInitialize() {
|
||||
|
@ -169,7 +169,7 @@ public class TestBlockManagerSafeMode {
|
|||
*
|
||||
* Once the block threshold is reached, the block manger leaves safe mode and
|
||||
* increment will be a no-op.
|
||||
* The safe mode status lifecycle: INITIALIZED -> PENDING_THRESHOLD -> OFF
|
||||
* The safe mode status lifecycle: OFF -> PENDING_THRESHOLD -> OFF
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testIncrementSafeBlockCount() {
|
||||
|
@ -195,7 +195,7 @@ public class TestBlockManagerSafeMode {
|
|||
*
|
||||
* Once the block threshold is reached, the block manger leaves safe mode and
|
||||
* increment will be a no-op.
|
||||
* The safe mode status lifecycle: INITIALIZED -> PENDING_THRESHOLD -> EXTENSION-> OFF
|
||||
* The safe mode status lifecycle: OFF -> PENDING_THRESHOLD -> EXTENSION-> OFF
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testIncrementSafeBlockCountWithExtension() throws Exception {
|
||||
|
@ -215,7 +215,7 @@ public class TestBlockManagerSafeMode {
|
|||
* Test that the block safe decreases the block safe.
|
||||
*
|
||||
* The block manager stays in safe mode.
|
||||
* The safe mode status lifecycle: INITIALIZED -> PENDING_THRESHOLD
|
||||
* The safe mode status lifecycle: OFF -> PENDING_THRESHOLD
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testDecrementSafeBlockCount() {
|
||||
|
@ -237,7 +237,7 @@ public class TestBlockManagerSafeMode {
|
|||
* Test when the block safe increment and decrement interleave.
|
||||
*
|
||||
* Both the increment and decrement will be a no-op if the safe mode is OFF.
|
||||
* The safe mode status lifecycle: INITIALIZED -> PENDING_THRESHOLD -> OFF
|
||||
* The safe mode status lifecycle: OFF -> PENDING_THRESHOLD -> OFF
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testIncrementAndDecrementSafeBlockCount() {
|
||||
|
@ -303,24 +303,31 @@ public class TestBlockManagerSafeMode {
|
|||
}
|
||||
|
||||
/**
|
||||
* Test block manager won't leave safe mode if there are orphan blocks.
|
||||
* Test block manager won't leave safe mode if there are blocks with
|
||||
* generation stamp (GS) in future.
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testStayInSafeModeWhenBytesInFuture() throws Exception {
|
||||
bmSafeMode.activate(BLOCK_TOTAL);
|
||||
|
||||
when(bm.getBytesInFuture()).thenReturn(1L);
|
||||
// Inject blocks with future GS
|
||||
injectBlocksWithFugureGS(100L);
|
||||
assertEquals(100L, bmSafeMode.getBytesInFuture());
|
||||
|
||||
// safe blocks are enough
|
||||
setBlockSafe(BLOCK_THRESHOLD);
|
||||
|
||||
// PENDING_THRESHOLD -> EXTENSION
|
||||
bmSafeMode.checkSafeMode();
|
||||
try {
|
||||
waitForExtensionPeriod();
|
||||
fail("Safe mode should not leave extension period with orphan blocks!");
|
||||
} catch (TimeoutException e) {
|
||||
assertEquals(BMSafeModeStatus.EXTENSION, getSafeModeStatus());
|
||||
}
|
||||
|
||||
assertFalse("Shouldn't leave safe mode in case of blocks with future GS! ",
|
||||
bmSafeMode.leaveSafeMode(false));
|
||||
assertTrue("Leaving safe mode forcefully should succeed regardless of " +
|
||||
"blocks with future GS.", bmSafeMode.leaveSafeMode(true));
|
||||
assertEquals("Number of blocks with future GS should have been cleared " +
|
||||
"after leaving safe mode", 0L, bmSafeMode.getBytesInFuture());
|
||||
assertTrue("Leaving safe mode should succeed after blocks with future GS " +
|
||||
"are cleared.", bmSafeMode.leaveSafeMode(false));
|
||||
}
|
||||
|
||||
/**
|
||||
|
@ -347,7 +354,7 @@ public class TestBlockManagerSafeMode {
|
|||
tip = bmSafeMode.getSafeModeTip();
|
||||
assertTrue(tip.contains(
|
||||
String.format("The reported blocks %d has reached the threshold"
|
||||
+ " %.4f of total blocks %d. ",
|
||||
+ " %.4f of total blocks %d. ",
|
||||
getblockSafe(), THRESHOLD, BLOCK_TOTAL)));
|
||||
assertTrue(tip.contains(
|
||||
String.format("The number of live datanodes %d has reached the " +
|
||||
|
@ -357,7 +364,6 @@ public class TestBlockManagerSafeMode {
|
|||
|
||||
waitForExtensionPeriod();
|
||||
tip = bmSafeMode.getSafeModeTip();
|
||||
System.out.println(tip);
|
||||
assertTrue(tip.contains(
|
||||
String.format("The reported blocks %d has reached the threshold"
|
||||
+ " %.4f of total blocks %d. ",
|
||||
|
@ -369,7 +375,55 @@ public class TestBlockManagerSafeMode {
|
|||
}
|
||||
|
||||
/**
|
||||
* Mock block manager internal state for decrement safe block
|
||||
* Test get safe mode tip in case of blocks with future GS.
