/**
* Keep accessing the given file until the namenode reports that the given block in the file
* contains the given number of corrupt replicas.
*/
public static void waitCorruptReplicas(
FileSystem fs, FSNamesystem ns, Path file, ExtendedBlock b, int corruptRepls)
throws IOException, TimeoutException {
int count = 0;
final int ATTEMPTS = 50;
int repls = ns.getBlockManager().numCorruptReplicas(b.getLocalBlock());
while (repls != corruptRepls && count < ATTEMPTS) {
try {
IOUtils.copyBytes(fs.open(file), new IOUtils.NullOutputStream(), 512, true);
} catch (IOException e) {
// Swallow exceptions
}
System.out.println("Waiting for " + corruptRepls + " corrupt replicas");
repls = ns.getBlockManager().numCorruptReplicas(b.getLocalBlock());
count++;
}
if (count == ATTEMPTS) {
throw new TimeoutException(
"Timed out waiting for corrupt replicas."
+ " Waiting for "
+ corruptRepls
+ ", but only found "
+ repls);
}
}
int addSomeBlocks(SimulatedFSDataset fsdataset, int startingBlockId) throws IOException {
int bytesAdded = 0;
for (int i = startingBlockId; i < startingBlockId + NUMBLOCKS; ++i) {
ExtendedBlock b = new ExtendedBlock(bpid, i, 0, 0);
// we pass expected len as zero, - fsdataset should use the sizeof actual
// data written
ReplicaInPipelineInterface bInfo =
fsdataset.createRbw(StorageType.DEFAULT, b, false).getReplica();
ReplicaOutputStreams out =
bInfo.createStreams(true, DataChecksum.newDataChecksum(DataChecksum.Type.CRC32, 512));
try {
OutputStream dataOut = out.getDataOut();
assertEquals(0, fsdataset.getLength(b));
for (int j = 1; j <= blockIdToLen(i); ++j) {
dataOut.write(j);
assertEquals(j, bInfo.getBytesOnDisk()); // correct length even as we write
bytesAdded++;
}
} finally {
out.close();
}
b.setNumBytes(blockIdToLen(i));
fsdataset.finalizeBlock(b);
assertEquals(blockIdToLen(i), fsdataset.getLength(b));
}
return bytesAdded;
}
/**
* The following test first creates a file. It verifies the block information from a datanode.
* Then, it updates the block with new information and verifies again.
*/
@Test
public void testBlockMetaDataInfo() throws Exception {
MiniDFSCluster cluster = null;
try {
cluster = new MiniDFSCluster.Builder(conf).numDataNodes(3).build();
cluster.waitActive();
// create a file
DistributedFileSystem dfs = (DistributedFileSystem) cluster.getFileSystem();
String filestr = "/foo";
Path filepath = new Path(filestr);
DFSTestUtil.createFile(dfs, filepath, 1024L, (short) 3, 0L);
assertTrue(dfs.exists(filepath));
// get block info
LocatedBlock locatedblock =
getLastLocatedBlock(DFSClientAdapter.getDFSClient(dfs).getNamenode(), filestr);
DatanodeInfo[] datanodeinfo = locatedblock.getLocations();
assertTrue(datanodeinfo.length > 0);
// connect to a data node
DataNode datanode = cluster.getDataNode(datanodeinfo[0].getIpcPort());
InterDatanodeProtocol idp =
DataNodeTestUtils.createInterDatanodeProtocolProxy(datanode, datanodeinfo[0], conf);
// stop block scanner, so we could compare lastScanTime
DataNodeTestUtils.shutdownBlockScanner(datanode);
// verify BlockMetaDataInfo
ExtendedBlock b = locatedblock.getBlock();
InterDatanodeProtocol.LOG.info("b=" + b + ", " + b.getClass());
checkMetaInfo(b, datanode);
long recoveryId = b.getGenerationStamp() + 1;
idp.initReplicaRecovery(new RecoveringBlock(b, locatedblock.getLocations(), recoveryId));
// verify updateBlock
ExtendedBlock newblock =
new ExtendedBlock(
b.getBlockPoolId(), b.getBlockId(), b.getNumBytes() / 2, b.getGenerationStamp() + 1);
idp.updateReplicaUnderRecovery(b, recoveryId, newblock.getNumBytes());
checkMetaInfo(newblock, datanode);
// Verify correct null response trying to init recovery for a missing block
ExtendedBlock badBlock = new ExtendedBlock("fake-pool", b.getBlockId(), 0, 0);
assertNull(
idp.initReplicaRecovery(
new RecoveringBlock(badBlock, locatedblock.getLocations(), recoveryId)));
} finally {
if (cluster != null) {
cluster.shutdown();
}
}
}
private int getTrueReplication(MiniDFSCluster cluster, ExtendedBlock block) throws IOException {
int count = 0;
for (DataNode dn : cluster.getDataNodes()) {
if (DataNodeTestUtils.getFSDataset(dn)
.getStoredBlock(block.getBlockPoolId(), block.getBlockId())
!= null) {
count++;
}
}
return count;
}
/** Convert an ExtendedBlock to a Json map. */
private static Map<String, Object> toJsonMap(final ExtendedBlock extendedblock) {
if (extendedblock == null) {
return null;
}
final Map<String, Object> m = new TreeMap<String, Object>();
m.put("blockPoolId", extendedblock.getBlockPoolId());
m.put("blockId", extendedblock.getBlockId());
m.put("numBytes", extendedblock.getNumBytes());
m.put("generationStamp", extendedblock.getGenerationStamp());
return m;
}
/**
* Corrupt a block on a data node. Replace the block file content with content of 1, 2,
* ...BLOCK_SIZE.
