/**
* Creates an <tt>ArrayBlockingQueue</tt> with the given (fixed) capacity and the specified access
* policy.
*
* @param capacity the capacity of this queue
* @param fair if <tt>true</tt> then queue accesses for threads blocked on insertion or removal,
* are processed in FIFO order; if <tt>false</tt> the access order is unspecified.
* @throws IllegalArgumentException if <tt>capacity</tt> is less than 1
*/
public ArrayBlockingQueue(int capacity, boolean fair) {
if (capacity <= 0) throw new IllegalArgumentException();
this.items = (E[]) new Object[capacity];
lock = new ReentrantLock(fair);
notEmpty = lock.newCondition();
notFull = lock.newCondition();
}
private void init() {
hasFreeThread = lock.newCondition();
hasTask = lock.newCondition();
pool = new LinkedList<WorkThread>();
monitor = new QueueExecuteThread();
freer = new FreeCleanThread();
killer = new KillCleanThread();
}
@SuppressWarnings("unchecked")
public PriorityBlockingDeque(SortedSet<? extends E> c) {
this.lock = new ReentrantLock();
this.notEmpty = lock.newCondition();
this.comparator = (Comparator<? super E>) c.comparator();
addAll(c);
}
/**
* Method to create a new TE LSP initiated in this node
*
* @param destinationId IP AddreStart LSP Errorss of the destination of the LSP
* @param bw Bandwidth requested
* @param bidirectional bidirectional
* @param OFcode
* @throws LSPCreationException
*/
public long addnewLSP(
Inet4Address destinationId, float bw, boolean bidirectional, int OFcode, int lspID)
throws LSPCreationException {
log.info("Adding New LSP to " + destinationId);
// FIXME: mirar esto
// meter structura --> RequestedLSPinformation --> Dependiente de cada tecnologia
// meter campo con el estado del LSP e ir cambiandolo
LSPTE lsp =
new LSPTE(
lspID,
localIP,
destinationId,
bidirectional,
OFcode,
bw,
PathStateParameters.creatingLPS);
LSPList.put(new LSPKey(localIP, lsp.getIdLSP()), lsp);
ReentrantLock lock = new ReentrantLock();
Condition lspEstablished = lock.newCondition();
// log.info("Metemos en Lock list con ID: "+lsp.getIdLSP());
lockList.put(lsp.getIdLSP(), lock);
conditionList.put(lsp.getIdLSP(), lspEstablished);
/*log.info("Size lockList : "+lockList.size());
log.info("Size conditionList : "+conditionList.size());*/
timeIni = System.nanoTime();
log.info("Start to establish path: " + System.nanoTime());
try {
startLSP(lsp);
} catch (LSPCreationException e) {
log.info("Start LSP Error!");
throw e;
}
return lsp.getIdLSP();
}
public GifDecoder(GifAction gifaction)
{
isDestroy = false;
f = 1;
A = new byte[256];
B = 0;
C = 0;
D = 0;
E = false;
F = 0;
O = new ArrayBlockingQueue(15);
P = new ReentrantLock();
Q = P.newCondition();
R = P.newCondition();
S = 0;
T = false;
U = new ArrayList(M);
V = 0;
W = false;
X = null;
Y = null;
Z = false;
aa = 0;
ab = null;
ac = 0;
ad = null;
ae = null;
af = new int[256];
X = gifaction;
}
/**
* Manages the election of which asynchronous saga event processor is responsible for creating a new
* Saga instance, when necessary.
*
* @author Allard Buijze
* @since 2.0
*/
class AsyncSagaCreationElector {
private static final Logger logger = LoggerFactory.getLogger(AsyncSagaCreationElector.class);
private final ReentrantLock votingLock = new ReentrantLock();
private final Condition allVotesCast = votingLock.newCondition();
// guarded by "votingLock"
private int castVotes = 0;
private volatile boolean invocationDetected = false;
/**
* Forces the current thread to wait for the voting to complete if it is responsible for creating
* the Saga. As soon as an invocation has been recorded, the waiting thread is released.
*
* @param didInvocation indicates whether the current processor found a Saga to process
* @param totalVotesExpected The total number of processors expected to cast a vote
* @param isSagaOwner Indicates whether the current processor "owns" the to-be-created saga
* instance.
* @return <code>true</code> if the current processor should create the new instance, <code>false
* </code> otherwise.
*/
public boolean waitForSagaCreationVote(
final boolean didInvocation, final int totalVotesExpected, final boolean isSagaOwner) {
votingLock.lock();
try {
invocationDetected = invocationDetected || didInvocation;
castVotes++;
while (isSagaOwner && !invocationDetected && castVotes < totalVotesExpected) {
try {
allVotesCast.await();
} catch (InterruptedException e) {
// interrupting this process is not supported.
logger.warn(
"This thread has been interrupted, but the interruption has "
+ "been ignored to prevent loss of information.");
}
}
if (isSagaOwner) {
return !invocationDetected;
}
allVotesCast.signalAll();
} finally {
votingLock.unlock();
}
return false;
}
/** Clears the voting counts for a new round. */
public void clear() {
votingLock.lock();
try {
castVotes = 0;
invocationDetected = false;
} finally {
votingLock.unlock();
}
}
}
/**
* Constructor of the class. Initialize all the objects
*
* @param maxSize The size of the buffer
*/
public Buffer(final int maxSize) {
this.maxSize = maxSize;
buffer = new LinkedList<>();
lock = new ReentrantLock();
lines = lock.newCondition();
space = lock.newCondition();
pendingLines = true;
}
public SimpleSemaphore(int permits, boolean fair) {
// TODO - you fill in here to initialize the SimpleSemaphore,
// making sure to allow both fair and non-fair Semaphore
// semantics.
this.permits = permits;
reentrantLock = new ReentrantLock(fair);
condition = reentrantLock.newCondition();
}
public PriorityBlockingDeque(int initialCapacity, Comparator<? super E> comparator) {
// Note: This restriction of at least one is not actually needed,
// but continues for 1.5 compatibility
if (initialCapacity < 1) throw new IllegalArgumentException();
this.lock = new ReentrantLock();
this.notEmpty = lock.newCondition();
this.comparator = comparator;
this.deque = new Object[initialCapacity];
}
private class SpiceArrayAdapterUnderTest extends SpiceArrayAdapter<DataUnderTest> {
private ReentrantLock reentrantLock = new ReentrantLock();
private Condition loadBitmapHasBeenCalledCondition = reentrantLock.newCondition();
private boolean loadBitmapHasBeenCalled = false;
public SpiceArrayAdapterUnderTest(
Context context, BitmapSpiceManager spiceManagerBinary, List<DataUnderTest> data) {
super(context, spiceManagerBinary, data);
}
@Override
public BitmapRequest createRequest(
DataUnderTest data, int imageIndex, int reqWidth, int reqHeight) {
return new BitmapRequest(
mockWebServer.getUrl("/" + data.getImageUrl()).toString(),
reqWidth,
reqHeight,
cacheFile);
}
// ----------------------------------------------------
// ----- Block Test thread until drawable is refreshed.
// ----------------------------------------------------
@Override
protected void loadBitmapAsynchronously(
DataUnderTest octo, ImageView thumbImageView, String tempThumbnailImageFileName) {
super.loadBitmapAsynchronously(octo, thumbImageView, tempThumbnailImageFileName);
reentrantLock.lock();
try {
loadBitmapHasBeenCalled = true;
loadBitmapHasBeenCalledCondition.signal();
} finally {
reentrantLock.unlock();
}
}
public void await(long millisecond) throws InterruptedException {
reentrantLock.lock();
try {
loadBitmapHasBeenCalledCondition.await(millisecond, TimeUnit.MILLISECONDS);
} finally {
reentrantLock.unlock();
}
}
public boolean isLoadBitmapHasBeenCalled() {
return loadBitmapHasBeenCalled;
}
@Override
public SpiceListItemView<DataUnderTest> createView(Context context, ViewGroup parent) {
return new ListItemViewStub(getContext());
}
}
class PausableThreadPoolExecutor extends ThreadPoolExecutor
implements ExecutorRemoteControllerService {
public PausableThreadPoolExecutor(
int corePoolSize,
int maximumPoolSize,
long keepAliveTime,
TimeUnit unit,
BlockingQueue<Runnable> workQueue,
ThreadFactory threadFactory,
RejectedExecutionHandler handler) {
super(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue, threadFactory, handler);
}
private boolean isPaused;
private ReentrantLock pauseLock = new ReentrantLock();
private Condition unpaused = pauseLock.newCondition();
@Override
protected void beforeExecute(Thread t, Runnable r) {
super.beforeExecute(t, r);
pauseLock.lock();
try {
while (isPaused) unpaused.await();
} catch (InterruptedException ie) {
t.interrupt();
} finally {
pauseLock.unlock();
}
}
@Override
public void pause() {
pauseLock.lock();
try {
isPaused = true;
} finally {
pauseLock.unlock();
}
}
@Override
public void resume() {
pauseLock.lock();
try {
isPaused = false;
unpaused.signalAll();
} finally {
pauseLock.unlock();
}
}
}
/**
* Constructor.