|
||||
*/
|
||||
@Test(timeout = 30000)
|
||||
public void testGetSafeModeTipForBlocksWithFutureGS() throws Exception {
|
||||
bmSafeMode.activate(BLOCK_TOTAL);
|
||||
|
||||
injectBlocksWithFugureGS(40L);
|
||||
String tip = bmSafeMode.getSafeModeTip();
|
||||
assertTrue(tip.contains(
|
||||
String.format(
|
||||
"The reported blocks %d needs additional %d blocks to reach the " +
|
||||
"threshold %.4f of total blocks %d.%n",
|
||||
0, BLOCK_THRESHOLD, THRESHOLD, BLOCK_TOTAL)));
|
||||
assertTrue(tip.contains(
|
||||
"Name node detected blocks with generation stamps " +
|
||||
"in future. This means that Name node metadata is inconsistent. " +
|
||||
"This can happen if Name node metadata files have been manually " +
|
||||
"replaced. Exiting safe mode will cause loss of " +
|
||||
40 + " byte(s). Please restart name node with " +
|
||||
"right metadata or use \"hdfs dfsadmin -safemode forceExit\" " +
|
||||
"if you are certain that the NameNode was started with the " +
|
||||
"correct FsImage and edit logs. If you encountered this during " +
|
||||
"a rollback, it is safe to exit with -safemode forceExit."
|
||||
));
|
||||
assertFalse(tip.contains("Safe mode will be turned off"));
|
||||
|
||||
// blocks with future GS were already injected before.
|
||||
setBlockSafe(BLOCK_THRESHOLD);
|
||||
tip = bmSafeMode.getSafeModeTip();
|
||||
assertTrue(tip.contains(
|
||||
String.format("The reported blocks %d has reached the threshold"
|
||||
+ " %.4f of total blocks %d. ",
|
||||
getblockSafe(), THRESHOLD, BLOCK_TOTAL)));
|
||||
assertTrue(tip.contains(
|
||||
"Name node detected blocks with generation stamps " +
|
||||
"in future. This means that Name node metadata is inconsistent. " +
|
||||
"This can happen if Name node metadata files have been manually " +
|
||||
"replaced. Exiting safe mode will cause loss of " +
|
||||
40 + " byte(s). Please restart name node with " +
|
||||
"right metadata or use \"hdfs dfsadmin -safemode forceExit\" " +
|
||||
"if you are certain that the NameNode was started with the " +
|
||||
"correct FsImage and edit logs. If you encountered this during " +
|
||||
"a rollback, it is safe to exit with -safemode forceExit."
|
||||
));
|
||||
assertFalse(tip.contains("Safe mode will be turned off"));
|
||||
}
|
||||
|
||||
/**
|
||||
* Mock block manager internal state for decrement safe block.