*
* @param block the ExtendedBlock to be corrupted
* @param dn the data node where the block needs to be corrupted
* @throws FileNotFoundException
* @throws IOException
*/
private static void corruptBlock(final ExtendedBlock block, final DataNode dn)
throws FileNotFoundException, IOException {
final File f =
DataNodeTestUtils.getBlockFile(dn, block.getBlockPoolId(), block.getLocalBlock());
final RandomAccessFile raFile = new RandomAccessFile(f, "rw");
final byte[] bytes = new byte[(int) BLOCK_SIZE];
for (int i = 0; i < BLOCK_SIZE; i++) {
bytes[i] = (byte) (i);
}
raFile.write(bytes);
raFile.close();
}
/**
* try to access a block on a data node. If fails - throws exception
*
* @param datanode
* @param lblock
* @throws IOException
*/
private void accessBlock(DatanodeInfo datanode, LocatedBlock lblock) throws IOException {
InetSocketAddress targetAddr = null;
ExtendedBlock block = lblock.getBlock();
targetAddr = NetUtils.createSocketAddr(datanode.getXferAddr());
BlockReader blockReader =
new BlockReaderFactory(new DfsClientConf(conf))
.setInetSocketAddress(targetAddr)
.setBlock(block)
.setFileName(
BlockReaderFactory.getFileName(targetAddr, "test-blockpoolid", block.getBlockId()))
.setBlockToken(lblock.getBlockToken())
.setStartOffset(0)
.setLength(-1)
.setVerifyChecksum(true)
.setClientName("TestDataNodeVolumeFailure")
.setDatanodeInfo(datanode)
.setCachingStrategy(CachingStrategy.newDefaultStrategy())
.setClientCacheContext(ClientContext.getFromConf(conf))
.setConfiguration(conf)
.setTracer(FsTracer.get(conf))
.setRemotePeerFactory(
new RemotePeerFactory() {
@Override
public Peer newConnectedPeer(
InetSocketAddress addr,
Token<BlockTokenIdentifier> blockToken,
DatanodeID datanodeId)
throws IOException {
Peer peer = null;
Socket sock = NetUtils.getDefaultSocketFactory(conf).createSocket();
try {
sock.connect(addr, HdfsConstants.READ_TIMEOUT);
sock.setSoTimeout(HdfsConstants.READ_TIMEOUT);
peer = DFSUtilClient.peerFromSocket(sock);
} finally {
if (peer == null) {
IOUtils.closeSocket(sock);
}
}
return peer;
}
})
.build();
blockReader.close();
}
/**
* Setup a {@link MiniDFSCluster}. Create a block with both {@link State#NORMAL} and {@link
* State#READ_ONLY_SHARED} replicas.
*/
@Before
public void setup() throws IOException, InterruptedException {
conf = new HdfsConfiguration();
SimulatedFSDataset.setFactory(conf);
Configuration[] overlays = new Configuration[NUM_DATANODES];
for (int i = 0; i < overlays.length; i++) {
overlays[i] = new Configuration();
if (i == RO_NODE_INDEX) {
overlays[i].setEnum(
SimulatedFSDataset.CONFIG_PROPERTY_STATE,
i == RO_NODE_INDEX ? READ_ONLY_SHARED : NORMAL);
}
}
cluster =
new MiniDFSCluster.Builder(conf)
.numDataNodes(NUM_DATANODES)
.dataNodeConfOverlays(overlays)
.build();
fs = cluster.getFileSystem();
blockManager = cluster.getNameNode().getNamesystem().getBlockManager();
datanodeManager = blockManager.getDatanodeManager();
client =
new DFSClient(
new InetSocketAddress("localhost", cluster.getNameNodePort()),
cluster.getConfiguration(0));
for (int i = 0; i < NUM_DATANODES; i++) {
DataNode dataNode = cluster.getDataNodes().get(i);
validateStorageState(
BlockManagerTestUtil.getStorageReportsForDatanode(
datanodeManager.getDatanode(dataNode.getDatanodeId())),
i == RO_NODE_INDEX ? READ_ONLY_SHARED : NORMAL);
}
// Create a 1 block file
DFSTestUtil.createFile(fs, PATH, BLOCK_SIZE, BLOCK_SIZE, BLOCK_SIZE, (short) 1, seed);
LocatedBlock locatedBlock = getLocatedBlock();
extendedBlock = locatedBlock.getBlock();
block = extendedBlock.getLocalBlock();
assertThat(locatedBlock.getLocations().length, is(1));
normalDataNode = locatedBlock.getLocations()[0];
readOnlyDataNode =
datanodeManager.getDatanode(cluster.getDataNodes().get(RO_NODE_INDEX).getDatanodeId());
assertThat(normalDataNode, is(not(readOnlyDataNode)));
validateNumberReplicas(1);
// Inject the block into the datanode with READ_ONLY_SHARED storage
cluster.injectBlocks(0, RO_NODE_INDEX, Collections.singleton(block));
// There should now be 2 *locations* for the block
// Must wait until the NameNode has processed the block report for the injected blocks
waitForLocations(2);
}
/**
* go to each block on the 2nd DataNode until it fails...