*
* @param thread The MoSync thread.
*/
public MoSyncCameraController(MoSyncThread thread) {
mMoSyncThread = thread;
lock = new ReentrantLock();
mPreview = null;
condition = lock.newCondition();
dataReady = false;
userWidths = new ArrayList<Integer>();
userHeights = new ArrayList<Integer>();
mCameraParametersList = new ArrayList<Camera.Parameters>();
mNumCameras = numberOfCameras();
initilizeCameras();
rawMode = false;
mCurrentCameraIndex = 0;
}
public class ManualFlag extends Flag {
private final ReentrantLock lock = new ReentrantLock();
private final Condition condition = lock.newCondition();
private boolean set;
public ManualFlag(boolean state) {
set = state;
}
public ManualFlag() {
this(false);
}
@Override
public void set() {
lock.lock();
try {
set = true;
condition.signalAll();
} finally {
lock.unlock();
}
}
public void reset() {
lock.lock();
try {
set = false;
} finally {
lock.unlock();
}
}
@Override
public void await() {
lock.lock();
while (true) {
try {
while (!set) condition.await();
return;
} catch (InterruptedException ex) {
continue;
} finally {
lock.unlock();
}
}
}
}
/**
* Transfer the local file to the hypervisor's importdir
*
* @param from Local file name
* @throws OccpException
*/
public void stageFile(String from) throws OccpException {
boolean transferSuccess = false;
/*
* Only VBox requires transfer to the host; Esxi allows uploads
*/
if (this.hv.getClass() == OccpVBoxHV.class) {
try {
completionLock.lock();
if (!completed_conditions.containsKey(from)) {
completed_conditions.put(from, completionLock.newCondition());
} else {
while (!completed_transfers.containsKey(from)) {
try {
completed_conditions.get(from).await();
} catch (InterruptedException e) {
throw new VMOperationFailedException(
hv.getName(), vm.getName(), ErrorCode.TRANSFER_TO, "Transfer interrupted", e);
}
}
// If it has been transferred, say we did it, otherwise try again
if (completed_transfers.get(from) == true) {
return;
}
}
} finally {
completionLock.unlock();
}
boolean hasPath = (from.lastIndexOf('/') >= 0);
String to = from;
if (hasPath) {
to = from.substring(from.lastIndexOf('/') + 1);
}
try {
this.hv.transferFileToVM(this.vm, from, "/mnt/" + OccpAdmin.scenarioName + "/" + to, false);
transferSuccess = true;
} finally {
completionLock.lock();
try {
completed_transfers.put(from, transferSuccess);
completed_conditions.get(from).signal();
} finally {
completionLock.unlock();
}
}
}
}
/** for array */
public class TestArrayBlockingQueueforList {
public ReentrantLock lock = new ReentrantLock();
public Condition condition = lock.newCondition();
public List<Integer> list = new ArrayList<Integer>();
boolean isEmpty = true;
public static void main(String[] args) {
TestArrayBlockingQueueforList test = new TestArrayBlockingQueueforList();
Resource res = new Resource();
new Thread(new ProducerThread(test.getCondition(), test.getLock(), res)).start();
new Thread(new ConsumerThread(test.getCondition(), test.getLock(), res)).start();
}
public ReentrantLock getLock() {
return lock;
}
public void setLock(ReentrantLock lock) {
this.lock = lock;
}
public Condition getCondition() {
return condition;
}
public void setCondition(Condition condition) {
this.condition = condition;
}
public List<Integer> getList() {
return list;
}
public void setList(List<Integer> list) {
this.list = list;
}
public boolean isEmpty() {
return isEmpty;
}
public void setEmpty(boolean isEmpty) {
this.isEmpty = isEmpty;
}
}
/**
* Constructs the consumer which will read from the given destination and is a child of the given
* context.
*
* @param destination the destination that this consumer will read from
* @param hazelcastMQContext the parent context of this consumer
*/
DefaultHazelcastMQConsumer(String destination, DefaultHazelcastMQContext hazelcastMQContext) {
super();
this.destination = destination;
this.receiveLock = new ReentrantLock();
this.receiveCondition = receiveLock.newCondition();
this.closed = false;
this.active = false;
this.hazelcastMQContext = hazelcastMQContext;
this.config = hazelcastMQContext.getHazelcastMQInstance().getConfig();
HazelcastInstance hazelcast =
this.hazelcastMQContext.getHazelcastMQInstance().getConfig().getHazelcastInstance();
IdGenerator idGenerator = hazelcast.getIdGenerator("hazelcastmqconsumer");
this.id = "hazelcastmqconsumer-" + String.valueOf(idGenerator.newId());
}
public class MyService {
private ReentrantLock lock = new ReentrantLock();
private Condition condition = lock.newCondition();
public void waitMethod() {
try {
lock.lock();
System.out.println("A");
condition.await();
System.out.println("B");
} catch (InterruptedException e) {
e.printStackTrace();
} finally {
lock.unlock();
System.out.println("锁释放了！");
}
}
}
private static final class Notifier {
private final ReentrantLock lock = new ReentrantLock();
private final Condition condition = lock.newCondition();
private volatile TimeValue timeout;
public Notifier(TimeValue timeout) {
assert timeout != null;
this.timeout = timeout;
}
public void await() {
lock.lock();
try {
condition.await(timeout.millis(), TimeUnit.MILLISECONDS);
} catch (InterruptedException e) {
// we intentionally do not want to restore the interruption flag, we're about to shutdown
// anyway
} finally {
lock.unlock();
}
}
public void setTimeout(TimeValue timeout) {
assert timeout != null;
this.timeout = timeout;
doNotify();
}
public TimeValue getTimeout() {
return timeout;
}
public void doNotify() {
lock.lock();
try {
condition.signalAll();
} finally {
lock.unlock();
}
}
}
public class BrokenOrderingReentrantLock implements Runnable {
private final ReentrantLock lock1 = new ReentrantLock();
private final ReentrantLock lock2 = new ReentrantLock();
private final Condition condition = lock1.newCondition();
public static void main(String[] args) throws InterruptedException {
BrokenOrderingReentrantLock runnable = new BrokenOrderingReentrantLock();
Thread thread1 = new Thread(runnable, "thread1");
Thread thread2 = new Thread(runnable, "thread2");
thread1.start();
Thread.sleep(500);
thread2.start();
}
@Override
public void run() {
try {
String threadName = Thread.currentThread().getName();
lock1.lock();
try {
System.out.println(threadName + " has lock1");
lock2.lock();
try {
System.out.println(threadName + " has lock2");
lock1.lock();
try {
System.out.println(threadName + " reenters lock1");
condition.await(1, TimeUnit.SECONDS);
} finally {
lock1.unlock();
}
} finally {
lock2.unlock();
}
} finally {
lock1.unlock();
}
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
@SuppressWarnings("unchecked")
public PriorityBlockingDeque(Collection<? extends E> c) {
this.lock = new ReentrantLock();
this.notEmpty = lock.newCondition();
if (c instanceof SortedSet<?>) {
SortedSet<? extends E> ss = (SortedSet<? extends E>) c;
this.comparator = (Comparator<? super E>) ss.comparator();
addAll(ss);
} else if (c instanceof PriorityDeque<?>) {
PriorityDeque<? extends E> pq = (PriorityDeque<? extends E>) c;
this.comparator = (Comparator<? super E>) pq.comparator();
initFromPriorityDeque(pq);
} else if (c instanceof PriorityBlockingDeque<?>) {
PriorityBlockingDeque<? extends E> pq = (PriorityBlockingDeque<? extends E>) c;
this.comparator = (Comparator<? super E>) pq.comparator();
initFromPriorityBlockingDeque(pq);
} else {
this.comparator = null;
addAll(c);
}
}
ConsumerImpl(
URI uri,
String submitQName,
String statusQName,
String statusTName,
String heartbeatTName,
String commandTName,
IEventConnectorService service,
IEventService eservice)
throws EventException {
super(uri, submitQName, statusQName, statusTName, commandTName, service, eservice);
this.lock = new ReentrantLock();
this.paused = lock.newCondition();
durable = true;
consumerId = UUID.randomUUID();
name = "Consumer " + consumerId; // This will hopefully be changed to something meaningful...