|
||||
*/
|
||||
private void mockBlockManagerForBlockSafeDecrement() {
|
||||
BlockInfo storedBlock = mock(BlockInfo.class);
|
||||
|
@ -396,6 +450,12 @@ public class TestBlockManagerSafeMode {
|
|||
}, EXTENSION / 10, EXTENSION * 2);
|
||||
}
|
||||
|
||||
private void injectBlocksWithFugureGS(long numBytesInFuture) {
|
||||
BlockReportReplica brr = mock(BlockReportReplica.class);
|
||||
when(brr.getBytesOnDisk()).thenReturn(numBytesInFuture);
|
||||
bmSafeMode.checkBlocksWithFutureGS(brr);
|
||||
}
|
||||
|
||||
private void setSafeModeStatus(BMSafeModeStatus status) {
|
||||
Whitebox.setInternalState(bmSafeMode, "status", status);
|
||||
}
|
||||
|
|
|
@ -91,7 +91,7 @@ public class NameNodeAdapter {
|
|||
}
|
||||
|
||||
public static void leaveSafeMode(NameNode namenode) {
|
||||
namenode.getNamesystem().leaveSafeMode();
|
||||
namenode.getNamesystem().leaveSafeMode(false);
|
||||
}
|
||||
|
||||
public static void abortEditLogs(NameNode nn) {
|
||||
|
|
|
@ -1610,7 +1610,7 @@ public class TestCheckpoint {
|
|||
FSNamesystem fsns = cluster.getNamesystem();
|
||||
fsns.enterSafeMode(false);
|
||||
fsns.saveNamespace();
|
||||
fsns.leaveSafeMode();
|
||||
fsns.leaveSafeMode(false);
|
||||
|
||||
secondary = startSecondaryNameNode(conf);
|
||||
|
||||
|
|
|
@ -306,7 +306,7 @@ public class TestEditLogRace {
|
|||
assertEquals(fsimage.getStorage().getMostRecentCheckpointTxId(),
|
||||
editLog.getLastWrittenTxId() - 1);
|
||||
|
||||
namesystem.leaveSafeMode();
|
||||
namesystem.leaveSafeMode(false);
|
||||
LOG.info("Save " + i + ": complete");
|
||||
}
|
||||
} finally {
|
||||
|
|
|
@ -105,7 +105,7 @@ public class TestFSNamesystem {
|
|||
Mockito.when(fsImage.getEditLog()).thenReturn(fsEditLog);
|
||||
FSNamesystem fsn = new FSNamesystem(conf, fsImage);
|
||||
|
||||
fsn.leaveSafeMode();
|
||||
fsn.leaveSafeMode(false);
|
||||
assertTrue("After leaving safemode FSNamesystem.isInStartupSafeMode still "
|
||||
+ "returned true", !fsn.isInStartupSafeMode());
|
||||
assertTrue("After leaving safemode FSNamesystem.isInSafeMode still returned"
|
||||
|
@ -145,7 +145,7 @@ public class TestFSNamesystem {
|
|||
assertTrue("FSNamesystem didn't enter safemode", fsn.isInSafeMode());
|
||||
assertTrue("Replication queues were being populated during very first "
|
||||
+ "safemode", !bm.isPopulatingReplQueues());
|
||||
fsn.leaveSafeMode();
|
||||
fsn.leaveSafeMode(false);
|
||||
assertTrue("FSNamesystem didn't leave safemode", !fsn.isInSafeMode());
|
||||
assertTrue("Replication queues weren't being populated even after leaving "
|
||||
+ "safemode", bm.isPopulatingReplQueues());
|
||||
|
|
|
@ -489,7 +489,7 @@ public class TestINodeFile {
|
|||
// Apply editlogs to fsimage, ensure inodeUnderConstruction is handled
|
||||
fsn.enterSafeMode(false);
|
||||
fsn.saveNamespace();
|
||||
fsn.leaveSafeMode();
|
||||
fsn.leaveSafeMode(false);
|
||||
|
||||
outStream.close();
|
||||
|
||||
|
|
|
@ -23,26 +23,26 @@ 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.MiniDFSCluster.DataNodeProperties;
|
||||
import org.apache.hadoop.hdfs.protocol.ExtendedBlock;
|
||||
import org.apache.hadoop.hdfs.server.blockmanagement.BlockInfo;
|
||||
import org.hamcrest.CoreMatchers;
|
||||
import org.apache.hadoop.hdfs.server.blockmanagement.BlockManagerTestUtil;
|
||||
|
||||
import org.junit.After;
|
||||
import org.junit.Before;
|
||||
import org.junit.Test;
|
||||
import org.mockito.Mockito;
|
||||
import org.mockito.internal.util.reflection.Whitebox;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.io.OutputStream;
|
||||
import java.util.concurrent.TimeUnit;
|
||||
|
||||
import static org.junit.Assert.assertEquals;
|
||||
import static org.junit.Assert.assertThat;
|
||||
import static org.mockito.Mockito.*;
|
||||
import static org.junit.Assert.assertTrue;
|
||||
|
||||
public class TestNameNodeMetadataConsistency {
|
||||
private static final Path filePath1 = new Path("/testdata1.txt");
|
||||
private static final Path filePath2 = new Path("/testdata2.txt");
|
||||
private static final String TEST_DATA_IN_FUTURE = "This is test data";
|
||||
|
||||
private static final int SCAN_INTERVAL = 1;
|
||||
private static final int SCAN_WAIT = 3;
|
||||
|
@ -75,59 +75,45 @@ public class TestNameNodeMetadataConsistency {
|
|||
@Test
|
||||
public void testGenerationStampInFuture() throws
|
||||
IOException, InterruptedException {
|
||||
|
||||
String testData = " This is test data";
|
||||
int datalen = testData.length();
|
||||
|
||||
cluster.waitActive();
|
||||
|
||||
FileSystem fs = cluster.getFileSystem();
|
||||
OutputStream ostream = fs.create(filePath1);
|
||||
ostream.write(testData.getBytes());
|
||||
ostream.write(TEST_DATA_IN_FUTURE.getBytes());
|
||||
ostream.close();
|
||||
|
||||
ExtendedBlock block = DFSTestUtil.getFirstBlock(fs, filePath1);
|
||||
long genStamp = block.getGenerationStamp();
|
||||
|
||||
// Re-write the Generation Stamp to a Generation Stamp in future.