*
* @param path
* @param size
* @throws IOException
*/
private void triggerFailure(String path, long size) throws IOException {
NamenodeProtocols nn = cluster.getNameNodeRpc();
List<LocatedBlock> locatedBlocks = nn.getBlockLocations(path, 0, size).getLocatedBlocks();
for (LocatedBlock lb : locatedBlocks) {
DatanodeInfo dinfo = lb.getLocations()[1];
ExtendedBlock b = lb.getBlock();
try {
accessBlock(dinfo, lb);
} catch (IOException e) {
System.out.println(
"Failure triggered, on block: "
+ b.getBlockId()
+ "; corresponding volume should be removed by now");
break;
}
}
}
/** Create a file with one block and corrupt some/all of the block replicas. */
private void createAFileWithCorruptedBlockReplicas(
Path filePath, short repl, int corruptBlockCount)
throws IOException, AccessControlException, FileNotFoundException, UnresolvedLinkException,
InterruptedException, TimeoutException {
DFSTestUtil.createFile(dfs, filePath, BLOCK_SIZE, repl, 0);
DFSTestUtil.waitReplication(dfs, filePath, repl);
// Locate the file blocks by asking name node
final LocatedBlocks locatedblocks =
dfs.dfs.getNamenode().getBlockLocations(filePath.toString(), 0L, BLOCK_SIZE);
Assert.assertEquals(repl, locatedblocks.get(0).getLocations().length);
// The file only has one block
LocatedBlock lblock = locatedblocks.get(0);
DatanodeInfo[] datanodeinfos = lblock.getLocations();
ExtendedBlock block = lblock.getBlock();
// corrupt some /all of the block replicas
for (int i = 0; i < corruptBlockCount; i++) {
DatanodeInfo dninfo = datanodeinfos[i];
final DataNode dn = cluster.getDataNode(dninfo.getIpcPort());
corruptBlock(block, dn);
LOG.debug("Corrupted block " + block.getBlockName() + " on data node " + dninfo);
}
}
/** Get a BlockReader for the given block. */
public BlockReader getBlockReader(LocatedBlock testBlock, int offset, int lenToRead)
throws IOException {
InetSocketAddress targetAddr = null;
Socket sock = null;
ExtendedBlock block = testBlock.getBlock();
DatanodeInfo[] nodes = testBlock.getLocations();
targetAddr = NetUtils.createSocketAddr(nodes[0].getName());
sock = NetUtils.getDefaultSocketFactory(conf).createSocket();
sock.connect(targetAddr, HdfsServerConstants.READ_TIMEOUT);
sock.setSoTimeout(HdfsServerConstants.READ_TIMEOUT);
return BlockReaderFactory.newBlockReader(
new DFSClient.Conf(conf),
sock,
targetAddr.toString() + ":" + block.getBlockId(),
block,
testBlock.getBlockToken(),
offset,
lenToRead,
conf.getInt("io.file.buffer.size", 4096),
true,
"");
}
/*
* Wait up to 20s for the given block to be replicated across
* the requested number of racks, with the requested number of
* replicas, and the requested number of replicas still needed.
*/
public static void waitForReplication(
MiniDFSCluster cluster, ExtendedBlock b, int racks, int replicas, int neededReplicas)
throws IOException, TimeoutException, InterruptedException {
int curRacks = 0;
int curReplicas = 0;
int curNeededReplicas = 0;
int count = 0;
final int ATTEMPTS = 20;
do {
Thread.sleep(1000);
int[] r = BlockManagerTestUtil.getReplicaInfo(cluster.getNamesystem(), b.getLocalBlock());
curRacks = r[0];
curReplicas = r[1];
curNeededReplicas = r[2];
count++;
} while ((curRacks != racks || curReplicas != replicas || curNeededReplicas != neededReplicas)
&& count < ATTEMPTS);
if (count == ATTEMPTS) {
throw new TimeoutException(
"Timed out waiting for replication."
+ " Needed replicas = "
+ neededReplicas
+ " Cur needed replicas = "
+ curNeededReplicas
+ " Replicas = "
+ replicas
+ " Cur replicas = "
+ curReplicas
+ " Racks = "
+ racks
+ " Cur racks = "
+ curRacks);
}
}
/*
* XXX (ab) Bulk of this method is copied verbatim from {@link DFSClient}, which is
* bad. Both places should be refactored to provide a method to copy blocks
* around.