this.processes = new Hashtable<>(7); // Synch!
this.heartbeatTopicName = heartbeatTName;
connect();
}
public class MyService {
private ReentrantLock lock = new ReentrantLock();
private Condition condition = lock.newCondition();
private boolean hasValue = false;
public void set() {
try {
lock.lock();
while (hasValue == true) {
condition.await();
}
System.out.println("打印★");
hasValue = true;
condition.signal();
} catch (InterruptedException e) {
e.printStackTrace();
} finally {
lock.unlock();
}
}
public void get() {
try {
lock.lock();
while (hasValue == false) {
condition.await();
}
System.out.println("打印☆");
hasValue = false;
condition.signal();
} catch (InterruptedException e) {
e.printStackTrace();
} finally {
lock.unlock();
}
}
}
class SuspendableThreadPoolExecutor extends ThreadPoolExecutor {
private boolean available = false;
private ReentrantLock suspendLock = new ReentrantLock();
private Condition availableCondition = suspendLock.newCondition();
public SuspendableThreadPoolExecutor(ThreadFactory threadFactory) {
super(1, 1, 0L, TimeUnit.MILLISECONDS, new LinkedBlockingQueue<Runnable>(), threadFactory);
}
@Override
protected void beforeExecute(Thread thread, Runnable task) {
super.beforeExecute(thread, task);
suspendLock.lock();
try {
while (!available) {
availableCondition.await();
}
} catch (InterruptedException interruptedException) {
thread.interrupt();
} finally {
suspendLock.unlock();
}
}
public void setAvailable(boolean available) {
suspendLock.lock();
try {
this.available = available;
if (available) {
availableCondition.signalAll();
}
} finally {
suspendLock.unlock();
}
}
}
// This only knows how to deal with a single srcIndex for a given targetIndex.
// In case the src task generates multiple outputs for the same target Index
// (multiple src-indices), modifications will be required.
public class ShuffleManager implements FetcherCallback {
private static final Log LOG = LogFactory.getLog(ShuffleManager.class);
private final TezInputContext inputContext;
private final int numInputs;
private final FetchedInputAllocator inputManager;
private final ListeningExecutorService fetcherExecutor;
private final ListeningExecutorService schedulerExecutor;
private final RunShuffleCallable schedulerCallable = new RunShuffleCallable();
private final BlockingQueue<FetchedInput> completedInputs;
private final AtomicBoolean inputReadyNotificationSent = new AtomicBoolean(false);
private final Set<InputIdentifier> completedInputSet;
private final ConcurrentMap<String, InputHost> knownSrcHosts;
private final BlockingQueue<InputHost> pendingHosts;
private final Set<InputAttemptIdentifier> obsoletedInputs;
private Set<Fetcher> runningFetchers;
private final AtomicInteger numCompletedInputs = new AtomicInteger(0);
private final long startTime;
private long lastProgressTime;
// Required to be held when manipulating pendingHosts
private final ReentrantLock lock = new ReentrantLock();
private final Condition wakeLoop = lock.newCondition();
private final int numFetchers;
// Parameters required by Fetchers
private final SecretKey shuffleSecret;
private final CompressionCodec codec;
private final int ifileBufferSize;
private final boolean ifileReadAhead;
private final int ifileReadAheadLength;
private final String srcNameTrimmed;
private final AtomicBoolean isShutdown = new AtomicBoolean(false);
private final TezCounter shuffledInputsCounter;
private final TezCounter failedShufflesCounter;
private final TezCounter bytesShuffledCounter;
private final TezCounter decompressedDataSizeCounter;
private final TezCounter bytesShuffledToDiskCounter;
private final TezCounter bytesShuffledToMemCounter;
private volatile Throwable shuffleError;
private final HttpConnectionParams httpConnectionParams;
// TODO More counters - FetchErrors, speed?
public ShuffleManager(
TezInputContext inputContext,
Configuration conf,
int numInputs,
int bufferSize,
boolean ifileReadAheadEnabled,
int ifileReadAheadLength,
CompressionCodec codec,
FetchedInputAllocator inputAllocator)
throws IOException {
this.inputContext = inputContext;
this.numInputs = numInputs;
this.shuffledInputsCounter =
inputContext.getCounters().findCounter(TaskCounter.NUM_SHUFFLED_INPUTS);
this.failedShufflesCounter =
inputContext.getCounters().findCounter(TaskCounter.NUM_FAILED_SHUFFLE_INPUTS);
this.bytesShuffledCounter = inputContext.getCounters().findCounter(TaskCounter.SHUFFLE_BYTES);
this.decompressedDataSizeCounter =
inputContext.getCounters().findCounter(TaskCounter.SHUFFLE_BYTES_DECOMPRESSED);
this.bytesShuffledToDiskCounter =
inputContext.getCounters().findCounter(TaskCounter.SHUFFLE_BYTES_TO_DISK);
this.bytesShuffledToMemCounter =
inputContext.getCounters().findCounter(TaskCounter.SHUFFLE_BYTES_TO_MEM);
this.ifileBufferSize = bufferSize;
this.ifileReadAhead = ifileReadAheadEnabled;
this.ifileReadAheadLength = ifileReadAheadLength;
this.codec = codec;
this.inputManager = inputAllocator;
this.srcNameTrimmed = TezUtils.cleanVertexName(inputContext.getSourceVertexName());
completedInputSet =
Collections.newSetFromMap(new ConcurrentHashMap<InputIdentifier, Boolean>(numInputs));
completedInputs = new LinkedBlockingQueue<FetchedInput>(numInputs);
knownSrcHosts = new ConcurrentHashMap<String, InputHost>();
pendingHosts = new LinkedBlockingQueue<InputHost>();
obsoletedInputs =
Collections.newSetFromMap(new ConcurrentHashMap<InputAttemptIdentifier, Boolean>());
runningFetchers = Collections.newSetFromMap(new ConcurrentHashMap<Fetcher, Boolean>());
int maxConfiguredFetchers =
conf.getInt(
TezJobConfig.TEZ_RUNTIME_SHUFFLE_PARALLEL_COPIES,
TezJobConfig.TEZ_RUNTIME_SHUFFLE_PARALLEL_COPIES_DEFAULT);
this.numFetchers = Math.min(maxConfiguredFetchers, numInputs);
ExecutorService fetcherRawExecutor =
Executors.newFixedThreadPool(
numFetchers,
new ThreadFactoryBuilder()
.setDaemon(true)
.setNameFormat("Fetcher [" + srcNameTrimmed + "] #%d")
.build());
this.fetcherExecutor = MoreExecutors.listeningDecorator(fetcherRawExecutor);
ExecutorService schedulerRawExecutor =
Executors.newFixedThreadPool(
1,
new ThreadFactoryBuilder()
.setDaemon(true)
.setNameFormat("ShuffleRunner [" + srcNameTrimmed + "]")
.build());
this.schedulerExecutor = MoreExecutors.listeningDecorator(schedulerRawExecutor);
this.startTime = System.currentTimeMillis();
this.lastProgressTime = startTime;
this.shuffleSecret =
ShuffleUtils.getJobTokenSecretFromTokenBytes(
inputContext.getServiceConsumerMetaData(
TezConfiguration.TEZ_SHUFFLE_HANDLER_SERVICE_ID));
httpConnectionParams = ShuffleUtils.constructHttpShuffleConnectionParams(conf);
LOG.info(
this.getClass().getSimpleName()
+ " : numInputs="
+ numInputs
+ ", compressionCodec="
+ (codec == null ? "NoCompressionCodec" : codec.getClass().getName())
+ ", numFetchers="
+ numFetchers
+ ", ifileBufferSize="
+ ifileBufferSize
+ ", ifileReadAheadEnabled="
+ ifileReadAhead
+ ", ifileReadAheadLength="
+ ifileReadAheadLength
+ ", "
+ httpConnectionParams.toString());
}
public void run() throws IOException {
Preconditions.checkState(inputManager != null, "InputManager must be configured");
ListenableFuture<Void> runShuffleFuture = schedulerExecutor.submit(schedulerCallable);
Futures.addCallback(runShuffleFuture, new SchedulerFutureCallback());
// Shutdown this executor once this task, and the callback complete.