|
||||
cluster.changeGenStampOfBlock(0, block, genStamp + 1);
|
||||
MiniDFSCluster.DataNodeProperties dnProps = cluster.stopDataNode(0);
|
||||
ExtendedBlock block = DFSTestUtil.getFirstBlock(fs, filePath1);
|
||||
final long genStamp = block.getGenerationStamp();
|
||||
final int datanodeIndex = 0;
|
||||
cluster.changeGenStampOfBlock(datanodeIndex, block, genStamp + 1);
|
||||
// stop the data node so that it won't remove block
|
||||
final DataNodeProperties dnProps = cluster.stopDataNode(datanodeIndex);
|
||||
|
||||
|
||||
// Simulate Namenode forgetting a Block
|
||||
// Simulate Namenode forgetting a Block
|
||||
cluster.restartNameNode(true);
|
||||
BlockInfo bInfo = cluster.getNameNode().getNamesystem().getBlockManager
|
||||
().getStoredBlock(block.getLocalBlock());
|
||||
cluster.getNameNode().getNamesystem().writeLock();
|
||||
BlockInfo bInfo = cluster.getNameNode().getNamesystem().getBlockManager()
|
||||
.getStoredBlock(block.getLocalBlock());
|
||||
cluster.getNameNode().getNamesystem().getBlockManager()
|
||||
.removeBlock(bInfo);
|
||||
cluster.getNameNode().getNamesystem().writeUnlock();
|
||||
|
||||
// we also need to tell block manager that we are in the startup path
|
||||
FSNamesystem spyNameSystem = spy(cluster.getNameNode().getNamesystem());
|
||||
Whitebox.setInternalState(cluster.getNameNode()
|
||||
.getNamesystem().getBlockManager(),
|
||||
"namesystem", spyNameSystem);
|
||||
Whitebox.setInternalState(cluster.getNameNode(),
|
||||
"namesystem", spyNameSystem);
|
||||
Mockito.doReturn(true).when(spyNameSystem).isInStartupSafeMode();
|
||||
BlockManagerTestUtil.setStartupSafeModeForTest(
|
||||
cluster.getNameNode().getNamesystem().getBlockManager());
|
||||
|
||||
|
||||
// Since Data Node is already shutdown we didn't remove blocks
|
||||
cluster.restartDataNode(dnProps);
|
||||
waitTil(TimeUnit.SECONDS.toMillis(SCAN_WAIT));
|
||||
cluster.triggerBlockReports();
|
||||
|
||||
// Give some buffer to process the block reports
|
||||
waitTil(TimeUnit.SECONDS.toMillis(SCAN_WAIT));
|
||||
|
||||
// Make sure that we find all written bytes in future block
|
||||
assertEquals(datalen, cluster.getNameNode().getBytesWithFutureGenerationStamps());
|
||||
|
||||
assertEquals(TEST_DATA_IN_FUTURE.length(),
|
||||
cluster.getNameNode().getBytesWithFutureGenerationStamps());
|
||||
// Assert safemode reason
|
||||
String safeModeMessage = cluster.getNameNode().getNamesystem()
|
||||
.getSafeModeTip();
|
||||
assertThat(safeModeMessage, CoreMatchers.containsString("Name node " +
|
||||
"detected blocks with generation stamps in future"));
|
||||
assertTrue(cluster.getNameNode().getNamesystem().getSafeModeTip().contains(
|
||||
"Name node detected blocks with generation stamps in future"));
|
||||
}
|
||||
|
||||
/**
|
||||
|
|
Loading…
Reference in New Issue