*/
private void copyBlock(DFSClient dfs, LocatedBlock lblock, OutputStream fos) throws Exception {
int failures = 0;
InetSocketAddress targetAddr = null;
TreeSet<DatanodeInfo> deadNodes = new TreeSet<DatanodeInfo>();
BlockReader blockReader = null;
ExtendedBlock block = lblock.getBlock();
while (blockReader == null) {
DatanodeInfo chosenNode;
try {
chosenNode = bestNode(dfs, lblock.getLocations(), deadNodes);
targetAddr = NetUtils.createSocketAddr(chosenNode.getXferAddr());
} catch (IOException ie) {
if (failures >= DFSConfigKeys.DFS_CLIENT_MAX_BLOCK_ACQUIRE_FAILURES_DEFAULT) {
throw new IOException("Could not obtain block " + lblock, ie);
}
LOG.info("Could not obtain block from any node: " + ie);
try {
Thread.sleep(10000);
} catch (InterruptedException iex) {
}
deadNodes.clear();
failures++;
continue;
}
try {
String file =
BlockReaderFactory.getFileName(targetAddr, block.getBlockPoolId(), block.getBlockId());
blockReader =
new BlockReaderFactory(dfs.getConf())
.setFileName(file)
.setBlock(block)
.setBlockToken(lblock.getBlockToken())
.setStartOffset(0)
.setLength(-1)
.setVerifyChecksum(true)
.setClientName("fsck")
.setDatanodeInfo(chosenNode)
.setInetSocketAddress(targetAddr)
.setCachingStrategy(CachingStrategy.newDropBehind())
.setClientCacheContext(dfs.getClientContext())
.setConfiguration(namenode.conf)
.setRemotePeerFactory(
new RemotePeerFactory() {
@Override
public Peer newConnectedPeer(InetSocketAddress addr) throws IOException {
Peer peer = null;
Socket s = NetUtils.getDefaultSocketFactory(conf).createSocket();
try {
s.connect(addr, HdfsServerConstants.READ_TIMEOUT);
s.setSoTimeout(HdfsServerConstants.READ_TIMEOUT);
peer =
TcpPeerServer.peerFromSocketAndKey(
s, namenode.getRpcServer().getDataEncryptionKey());
} finally {
if (peer == null) {
IOUtils.closeQuietly(s);
}
}
return peer;
}
})
.build();
} catch (IOException ex) {
// Put chosen node into dead list, continue
LOG.info("Failed to connect to " + targetAddr + ":" + ex);
deadNodes.add(chosenNode);
}
}
byte[] buf = new byte[1024];
int cnt = 0;
boolean success = true;
long bytesRead = 0;
try {
while ((cnt = blockReader.read(buf, 0, buf.length)) > 0) {
fos.write(buf, 0, cnt);
bytesRead += cnt;
}
if (bytesRead != block.getNumBytes()) {
throw new IOException(
"Recorded block size is "
+ block.getNumBytes()
+ ", but datanode returned "
+ bytesRead
+ " bytes");
}
} catch (Exception e) {
LOG.error("Error reading block", e);
success = false;
} finally {
blockReader.close();
}
if (!success) {
throw new Exception("Could not copy block data for " + lblock.getBlock());
}
}
/** Test for {@link FsDatasetImpl#updateReplicaUnderRecovery(ExtendedBlock, long, long)} */
@Test
public void testUpdateReplicaUnderRecovery() throws IOException {
MiniDFSCluster cluster = null;
try {
cluster = new MiniDFSCluster.Builder(conf).numDataNodes(3).build();
cluster.waitActive();
String bpid = cluster.getNamesystem().getBlockPoolId();
// create a file
DistributedFileSystem dfs = (DistributedFileSystem) cluster.getFileSystem();
String filestr = "/foo";
Path filepath = new Path(filestr);
DFSTestUtil.createFile(dfs, filepath, 1024L, (short) 3, 0L);
// get block info
final LocatedBlock locatedblock =
getLastLocatedBlock(DFSClientAdapter.getDFSClient(dfs).getNamenode(), filestr);
final DatanodeInfo[] datanodeinfo = locatedblock.getLocations();
Assert.assertTrue(datanodeinfo.length > 0);
// get DataNode and FSDataset objects
final DataNode datanode = cluster.getDataNode(datanodeinfo[0].getIpcPort());
Assert.assertTrue(datanode != null);
// initReplicaRecovery
final ExtendedBlock b = locatedblock.getBlock();
final long recoveryid = b.getGenerationStamp() + 1;
final long newlength = b.getNumBytes() - 1;
final FsDatasetSpi<?> fsdataset = DataNodeTestUtils.getFSDataset(datanode);
final ReplicaRecoveryInfo rri =
fsdataset.initReplicaRecovery(new RecoveringBlock(b, null, recoveryid));
// check replica
final ReplicaInfo replica =
FsDatasetTestUtil.fetchReplicaInfo(fsdataset, bpid, b.getBlockId());
Assert.assertEquals(ReplicaState.RUR, replica.getState());
// check meta data before update
FsDatasetImpl.checkReplicaFiles(replica);
// case "THIS IS NOT SUPPOSED TO HAPPEN"
// with (block length) != (stored replica's on disk length).