schedulerExecutor.shutdown();
}
private class RunShuffleCallable implements Callable<Void> {
@Override
public Void call() throws Exception {
while (!isShutdown.get() && numCompletedInputs.get() < numInputs) {
lock.lock();
try {
if (runningFetchers.size() >= numFetchers || pendingHosts.isEmpty()) {
if (numCompletedInputs.get() < numInputs) {
wakeLoop.await();
}
}
} finally {
lock.unlock();
}
if (shuffleError != null) {
// InputContext has already been informed of a fatal error. Relying on
// tez to kill the task.
break;
}
if (LOG.isDebugEnabled()) {
LOG.debug("NumCompletedInputs: " + numCompletedInputs);
}
if (numCompletedInputs.get() < numInputs && !isShutdown.get()) {
lock.lock();
try {
int maxFetchersToRun = numFetchers - runningFetchers.size();
int count = 0;
while (pendingHosts.peek() != null && !isShutdown.get()) {
InputHost inputHost = null;
try {
inputHost = pendingHosts.take();
} catch (InterruptedException e) {
if (isShutdown.get()) {
LOG.info(
"Interrupted and hasBeenShutdown, Breaking out of ShuffleScheduler Loop");
break;
} else {
throw e;
}
}
if (LOG.isDebugEnabled()) {
LOG.debug("Processing pending host: " + inputHost.toDetailedString());
}
if (inputHost.getNumPendingInputs() > 0 && !isShutdown.get()) {
LOG.info("Scheduling fetch for inputHost: " + inputHost.getIdentifier());
Fetcher fetcher = constructFetcherForHost(inputHost);
runningFetchers.add(fetcher);
if (isShutdown.get()) {
LOG.info("hasBeenShutdown, Breaking out of ShuffleScheduler Loop");
}
ListenableFuture<FetchResult> future = fetcherExecutor.submit(fetcher);
Futures.addCallback(future, new FetchFutureCallback(fetcher));
if (++count >= maxFetchersToRun) {
break;
}
} else {
if (LOG.isDebugEnabled()) {
LOG.debug(
"Skipping host: "
+ inputHost.getIdentifier()
+ " since it has no inputs to process");
}
}
}
} finally {
lock.unlock();
}
}
}
LOG.info(
"Shutting down FetchScheduler, Was Interrupted: "
+ Thread.currentThread().isInterrupted());
// TODO NEWTEZ Maybe clean up inputs.
if (!fetcherExecutor.isShutdown()) {
fetcherExecutor.shutdownNow();
}
return null;
}
}
private Fetcher constructFetcherForHost(InputHost inputHost) {
FetcherBuilder fetcherBuilder =
new FetcherBuilder(
ShuffleManager.this,
httpConnectionParams,
inputManager,
inputContext.getApplicationId(),
shuffleSecret,
srcNameTrimmed);
if (codec != null) {
fetcherBuilder.setCompressionParameters(codec);
}
fetcherBuilder.setIFileParams(ifileReadAhead, ifileReadAheadLength);
// Remove obsolete inputs from the list being given to the fetcher. Also
// remove from the obsolete list.
List<InputAttemptIdentifier> pendingInputsForHost = inputHost.clearAndGetPendingInputs();
for (Iterator<InputAttemptIdentifier> inputIter = pendingInputsForHost.iterator();
inputIter.hasNext(); ) {
InputAttemptIdentifier input = inputIter.next();
// Avoid adding attempts which have already completed.
if (completedInputSet.contains(input.getInputIdentifier())) {
inputIter.remove();
continue;
}
// Avoid adding attempts which have been marked as OBSOLETE
if (obsoletedInputs.contains(input)) {
inputIter.remove();
}
}
// TODO NEWTEZ Maybe limit the number of inputs being given to a single
// fetcher, especially in the case where #hosts < #fetchers
fetcherBuilder.assignWork(
inputHost.getHost(),
inputHost.getPort(),
inputHost.getSrcPhysicalIndex(),
pendingInputsForHost);
LOG.info(
"Created Fetcher for host: "
+ inputHost.getHost()
+ ", with inputs: "
+ pendingInputsForHost);
return fetcherBuilder.build();
}
/////////////////// Methods for InputEventHandler
public void addKnownInput(
String hostName,
int port,
InputAttemptIdentifier srcAttemptIdentifier,
int srcPhysicalIndex) {
String identifier = InputHost.createIdentifier(hostName, port);
InputHost host = knownSrcHosts.get(identifier);
if (host == null) {
host = new InputHost(hostName, port, inputContext.getApplicationId(), srcPhysicalIndex);
assert identifier.equals(host.getIdentifier());
InputHost old = knownSrcHosts.putIfAbsent(identifier, host);
if (old != null) {
host = old;
}
}
if (LOG.isDebugEnabled()) {
LOG.debug("Adding input: " + srcAttemptIdentifier + ", to host: " + host);
}
host.addKnownInput(srcAttemptIdentifier);
lock.lock();
try {
boolean added = pendingHosts.offer(host);
if (!added) {
String errorMessage = "Unable to add host: " + host.getIdentifier() + " to pending queue";
LOG.error(errorMessage);
throw new TezUncheckedException(errorMessage);
}
wakeLoop.signal();
} finally {
lock.unlock();
}
}
public void addCompletedInputWithNoData(InputAttemptIdentifier srcAttemptIdentifier) {
InputIdentifier inputIdentifier = srcAttemptIdentifier.getInputIdentifier();
LOG.info("No input data exists for SrcTask: " + inputIdentifier + ". Marking as complete.");
if (!completedInputSet.contains(inputIdentifier)) {
synchronized (completedInputSet) {
if (!completedInputSet.contains(inputIdentifier)) {
registerCompletedInput(new NullFetchedInput(srcAttemptIdentifier));
}
}
}
// Awake the loop to check for termination.
lock.lock();
try {
wakeLoop.signal();
} finally {
lock.unlock();
}
}
public void addCompletedInputWithData(
InputAttemptIdentifier srcAttemptIdentifier, FetchedInput fetchedInput) throws IOException {
InputIdentifier inputIdentifier = srcAttemptIdentifier.getInputIdentifier();
LOG.info("Received Data via Event: " + srcAttemptIdentifier + " to " + fetchedInput.getType());
// Count irrespective of whether this is a copy of an already fetched input
lock.lock();
try {
lastProgressTime = System.currentTimeMillis();
} finally {
lock.unlock();
}
boolean committed = false;
if (!completedInputSet.contains(inputIdentifier)) {
synchronized (completedInputSet) {
if (!completedInputSet.contains(inputIdentifier)) {
fetchedInput.commit();
committed = true;
registerCompletedInput(fetchedInput);
}
}
}
if (!committed) {
fetchedInput.abort(); // If this fails, the fetcher may attempt another
// abort.
} else {
lock.lock();
try {
// Signal the wakeLoop to check for termination.
wakeLoop.signal();
} finally {
lock.unlock();
}
}
}
public synchronized void obsoleteKnownInput(InputAttemptIdentifier srcAttemptIdentifier) {
obsoletedInputs.add(srcAttemptIdentifier);
// TODO NEWTEZ Maybe inform the fetcher about this. For now, this is used during the initial
// fetch list construction.
}
/////////////////// End of Methods for InputEventHandler
/////////////////// Methods from FetcherCallbackHandler
@Override
public void fetchSucceeded(
String host,
InputAttemptIdentifier srcAttemptIdentifier,
FetchedInput fetchedInput,
long fetchedBytes,
long decompressedLength,
long copyDuration)
throws IOException {
InputIdentifier inputIdentifier = srcAttemptIdentifier.getInputIdentifier();
LOG.info(
"Completed fetch for attempt: " + srcAttemptIdentifier + " to " + fetchedInput.getType());
// Count irrespective of whether this is a copy of an already fetched input
lock.lock();
try {
lastProgressTime = System.currentTimeMillis();
} finally {
lock.unlock();
}
boolean committed = false;
if (!completedInputSet.contains(inputIdentifier)) {
synchronized (completedInputSet) {
if (!completedInputSet.contains(inputIdentifier)) {
fetchedInput.commit();
committed = true;
// Processing counters for completed and commit fetches only. Need
// additional counters for excessive fetches - which primarily comes
// in after speculation or retries.
shuffledInputsCounter.increment(1);
bytesShuffledCounter.increment(fetchedBytes);
if (fetchedInput.getType() == Type.MEMORY) {
bytesShuffledToMemCounter.increment(fetchedBytes);
} else {
bytesShuffledToDiskCounter.increment(fetchedBytes);
}
decompressedDataSizeCounter.increment(decompressedLength);
registerCompletedInput(fetchedInput);
}
}
}
if (!committed) {
fetchedInput.abort(); // If this fails, the fetcher may attempt another abort.