{
// create a block with same id and gs but different length.
final ExtendedBlock tmp =
new ExtendedBlock(
b.getBlockPoolId(),
rri.getBlockId(),
rri.getNumBytes() - 1,
rri.getGenerationStamp());
try {
// update should fail
fsdataset.updateReplicaUnderRecovery(tmp, recoveryid, newlength);
Assert.fail();
} catch (IOException ioe) {
System.out.println("GOOD: getting " + ioe);
}
}
// update
final String storageID =
fsdataset.updateReplicaUnderRecovery(
new ExtendedBlock(b.getBlockPoolId(), rri), recoveryid, newlength);
assertTrue(storageID != null);
} finally {
if (cluster != null) cluster.shutdown();
}
}
public static void checkMetaInfo(ExtendedBlock b, DataNode dn) throws IOException {
Block metainfo =
DataNodeTestUtils.getFSDataset(dn).getStoredBlock(b.getBlockPoolId(), b.getBlockId());
Assert.assertEquals(b.getBlockId(), metainfo.getBlockId());
Assert.assertEquals(b.getNumBytes(), metainfo.getNumBytes());
}
// try reading a block using a BlockReader directly
protected void tryRead(final Configuration conf, LocatedBlock lblock, boolean shouldSucceed) {
InetSocketAddress targetAddr = null;
IOException ioe = null;
BlockReader blockReader = null;
ExtendedBlock block = lblock.getBlock();
try {
DatanodeInfo[] nodes = lblock.getLocations();
targetAddr = NetUtils.createSocketAddr(nodes[0].getXferAddr());
blockReader =
new BlockReaderFactory(new DfsClientConf(conf))
.setFileName(
BlockReaderFactory.getFileName(
targetAddr, "test-blockpoolid", block.getBlockId()))
.setBlock(block)
.setBlockToken(lblock.getBlockToken())
.setInetSocketAddress(targetAddr)
.setStartOffset(0)
.setLength(-1)
.setVerifyChecksum(true)
.setClientName("TestBlockTokenWithDFS")
.setDatanodeInfo(nodes[0])
.setCachingStrategy(CachingStrategy.newDefaultStrategy())
.setClientCacheContext(ClientContext.getFromConf(conf))
.setConfiguration(conf)
.setTracer(FsTracer.get(conf))
.setRemotePeerFactory(
new RemotePeerFactory() {
@Override
public Peer newConnectedPeer(
InetSocketAddress addr,
Token<BlockTokenIdentifier> blockToken,
DatanodeID datanodeId)
throws IOException {
Peer peer = null;
Socket sock = NetUtils.getDefaultSocketFactory(conf).createSocket();
try {
sock.connect(addr, HdfsConstants.READ_TIMEOUT);
sock.setSoTimeout(HdfsConstants.READ_TIMEOUT);
peer = DFSUtilClient.peerFromSocket(sock);
} finally {
if (peer == null) {
IOUtils.closeSocket(sock);
}
}
return peer;
}
})
.build();
} catch (IOException ex) {
ioe = ex;
} finally {
if (blockReader != null) {
try {
blockReader.close();
} catch (IOException e) {
throw new RuntimeException(e);
}
}
}
if (shouldSucceed) {
Assert.assertNotNull(
"OP_READ_BLOCK: access token is invalid, " + "when it is expected to be valid",
blockReader);
} else {
Assert.assertNotNull(
"OP_READ_BLOCK: access token is valid, " + "when it is expected to be invalid", ioe);
Assert.assertTrue(
"OP_READ_BLOCK failed due to reasons other than access token: ",
ioe instanceof InvalidBlockTokenException);
}
}
/**
* Open a DataInputStream to a DataNode so that it can be read from. We get block ID and the IDs
* of the destinations at startup, from the namenode.
*/
private synchronized DatanodeInfo blockSeekTo(long target) throws IOException {
if (target >= getFileLength()) {
throw new IOException("Attempted to read past end of file");
}
// Will be getting a new BlockReader.
if (blockReader != null) {
closeBlockReader(blockReader);
blockReader = null;
}
//
// Connect to best DataNode for desired Block, with potential offset
//
DatanodeInfo chosenNode = null;
int refetchToken = 1; // only need to get a new access token once
boolean connectFailedOnce = false;
while (true) {
//
// Compute desired block
//
LocatedBlock targetBlock = getBlockAt(target, true);
assert (target == pos) : "Wrong postion " + pos + " expect " + target;
long offsetIntoBlock = target - targetBlock.getStartOffset();
DNAddrPair retval = chooseDataNode(targetBlock);
chosenNode = retval.info;
InetSocketAddress targetAddr = retval.addr;
try {
ExtendedBlock blk = targetBlock.getBlock();
Token<BlockTokenIdentifier> accessToken = targetBlock.getBlockToken();
blockReader =
getBlockReader(
targetAddr,
chosenNode,
src,
blk,
accessToken,
offsetIntoBlock,
blk.getNumBytes() - offsetIntoBlock,
buffersize,
verifyChecksum,
dfsClient.clientName);
if (connectFailedOnce) {
DFSClient.LOG.info(
"Successfully connected to " + targetAddr + " for block " + blk.getBlockId());
}
return chosenNode;
} catch (IOException ex) {
if (ex instanceof InvalidBlockTokenException && refetchToken > 0) {
DFSClient.LOG.info(
"Will fetch a new access token and retry, "
+ "access token was invalid when connecting to "
+ targetAddr
+ " : "
+ ex);
/*
* Get a new access token and retry. Retry is needed in 2 cases. 1)
* When both NN and DN re-started while DFSClient holding a cached
* access token. 2) In the case that NN fails to update its
* access key at pre-set interval (by a wide margin) and
* subsequently restarts. In this case, DN re-registers itself with
* NN and receives a new access key, but DN will delete the old
* access key from its memory since it's considered expired based on
* the estimated expiration date.