} else {
lock.lock();
try {
// Signal the wakeLoop to check for termination.
wakeLoop.signal();
} finally {
lock.unlock();
}
}
// TODO NEWTEZ Maybe inform fetchers, in case they have an alternate attempt of the same task in
// their queue.
}
@Override
public void fetchFailed(
String host, InputAttemptIdentifier srcAttemptIdentifier, boolean connectFailed) {
// TODO NEWTEZ. Implement logic to report fetch failures after a threshold.
// For now, reporting immediately.
LOG.info(
"Fetch failed for src: "
+ srcAttemptIdentifier
+ "InputIdentifier: "
+ srcAttemptIdentifier
+ ", connectFailed: "
+ connectFailed);
failedShufflesCounter.increment(1);
if (srcAttemptIdentifier == null) {
String message = "Received fetchFailure for an unknown src (null)";
LOG.fatal(message);
inputContext.fatalError(null, message);
} else {
InputReadErrorEvent readError =
new InputReadErrorEvent(
"Fetch failure while fetching from "
+ TezRuntimeUtils.getTaskAttemptIdentifier(
inputContext.getSourceVertexName(),
srcAttemptIdentifier.getInputIdentifier().getInputIndex(),
srcAttemptIdentifier.getAttemptNumber()),
srcAttemptIdentifier.getInputIdentifier().getInputIndex(),
srcAttemptIdentifier.getAttemptNumber());
List<Event> failedEvents = Lists.newArrayListWithCapacity(1);
failedEvents.add(readError);
inputContext.sendEvents(failedEvents);
}
}
/////////////////// End of Methods from FetcherCallbackHandler
public void shutdown() throws InterruptedException {
if (!isShutdown.getAndSet(true)) {
// Shut down any pending fetchers
LOG.info(
"Shutting down pending fetchers on source"
+ srcNameTrimmed
+ ": "
+ runningFetchers.size());
lock.lock();
try {
wakeLoop.signal(); // signal the fetch-scheduler
for (Fetcher fetcher : runningFetchers) {
fetcher.shutdown(); // This could be parallelized.
}
} finally {
lock.unlock();
}
if (this.schedulerExecutor != null && !this.schedulerExecutor.isShutdown()) {
this.schedulerExecutor.shutdownNow();
}
if (this.fetcherExecutor != null && !this.fetcherExecutor.isShutdown()) {
this.fetcherExecutor.shutdownNow(); // Interrupts all running fetchers.
}
}
// All threads are shutdown. It is safe to shutdown SSL factory
if (httpConnectionParams.isSSLShuffleEnabled()) {
HttpConnection.cleanupSSLFactory();
}
}
private void registerCompletedInput(FetchedInput fetchedInput) {
lock.lock();
try {
completedInputSet.add(fetchedInput.getInputAttemptIdentifier().getInputIdentifier());
completedInputs.add(fetchedInput);
if (!inputReadyNotificationSent.getAndSet(true)) {
// TODO Should eventually be controlled by Inputs which are processing the data.
inputContext.inputIsReady();
}
int numComplete = numCompletedInputs.incrementAndGet();
if (numComplete == numInputs) {
LOG.info("All inputs fetched for input vertex : " + inputContext.getSourceVertexName());
}
} finally {
lock.unlock();
}
}
/////////////////// Methods for walking the available inputs
/** @return true if there is another input ready for consumption. */
public boolean newInputAvailable() {
FetchedInput head = completedInputs.peek();
if (head == null || head instanceof NullFetchedInput) {
return false;
} else {
return true;
}
}
/** @return true if all of the required inputs have been fetched. */
public boolean allInputsFetched() {
lock.lock();
try {
return numCompletedInputs.get() == numInputs;
} finally {
lock.unlock();
}
}
/**
* @return the next available input, or null if there are no available inputs. This method will
* block if there are currently no available inputs, but more may become available.
*/
public FetchedInput getNextInput() throws InterruptedException {
FetchedInput input = null;
do {
// Check for no additional inputs
lock.lock();
try {
input = completedInputs.peek();
if (input == null && allInputsFetched()) {
break;
}
} finally {
lock.unlock();
}
input = completedInputs.take(); // block
} while (input instanceof NullFetchedInput);
return input;
}
/////////////////// End of methods for walking the available inputs
/**
* Fake input that is added to the completed input list in case an input does not have any data.
*/
private class NullFetchedInput extends FetchedInput {
public NullFetchedInput(InputAttemptIdentifier inputAttemptIdentifier) {
super(Type.MEMORY, -1, -1, inputAttemptIdentifier, null);
}
@Override
public OutputStream getOutputStream() throws IOException {
throw new UnsupportedOperationException("Not supported for NullFetchedInput");
}
@Override
public InputStream getInputStream() throws IOException {
throw new UnsupportedOperationException("Not supported for NullFetchedInput");
}
@Override
public void commit() throws IOException {
throw new UnsupportedOperationException("Not supported for NullFetchedInput");
}
@Override
public void abort() throws IOException {
throw new UnsupportedOperationException("Not supported for NullFetchedInput");
}
@Override
public void free() {
throw new UnsupportedOperationException("Not supported for NullFetchedInput");
}
}
private class SchedulerFutureCallback implements FutureCallback<Void> {
@Override
public void onSuccess(Void result) {
LOG.info("Scheduler thread completed");
}
@Override
public void onFailure(Throwable t) {
if (isShutdown.get()) {
LOG.info("Already shutdown. Ignoring error: " + t);
} else {
LOG.error("Scheduler failed with error: ", t);
inputContext.fatalError(t, "Shuffle Scheduler Failed");
}
}
}
private class FetchFutureCallback implements FutureCallback<FetchResult> {
private final Fetcher fetcher;
public FetchFutureCallback(Fetcher fetcher) {
this.fetcher = fetcher;
}
private void doBookKeepingForFetcherComplete() {
lock.lock();
try {
runningFetchers.remove(fetcher);
wakeLoop.signal();
} finally {
lock.unlock();
}
}
@Override
public void onSuccess(FetchResult result) {
fetcher.shutdown();
if (isShutdown.get()) {
LOG.info("Already shutdown. Ignoring event from fetcher");
} else {
Iterable<InputAttemptIdentifier> pendingInputs = result.getPendingInputs();
if (pendingInputs != null && pendingInputs.iterator().hasNext()) {
InputHost inputHost =
knownSrcHosts.get(InputHost.createIdentifier(result.getHost(), result.getPort()));
assert inputHost != null;
for (InputAttemptIdentifier input : pendingInputs) {
inputHost.addKnownInput(input);
}
pendingHosts.add(inputHost);
}
doBookKeepingForFetcherComplete();
}
}
@Override
public void onFailure(Throwable t) {
// Unsuccessful - the fetcher may not have shutdown correctly. Try shutting it down.