*/
refetchToken--;
fetchBlockAt(target);
} else {
connectFailedOnce = true;
DFSClient.LOG.warn(
"Failed to connect to "
+ targetAddr
+ " for block"
+ ", add to deadNodes and continue. "
+ ex,
ex);
// Put chosen node into dead list, continue
addToDeadNodes(chosenNode);
}
}
}
}
/**
* Test that copy on write for blocks works correctly
*
* @throws IOException an exception might be thrown
*/
@Test
public void testCopyOnWrite() throws IOException {
Configuration conf = new HdfsConfiguration();
if (simulatedStorage) {
conf.setBoolean(SimulatedFSDataset.CONFIG_PROPERTY_SIMULATED, true);
}
MiniDFSCluster cluster = new MiniDFSCluster.Builder(conf).build();
FileSystem fs = cluster.getFileSystem();
InetSocketAddress addr = new InetSocketAddress("localhost", cluster.getNameNodePort());
DFSClient client = new DFSClient(addr, conf);
try {
// create a new file, write to it and close it.
//
Path file1 = new Path("/filestatus.dat");
FSDataOutputStream stm = AppendTestUtil.createFile(fs, file1, 1);
writeFile(stm);
stm.close();
// Get a handle to the datanode
DataNode[] dn = cluster.listDataNodes();
assertTrue("There should be only one datanode but found " + dn.length, dn.length == 1);
LocatedBlocks locations =
client.getNamenode().getBlockLocations(file1.toString(), 0, Long.MAX_VALUE);
List<LocatedBlock> blocks = locations.getLocatedBlocks();
FSDataset dataset = (FSDataset) dn[0].data;
//
// Create hard links for a few of the blocks
//
for (int i = 0; i < blocks.size(); i = i + 2) {
ExtendedBlock b = blocks.get(i).getBlock();
File f = dataset.getFile(b.getBlockPoolId(), b.getLocalBlock());
File link = new File(f.toString() + ".link");
System.out.println("Creating hardlink for File " + f + " to " + link);
HardLink.createHardLink(f, link);
}
//
// Detach all blocks. This should remove hardlinks (if any)
//
for (int i = 0; i < blocks.size(); i++) {
ExtendedBlock b = blocks.get(i).getBlock();
System.out.println("testCopyOnWrite detaching block " + b);
assertTrue(
"Detaching block " + b + " should have returned true", dataset.unlinkBlock(b, 1));
}
// Since the blocks were already detached earlier, these calls should
// return false
//
for (int i = 0; i < blocks.size(); i++) {
ExtendedBlock b = blocks.get(i).getBlock();
System.out.println("testCopyOnWrite detaching block " + b);
assertTrue(
"Detaching block " + b + " should have returned false", !dataset.unlinkBlock(b, 1));
}
} finally {
fs.close();
cluster.shutdown();
}
}
@VisibleForTesting
void check(String parent, HdfsFileStatus file, Result res) throws IOException {
String path = file.getFullName(parent);
boolean isOpen = false;
if (file.isDir()) {
if (snapshottableDirs != null && snapshottableDirs.contains(path)) {
String snapshotPath =
(path.endsWith(Path.SEPARATOR) ? path : path + Path.SEPARATOR)
+ HdfsConstants.DOT_SNAPSHOT_DIR;
HdfsFileStatus snapshotFileInfo = namenode.getRpcServer().getFileInfo(snapshotPath);
check(snapshotPath, snapshotFileInfo, res);
}
byte[] lastReturnedName = HdfsFileStatus.EMPTY_NAME;
DirectoryListing thisListing;
if (showFiles) {
out.println(path + " <dir>");
}
res.totalDirs++;
do {
assert lastReturnedName != null;
thisListing = namenode.getRpcServer().getListing(path, lastReturnedName, false);
if (thisListing == null) {
return;
}
HdfsFileStatus[] files = thisListing.getPartialListing();
for (int i = 0; i < files.length; i++) {
check(path, files[i], res);
}
lastReturnedName = thisListing.getLastName();
} while (thisListing.hasMore());
return;
}
if (file.isSymlink()) {
if (showFiles) {
out.println(path + " <symlink>");
}
res.totalSymlinks++;
return;
}
long fileLen = file.getLen();
// Get block locations without updating the file access time
// and without block access tokens
LocatedBlocks blocks;
try {
blocks = namenode.getNamesystem().getBlockLocations(path, 0, fileLen, false, false, false);
} catch (FileNotFoundException fnfe) {
blocks = null;
}
if (blocks == null) { // the file is deleted
return;
}
isOpen = blocks.isUnderConstruction();
if (isOpen && !showOpenFiles) {
// We collect these stats about open files to report with default options
res.totalOpenFilesSize += fileLen;
res.totalOpenFilesBlocks += blocks.locatedBlockCount();
res.totalOpenFiles++;
return;
}
res.totalFiles++;
res.totalSize += fileLen;
res.totalBlocks += blocks.locatedBlockCount();
if (showOpenFiles && isOpen) {
out.print(
path
+ " "
+ fileLen
+ " bytes, "
+ blocks.locatedBlockCount()
+ " block(s), OPENFORWRITE: ");
} else if (showFiles) {
out.print(path + " " + fileLen + " bytes, " + blocks.locatedBlockCount() + " block(s): ");
} else {
out.print('.');
}
if (res.totalFiles % 100 == 0) {
out.println();
out.flush();
}
int missing = 0;
int corrupt = 0;
long missize = 0;
int underReplicatedPerFile = 0;
int misReplicatedPerFile = 0;
StringBuilder report = new StringBuilder();
int i = 0;
for (LocatedBlock lBlk : blocks.getLocatedBlocks()) {
ExtendedBlock block = lBlk.getBlock();
boolean isCorrupt = lBlk.isCorrupt();
String blkName = block.toString();
DatanodeInfo[] locs = lBlk.getLocations();
NumberReplicas numberReplicas =
namenode.getNamesystem().getBlockManager().countNodes(block.getLocalBlock());
int liveReplicas = numberReplicas.liveReplicas();
res.totalReplicas += liveReplicas;
short targetFileReplication = file.getReplication();
res.numExpectedReplicas += targetFileReplication;
if (liveReplicas > targetFileReplication) {
res.excessiveReplicas += (liveReplicas - targetFileReplication);
res.numOverReplicatedBlocks += 1;
}
// Check if block is Corrupt
if (isCorrupt) {
corrupt++;
res.corruptBlocks++;
out.print(
"\n"
+ path
+ ": CORRUPT blockpool "
+ block.getBlockPoolId()
+ " block "
+ block.getBlockName()
+ "\n");
}
if (liveReplicas >= minReplication) res.numMinReplicatedBlocks++;
if (liveReplicas < targetFileReplication && liveReplicas > 0) {
res.missingReplicas += (targetFileReplication - liveReplicas);
res.numUnderReplicatedBlocks += 1;
underReplicatedPerFile++;
if (!showFiles) {
out.print("\n" + path + ": ");
}
out.println(
" Under replicated "
+ block
+ ". Target Replicas is "
+ targetFileReplication
+ " but found "
+ liveReplicas
+ " replica(s).");
}
// verify block placement policy
BlockPlacementStatus blockPlacementStatus =
bpPolicy.verifyBlockPlacement(path, lBlk, targetFileReplication);
if (!blockPlacementStatus.isPlacementPolicySatisfied()) {
res.numMisReplicatedBlocks++;
misReplicatedPerFile++;
if (!showFiles) {
if (underReplicatedPerFile == 0) out.println();
out.print(path + ": ");
}
out.println(
" Replica placement policy is violated for "
+ block
+ ". "
+ blockPlacementStatus.getErrorDescription());
}
report.append(i + ". " + blkName + " len=" + block.getNumBytes());
if (liveReplicas == 0) {
report.append(" MISSING!");
res.addMissing(block.toString(), block.getNumBytes());
missing++;
missize += block.getNumBytes();
} else {
report.append(" repl=" + liveReplicas);
if (showLocations || showRacks) {
StringBuilder sb = new StringBuilder("[");
for (int j = 0; j < locs.length; j++) {
if (j > 0) {
sb.append(", ");
}
if (showRacks) sb.append(NodeBase.getPath(locs[j]));
else sb.append(locs[j]);
}
sb.append(']');
report.append(" " + sb.toString());
}
}
report.append('\n');
i++;
}
if ((missing > 0) || (corrupt > 0)) {
if (!showFiles && (missing > 0)) {
out.print(
"\n" + path + ": MISSING " + missing + " blocks of total size " + missize + " B.");
}
res.corruptFiles++;
if (isOpen) {
LOG.info("Fsck: ignoring open file " + path);
} else {
if (doMove) copyBlocksToLostFound(parent, file, blocks);
if (doDelete) deleteCorruptedFile(path);
}
}
if (showFiles) {
if (missing > 0) {
out.print(" MISSING " + missing + " blocks of total size " + missize + " B\n");
} else if (underReplicatedPerFile == 0 && misReplicatedPerFile == 0) {
out.print(" OK\n");
}
if (showBlocks) {
out.print(report.toString() + "\n");
}
}
}
/**
* The following test first creates a file with a few blocks. It randomly truncates the replica of
* the last block stored in each datanode. Finally, it triggers block synchronization to
* synchronize all stored block.
*/
public void testBlockSynchronization() throws Exception {
final int ORG_FILE_SIZE = 3000;
Configuration conf = new HdfsConfiguration();
conf.setLong(DFSConfigKeys.DFS_BLOCK_SIZE_KEY, BLOCK_SIZE);
MiniDFSCluster cluster = null;
try {
cluster = new MiniDFSCluster.Builder(conf).numDataNodes(5).build();
cluster.waitActive();
// create a file
DistributedFileSystem dfs = (DistributedFileSystem) cluster.getFileSystem();
String filestr = "/foo";
Path filepath = new Path(filestr);
DFSTestUtil.createFile(dfs, filepath, ORG_FILE_SIZE, REPLICATION_NUM, 0L);
assertTrue(dfs.exists(filepath));
DFSTestUtil.waitReplication(dfs, filepath, REPLICATION_NUM);
// get block info for the last block
LocatedBlock locatedblock =
TestInterDatanodeProtocol.getLastLocatedBlock(dfs.dfs.getNamenode(), filestr);
DatanodeInfo[] datanodeinfos = locatedblock.getLocations();
assertEquals(REPLICATION_NUM, datanodeinfos.length);
// connect to data nodes
DataNode[] datanodes = new DataNode[REPLICATION_NUM];
for (int i = 0; i < REPLICATION_NUM; i++) {
datanodes[i] = cluster.getDataNode(datanodeinfos[i].getIpcPort());
assertTrue(datanodes[i] != null);
}
// verify Block Info
ExtendedBlock lastblock = locatedblock.getBlock();
DataNode.LOG.info("newblocks=" + lastblock);
for (int i = 0; i < REPLICATION_NUM; i++) {
checkMetaInfo(lastblock, datanodes[i]);
}
DataNode.LOG.info("dfs.dfs.clientName=" + dfs.dfs.clientName);
cluster.getNameNodeRpc().append(filestr, dfs.dfs.clientName);
// expire lease to trigger block recovery.