fetcher.shutdown();
if (isShutdown.get()) {
LOG.info("Already shutdown. Ignoring error from fetcher: " + t);
} else {
LOG.error("Fetcher failed with error: ", t);
shuffleError = t;
inputContext.fatalError(t, "Fetch failed");
doBookKeepingForFetcherComplete();
}
}
}
}
/** @author tolgam */
public class Optimizer extends Observable implements Runnable {
/** Population size */
private static final int POPULATION_SIZE = 5;
/** Maximum generation to wait before finding an optima */
private static final int MAXIMUM_GENERATION = 40;
/** Logger */
protected final Logger logger = LoggerFactory.getLogger(Optimizer.class);
/** Population */
protected final SortedSet<Solution> population = new TreeSet<Solution>();
/** Hall of fame to put all the results found */
private final Set<Triple> blackListedTriples = new HashSet<Triple>();
/** Mutation operator used to generate new populations */
private final Generate generateOp;
/** Evaluation operator to evaluate all the candidates */
private final Evaluate evaluateOp;
/** Counter for statistics about the number of evaluations */
private int evaluationsCounter = 0;
private final Request request;
/** Activity control */
private boolean isPaused = false;
private boolean isTerminated = false;
private ReentrantLock pauseLock = new ReentrantLock();
private Condition unpaused = pauseLock.newCondition();
// Generation counter
private int generation = 0;
private DataLayer datalayer;
/**
* Optimizer
*
* @param datalayer
* @param request
* @param executor
*/
public Optimizer(
final DataLayer datalayer, final Request request, final ExecutorService executor) {
// Save a pointer to the request and the datalayer
this.request = request;
this.datalayer = datalayer;
// Create the operators
this.generateOp = new Generate(datalayer, request);
this.evaluateOp = new Evaluate(request, blackListedTriples, executor);
}
/*
* (non-Javadoc)
*
* @see java.lang.Runnable#run()
*/
public void run() {
// Do not run something terminated
if (isTerminated()) return;
logger.info("Run optimizer");
generation = 0;
while (!isTerminated()) {
pauseLock.lock();
try {
while (isPaused) unpaused.await();
if (isTerminated) return;
} catch (InterruptedException ie) {
// Finish
return;
} finally {
pauseLock.unlock();
}
//
// Initialise the population with a dummy individual
//
if (population.isEmpty()) {
Solution solution = new Solution();
for (Node_Variable variable : request.variables())
solution.add(new Binding(variable, Node.NULL));
population.add(solution);
}
// Increment the generation counter
++generation;
//
// Generate a new set of offspring and copy the parents into it
// first
//
// logger.info("Generate");
Set<Solution> newPopulation = new HashSet<Solution>();
newPopulation.addAll(population); // Add the parents
generateOp.createPopulation(population, newPopulation);
//
// Evaluate all of them
//
// logger.info("Evaluate " + newPopulation.size());
// Counts the number of different solutions
evaluationsCounter += newPopulation.size() - population.size();
evaluateOp.evaluatePopulation(newPopulation);
/*
* String buffer = "Fitnesses "; for (Solution s : newPopulation)
* buffer += s.getFitness() + " "; logger.info(buffer);
*/
// Provide feed back to the generation operator
generateOp.updateProviderRewards(newPopulation);
//
// Get rid of the previous population and insert the kids
//
// logger.info("Cut");
population.clear();
population.addAll(newPopulation);
while (population.size() > POPULATION_SIZE) population.remove(population.first());
//
// Track for optimality
//
double topFitness = population.last().getFitness();
for (Solution s : population) {
// Increment age
if (s.getFitness() != topFitness) s.resetAge();
s.incrementAge();
// Check optimality
s.setOptimal(false);
if (s.getAge() >= MAXIMUM_GENERATION && s.getFitness() > 0) s.setOptimal(true);
if (s.getFitness() == 1.0d) s.setOptimal(true);
// If the solution is optimal add its (valid!) triples to the
// black
// list
if (s.isOptimal()) {
synchronized (blackListedTriples) {
blackListedTriples.addAll(request.getTripleSet(s));
}
}
// Print solution
// logger.info(s.toString());
}
logger.info("Generation " + generation + ", best fitness=" + topFitness);
for (Solution s : population) logger.info(s.toString());
//
// Notify observers that a loop has been done
//
setChanged();
notifyObservers(population);
// for (Solution s : population)
// if (s.isOptimal())
// this.terminate();
//
// Wait a bit for the data layer
//
datalayer.waitForLatencyBuffer();
//
// Remove all optimum individuals from the population
//
List<Solution> toRemove = new ArrayList<Solution>();
for (Solution s : population) if (s.isOptimal()) toRemove.add(s);
population.removeAll(toRemove);
}
}
/** Stop the execution of the optimizer */
public void terminate() {
logger.info("Terminate optimizer");
pauseLock.lock();
try {
// Set the status to true
isTerminated = true;
} finally {
pauseLock.unlock();
}
}
/** @return true if the optimizer is stopped */
public boolean isTerminated() {
boolean res;
pauseLock.lock();
try {
res = isTerminated;
} finally {
pauseLock.unlock();
}
return res;
}
/** Pause the algorithm */
public void pause() {
logger.info("Pause optimizer " + this);
pauseLock.lock();
try {
isPaused = true;
} finally {
pauseLock.unlock();
}
}
/** @return true if the search algorithm is paused */
public boolean isPaused() {
boolean res;
pauseLock.lock();
try {
res = isPaused;
} finally {
pauseLock.unlock();
}
return res;
}
/** Continue the execution */
public void resume() {
logger.info("Resume optimizer " + this);
pauseLock.lock();
try {
isPaused = false;
unpaused.signalAll();
} finally {
pauseLock.unlock();
}
}
/** @return the evaluations counter */
public int getEvaluationsCounter() {
return evaluationsCounter;
}
/** @return the generations counter */
public int getGenerationsCounter() {
return generation;
}
}
@SuppressWarnings({"unchecked", "rawtypes"})
public class DefaultSorter extends ExternalSorter implements IndexedSortable {
private static final Log LOG = LogFactory.getLog(DefaultSorter.class);
// TODO NEWTEZ Progress reporting to Tez framework. (making progress vs %complete)
/** The size of each record in the index file for the map-outputs. */
public static final int MAP_OUTPUT_INDEX_RECORD_LENGTH = 24;
private static final int APPROX_HEADER_LENGTH = 150;
// k/v accounting
private final IntBuffer kvmeta; // metadata overlay on backing store
int kvstart; // marks origin of spill metadata
int kvend; // marks end of spill metadata
int kvindex; // marks end of fully serialized records
int equator; // marks origin of meta/serialization
int bufstart; // marks beginning of spill
int bufend; // marks beginning of collectable
int bufmark; // marks end of record
int bufindex; // marks end of collected
int bufvoid; // marks the point where we should stop
// reading at the end of the buffer
private final byte[] kvbuffer; // main output buffer
private final byte[] b0 = new byte[0];
protected static final int VALSTART = 0; // val offset in acct
protected static final int KEYSTART = 1; // key offset in acct
protected static final int PARTITION = 2; // partition offset in acct
protected static final int VALLEN = 3; // length of value
protected static final int NMETA = 4; // num meta ints
protected static final int METASIZE = NMETA * 4; // size in bytes
// spill accounting
final int maxRec;
final int softLimit;
boolean spillInProgress;
int bufferRemaining;
volatile Throwable sortSpillException = null;
int numSpills = 0;
final int minSpillsForCombine;
final ReentrantLock spillLock = new ReentrantLock();
final Condition spillDone = spillLock.newCondition();
final Condition spillReady = spillLock.newCondition();
final BlockingBuffer bb = new BlockingBuffer();
volatile boolean spillThreadRunning = false;
final SpillThread spillThread = new SpillThread();
final ArrayList<TezSpillRecord> indexCacheList = new ArrayList<TezSpillRecord>();
private final int indexCacheMemoryLimit;
private int totalIndexCacheMemory;
public DefaultSorter(
OutputContext outputContext, Configuration conf, int numOutputs, long initialMemoryAvailable)
throws IOException {
super(outputContext, conf, numOutputs, initialMemoryAvailable);
// sanity checks
final float spillper =
this.conf.getFloat(
TezRuntimeConfiguration.TEZ_RUNTIME_SORT_SPILL_PERCENT,
TezRuntimeConfiguration.TEZ_RUNTIME_SORT_SPILL_PERCENT_DEFAULT);
final int sortmb = this.availableMemoryMb;
if (spillper > (float) 1.0 || spillper <= (float) 0.0) {
throw new IOException(
"Invalid \""
+ TezRuntimeConfiguration.TEZ_RUNTIME_SORT_SPILL_PERCENT
+ "\": "
+ spillper);
}
if ((sortmb & 0x7FF) != sortmb) {
throw new IOException(
"Invalid \"" + TezRuntimeConfiguration.TEZ_RUNTIME_IO_SORT_MB + "\": " + sortmb);
}
indexCacheMemoryLimit =
this.conf.getInt(
TezRuntimeConfiguration.TEZ_RUNTIME_INDEX_CACHE_MEMORY_LIMIT_BYTES,
TezRuntimeConfiguration.TEZ_RUNTIME_INDEX_CACHE_MEMORY_LIMIT_BYTES_DEFAULT);
// buffers and accounting
int maxMemUsage = sortmb << 20;
maxMemUsage -= maxMemUsage % METASIZE;
kvbuffer = new byte[maxMemUsage];
bufvoid = kvbuffer.length;
kvmeta = ByteBuffer.wrap(kvbuffer).order(ByteOrder.nativeOrder()).asIntBuffer();
setEquator(0);
bufstart = bufend = bufindex = equator;
kvstart = kvend = kvindex;
maxRec = kvmeta.capacity() / NMETA;
softLimit = (int) (kvbuffer.length * spillper);
bufferRemaining = softLimit;
if (LOG.isInfoEnabled()) {
LOG.info(TezRuntimeConfiguration.TEZ_RUNTIME_IO_SORT_MB + ": " + sortmb);
LOG.info("soft limit at " + softLimit);
LOG.info("bufstart = " + bufstart + "; bufvoid = " + bufvoid);
LOG.info("kvstart = " + kvstart + "; length = " + maxRec);
}
// k/v serialization
valSerializer.open(bb);
keySerializer.open(bb);
spillInProgress = false;
minSpillsForCombine =
this.conf.getInt(TezRuntimeConfiguration.TEZ_RUNTIME_COMBINE_MIN_SPILLS, 3);
spillThread.setDaemon(true);
spillThread.setName(
"SpillThread ["
+ TezUtilsInternal.cleanVertexName(outputContext.getDestinationVertexName() + "]"));
spillLock.lock();
try {
spillThread.start();
while (!spillThreadRunning) {
spillDone.await();
}
} catch (InterruptedException e) {
throw new IOException("Spill thread failed to initialize", e);
} finally {
spillLock.unlock();
}
if (sortSpillException != null) {
throw new IOException("Spill thread failed to initialize", sortSpillException);
}
}
@Override
public void write(Object key, Object value) throws IOException {
collect(key, value, partitioner.getPartition(key, value, partitions));
}
/**
* Serialize the key, value to intermediate storage. When this method returns, kvindex must refer
* to sufficient unused storage to store one METADATA.
*/
synchronized void collect(Object key, Object value, final int partition) throws IOException {
if (key.getClass() != keyClass) {
throw new IOException(
"Type mismatch in key from map: expected "
+ keyClass.getName()
+ ", received "
+ key.getClass().getName());
}
if (value.getClass() != valClass) {
throw new IOException(
"Type mismatch in value from map: expected "
+ valClass.getName()
+ ", received "
+ value.getClass().getName());
}
if (partition < 0 || partition >= partitions) {
throw new IOException(
"Illegal partition for "
+ key
+ " ("
+ partition
+ ")"
+ ", TotalPartitions: "
+ partitions);
}
checkSpillException();
bufferRemaining -= METASIZE;
if (bufferRemaining <= 0) {
// start spill if the thread is not running and the soft limit has been
// reached
spillLock.lock();
try {
do {
if (!spillInProgress) {
final int kvbidx = 4 * kvindex;
final int kvbend = 4 * kvend;
// serialized, unspilled bytes always lie between kvindex and
// bufindex, crossing the equator. Note that any void space
// created by a reset must be included in "used" bytes
final int bUsed = distanceTo(kvbidx, bufindex);
final boolean bufsoftlimit = bUsed >= softLimit;
if ((kvbend + METASIZE) % kvbuffer.length != equator - (equator % METASIZE)) {
// spill finished, reclaim space
resetSpill();
bufferRemaining =
Math.min(distanceTo(bufindex, kvbidx) - 2 * METASIZE, softLimit - bUsed)
- METASIZE;
continue;
} else if (bufsoftlimit && kvindex != kvend) {
// spill records, if any collected; check latter, as it may
// be possible for metadata alignment to hit spill pcnt
startSpill();
final int avgRec =
(int) (mapOutputByteCounter.getValue() / mapOutputRecordCounter.getValue());
// leave at least half the split buffer for serialization data
// ensure that kvindex >= bufindex
final int distkvi = distanceTo(bufindex, kvbidx);
final int newPos =
(bufindex
+ Math.max(
2 * METASIZE - 1,
Math.min(distkvi / 2, distkvi / (METASIZE + avgRec) * METASIZE)))
% kvbuffer.length;
setEquator(newPos);
bufmark = bufindex = newPos;
final int serBound = 4 * kvend;
// bytes remaining before the lock must be held and limits
// checked is the minimum of three arcs: the metadata space, the
// serialization space, and the soft limit
bufferRemaining =
Math.min(
// metadata max
distanceTo(bufend, newPos),
Math.min(
// serialization max
distanceTo(newPos, serBound),
// soft limit
softLimit))
- 2 * METASIZE;
}
}
} while (false);
} finally {
spillLock.unlock();
}
}
try {
// serialize key bytes into buffer
int keystart = bufindex;
keySerializer.serialize(key);
if (bufindex < keystart) {
// wrapped the key; must make contiguous
bb.shiftBufferedKey();
keystart = 0;
}
// serialize value bytes into buffer
final int valstart = bufindex;
valSerializer.serialize(value);
// It's possible for records to have zero length, i.e. the serializer
// will perform no writes. To ensure that the boundary conditions are
// checked and that the kvindex invariant is maintained, perform a
// zero-length write into the buffer. The logic monitoring this could be
// moved into collect, but this is cleaner and inexpensive. For now, it
// is acceptable.
bb.write(b0, 0, 0);
// the record must be marked after the preceding write, as the metadata
// for this record are not yet written
int valend = bb.markRecord();
mapOutputRecordCounter.increment(1);
mapOutputByteCounter.increment(distanceTo(keystart, valend, bufvoid));
// write accounting info
kvmeta.put(kvindex + PARTITION, partition);
kvmeta.put(kvindex + KEYSTART, keystart);
kvmeta.put(kvindex + VALSTART, valstart);
kvmeta.put(kvindex + VALLEN, distanceTo(valstart, valend));
// advance kvindex
kvindex = (int) (((long) kvindex - NMETA + kvmeta.capacity()) % kvmeta.capacity());
} catch (MapBufferTooSmallException e) {
LOG.info("Record too large for in-memory buffer: " + e.getMessage());
spillSingleRecord(key, value, partition);
mapOutputRecordCounter.increment(1);
return;
}
}
/**
* Set the point from which meta and serialization data expand. The meta indices are aligned with
* the buffer, so metadata never spans the ends of the circular buffer.
*/
private void setEquator(int pos) {
equator = pos;
// set index prior to first entry, aligned at meta boundary
final int aligned = pos - (pos % METASIZE);
// Cast one of the operands to long to avoid integer overflow
kvindex = (int) (((long) aligned - METASIZE + kvbuffer.length) % kvbuffer.length) / 4;
if (LOG.isInfoEnabled()) {
LOG.info("(EQUATOR) " + pos + " kvi " + kvindex + "(" + (kvindex * 4) + ")");
}
}
/**
* The spill is complete, so set the buffer and meta indices to be equal to the new equator to
* free space for continuing collection. Note that when kvindex == kvend == kvstart, the buffer is
* empty.
*/
private void resetSpill() {
final int e = equator;
bufstart = bufend = e;
final int aligned = e - (e % METASIZE);
// set start/end to point to first meta record
// Cast one of the operands to long to avoid integer overflow
kvstart = kvend = (int) (((long) aligned - METASIZE + kvbuffer.length) % kvbuffer.length) / 4;
if (LOG.isInfoEnabled()) {
LOG.info(
"(RESET) equator "
+ e
+ " kv "
+ kvstart
+ "("
+ (kvstart * 4)
+ ")"
+ " kvi "
+ kvindex
+ "("
+ (kvindex * 4)
+ ")");
}
}
/**
* Compute the distance in bytes between two indices in the serialization buffer.
*
* @see #distanceTo(int,int,int)
*/
final int distanceTo(final int i, final int j) {
return distanceTo(i, j, kvbuffer.length);
}
/** Compute the distance between two indices in the circular buffer given the max distance. */
int distanceTo(final int i, final int j, final int mod) {
return i <= j ? j - i : mod - i + j;
}
/** For the given meta position, return the offset into the int-sized kvmeta buffer. */
int offsetFor(int metapos) {
return (metapos % maxRec) * NMETA;
}
/**
* Compare logical range, st i, j MOD offset capacity. Compare by partition, then by key.