waitLeaseRecovery(cluster);
Block[] updatedmetainfo = new Block[REPLICATION_NUM];
long oldSize = lastblock.getNumBytes();
lastblock =
TestInterDatanodeProtocol.getLastLocatedBlock(dfs.dfs.getNamenode(), filestr).getBlock();
long currentGS = lastblock.getGenerationStamp();
for (int i = 0; i < REPLICATION_NUM; i++) {
updatedmetainfo[i] =
DataNodeTestUtils.getFSDataset(datanodes[i])
.getStoredBlock(lastblock.getBlockPoolId(), lastblock.getBlockId());
assertEquals(lastblock.getBlockId(), updatedmetainfo[i].getBlockId());
assertEquals(oldSize, updatedmetainfo[i].getNumBytes());
assertEquals(currentGS, updatedmetainfo[i].getGenerationStamp());
}
// verify that lease recovery does not occur when namenode is in safemode
System.out.println("Testing that lease recovery cannot happen during safemode.");
filestr = "/foo.safemode";
filepath = new Path(filestr);
dfs.create(filepath, (short) 1);
cluster.getNameNodeRpc().setSafeMode(HdfsConstants.SafeModeAction.SAFEMODE_ENTER);
assertTrue(dfs.dfs.exists(filestr));
DFSTestUtil.waitReplication(dfs, filepath, (short) 1);
waitLeaseRecovery(cluster);
// verify that we still cannot recover the lease
LeaseManager lm = NameNodeAdapter.getLeaseManager(cluster.getNamesystem());
assertTrue("Found " + lm.countLease() + " lease, expected 1", lm.countLease() == 1);
cluster.getNameNodeRpc().setSafeMode(HdfsConstants.SafeModeAction.SAFEMODE_LEAVE);
} finally {
if (cluster != null) {
cluster.shutdown();
}
}
}
/**
* Test the case that a replica is reported corrupt while it is not in blocksMap. Make sure that
* ArrayIndexOutOfBounds does not thrown. See Hadoop-4351.
*
* <p>TODO HOPS This test fails as it tries to remove a non-existing replica. Calling
* findAndMarkBlockAsCorrupt from a DataNode that does not store any replica for this specific
* block will lead to a tuple did not exist exception. The reason for this is that
* BlockManager.removeStoredBlock is called with a node that does not store a replica and hence
* the delete will not be able to succeed during commit.
*/
@Test
public void testArrayOutOfBoundsException() throws Exception {
MiniDFSCluster cluster = null;
try {
Configuration conf = new HdfsConfiguration();
cluster = new MiniDFSCluster.Builder(conf).numDataNodes(2).build();
cluster.waitActive();
FileSystem fs = cluster.getFileSystem();
final Path FILE_PATH = new Path("/tmp.txt");
final long FILE_LEN = 1L;
DFSTestUtil.createFile(fs, FILE_PATH, FILE_LEN, (short) 2, 1L);
// get the block
final String bpid = cluster.getNamesystem().getBlockPoolId();
File storageDir = cluster.getInstanceStorageDir(0, 0);
File dataDir = MiniDFSCluster.getFinalizedDir(storageDir, bpid);
assertTrue("Data directory does not exist", dataDir.exists());
ExtendedBlock blk = getBlock(bpid, dataDir);
if (blk == null) {
storageDir = cluster.getInstanceStorageDir(0, 1);
dataDir = MiniDFSCluster.getFinalizedDir(storageDir, bpid);
blk = getBlock(bpid, dataDir);
}
assertFalse(
"Data directory does not contain any blocks or there was an " + "IO error", blk == null);
// start a third datanode
cluster.startDataNodes(conf, 1, true, null, null);
ArrayList<DataNode> datanodes = cluster.getDataNodes();
assertEquals(datanodes.size(), 3);
DataNode dataNode = datanodes.get(2);
// report corrupted block by the third datanode
DatanodeRegistration dnR =
DataNodeTestUtils.getDNRegistrationForBP(dataNode, blk.getBlockPoolId());
// Get the storage id of one of the storages on the datanode
String storageId =
cluster
.getNamesystem()
.getBlockManager()
.getDatanodeManager()
.getDatanode(dataNode.getDatanodeId())
.getStorageInfos()[0]
.getStorageID();
cluster
.getNamesystem()
.getBlockManager()
.findAndMarkBlockAsCorrupt(blk, new DatanodeInfo(dnR), storageId, "some test reason");
// open the file
fs.open(FILE_PATH);
// clean up
fs.delete(FILE_PATH, false);
} finally {
if (cluster != null) {
cluster.shutdown();
}
}
}