*
* @see IndexedSortable#compare
*/
public int compare(final int mi, final int mj) {
final int kvi = offsetFor(mi);
final int kvj = offsetFor(mj);
final int kvip = kvmeta.get(kvi + PARTITION);
final int kvjp = kvmeta.get(kvj + PARTITION);
// sort by partition
if (kvip != kvjp) {
return kvip - kvjp;
}
// sort by key
return comparator.compare(
kvbuffer,
kvmeta.get(kvi + KEYSTART),
kvmeta.get(kvi + VALSTART) - kvmeta.get(kvi + KEYSTART),
kvbuffer,
kvmeta.get(kvj + KEYSTART),
kvmeta.get(kvj + VALSTART) - kvmeta.get(kvj + KEYSTART));
}
final byte META_BUFFER_TMP[] = new byte[METASIZE];
/**
* Swap metadata for items i,j
*
* @see IndexedSortable#swap
*/
public void swap(final int mi, final int mj) {
int iOff = (mi % maxRec) * METASIZE;
int jOff = (mj % maxRec) * METASIZE;
System.arraycopy(kvbuffer, iOff, META_BUFFER_TMP, 0, METASIZE);
System.arraycopy(kvbuffer, jOff, kvbuffer, iOff, METASIZE);
System.arraycopy(META_BUFFER_TMP, 0, kvbuffer, jOff, METASIZE);
}
/** Inner class managing the spill of serialized records to disk. */
protected class BlockingBuffer extends DataOutputStream {
public BlockingBuffer() {
super(new Buffer());
}
/**
* Mark end of record. Note that this is required if the buffer is to cut the spill in the
* proper place.
*/
public int markRecord() {
bufmark = bufindex;
return bufindex;
}
/**
* Set position from last mark to end of writable buffer, then rewrite the data between last
* mark and kvindex. This handles a special case where the key wraps around the buffer. If the
* key is to be passed to a RawComparator, then it must be contiguous in the buffer. This
* recopies the data in the buffer back into itself, but starting at the beginning of the
* buffer. Note that this method should <b>only</b> be called immediately after detecting this
* condition. To call it at any other time is undefined and would likely result in data loss or
* corruption.
*
* @see #markRecord()
*/
protected void shiftBufferedKey() throws IOException {
// spillLock unnecessary; both kvend and kvindex are current
int headbytelen = bufvoid - bufmark;
bufvoid = bufmark;
final int kvbidx = 4 * kvindex;
final int kvbend = 4 * kvend;
final int avail = Math.min(distanceTo(0, kvbidx), distanceTo(0, kvbend));
if (bufindex + headbytelen < avail) {
System.arraycopy(kvbuffer, 0, kvbuffer, headbytelen, bufindex);
System.arraycopy(kvbuffer, bufvoid, kvbuffer, 0, headbytelen);
bufindex += headbytelen;
bufferRemaining -= kvbuffer.length - bufvoid;
} else {
byte[] keytmp = new byte[bufindex];
System.arraycopy(kvbuffer, 0, keytmp, 0, bufindex);
bufindex = 0;
out.write(kvbuffer, bufmark, headbytelen);
out.write(keytmp);
}
}
}
public class Buffer extends OutputStream {
private final byte[] scratch = new byte[1];
@Override
public void write(int v) throws IOException {
scratch[0] = (byte) v;
write(scratch, 0, 1);
}
/**
* Attempt to write a sequence of bytes to the collection buffer. This method will block if the
* spill thread is running and it cannot write.
*
* @throws MapBufferTooSmallException if record is too large to deserialize into the collection
* buffer.
*/
@Override
public void write(byte b[], int off, int len) throws IOException {
// must always verify the invariant that at least METASIZE bytes are
// available beyond kvindex, even when len == 0
bufferRemaining -= len;
if (bufferRemaining <= 0) {
// writing these bytes could exhaust available buffer space or fill
// the buffer to soft limit. check if spill or blocking are necessary
boolean blockwrite = false;
spillLock.lock();
try {
do {
checkSpillException();
final int kvbidx = 4 * kvindex;
final int kvbend = 4 * kvend;
// ser distance to key index
final int distkvi = distanceTo(bufindex, kvbidx);
// ser distance to spill end index
final int distkve = distanceTo(bufindex, kvbend);
// if kvindex is closer than kvend, then a spill is neither in
// progress nor complete and reset since the lock was held. The
// write should block only if there is insufficient space to
// complete the current write, write the metadata for this record,
// and write the metadata for the next record. If kvend is closer,
// then the write should block if there is too little space for
// either the metadata or the current write. Note that collect
// ensures its metadata requirement with a zero-length write
blockwrite =
distkvi <= distkve
? distkvi <= len + 2 * METASIZE
: distkve <= len || distanceTo(bufend, kvbidx) < 2 * METASIZE;
if (!spillInProgress) {
if (blockwrite) {
if ((kvbend + METASIZE) % kvbuffer.length != equator - (equator % METASIZE)) {
// spill finished, reclaim space
// need to use meta exclusively; zero-len rec & 100% spill
// pcnt would fail
resetSpill(); // resetSpill doesn't move bufindex, kvindex
bufferRemaining =
Math.min(distkvi - 2 * METASIZE, softLimit - distanceTo(kvbidx, bufindex))
- len;
continue;
}
// we have records we can spill; only spill if blocked
if (kvindex != kvend) {
startSpill();
// Blocked on this write, waiting for the spill just
// initiated to finish. Instead of repositioning the marker
// and copying the partial record, we set the record start
// to be the new equator
setEquator(bufmark);
} else {
// We have no buffered records, and this record is too large
// to write into kvbuffer. We must spill it directly from
// collect
final int size = distanceTo(bufstart, bufindex) + len;
setEquator(0);
bufstart = bufend = bufindex = equator;
kvstart = kvend = kvindex;
bufvoid = kvbuffer.length;
throw new MapBufferTooSmallException(size + " bytes");
}
}
}
if (blockwrite) {
// wait for spill
try {
while (spillInProgress) {
spillDone.await();
}
} catch (InterruptedException e) {
throw new IOException("Buffer interrupted while waiting for the writer", e);
}
}
} while (blockwrite);
} finally {
spillLock.unlock();
}
}
// here, we know that we have sufficient space to write
if (bufindex + len > bufvoid) {
final int gaplen = bufvoid - bufindex;
System.arraycopy(b, off, kvbuffer, bufindex, gaplen);
len -= gaplen;
off += gaplen;
bufindex = 0;
}
System.arraycopy(b, off, kvbuffer, bufindex, len);
bufindex += len;
}
}
@Override
public void flush() throws IOException {
LOG.info("Starting flush of map output");
spillLock.lock();
try {
while (spillInProgress) {
spillDone.await();
}
checkSpillException();
final int kvbend = 4 * kvend;
if ((kvbend + METASIZE) % kvbuffer.length != equator - (equator % METASIZE)) {
// spill finished
resetSpill();
}
if (kvindex != kvend) {
kvend = (kvindex + NMETA) % kvmeta.capacity();
bufend = bufmark;
if (LOG.isInfoEnabled()) {
LOG.info("Sorting & Spilling map output");
LOG.info("bufstart = " + bufstart + "; bufend = " + bufmark + "; bufvoid = " + bufvoid);
LOG.info(
"kvstart = "
+ kvstart
+ "("
+ (kvstart * 4)
+ "); kvend = "
+ kvend
+ "("
+ (kvend * 4)
+ "); length = "
+ (distanceTo(kvend, kvstart, kvmeta.capacity()) + 1)
+ "/"
+ maxRec);
}
sortAndSpill();
}
} catch (InterruptedException e) {
throw new IOException("Interrupted while waiting for the writer", e);
} finally {
spillLock.unlock();
}
assert !spillLock.isHeldByCurrentThread();
// shut down spill thread and wait for it to exit. Since the preceding
// ensures that it is finished with its work (and sortAndSpill did not
// throw), we elect to use an interrupt instead of setting a flag.
// Spilling simultaneously from this thread while the spill thread
// finishes its work might be both a useful way to extend this and also
// sufficient motivation for the latter approach.
try {
spillThread.interrupt();
spillThread.join();
} catch (InterruptedException e) {
throw new IOException("Spill failed", e);
}
// release sort buffer before the merge
// FIXME
// kvbuffer = null;
mergeParts();
Path outputPath = mapOutputFile.getOutputFile();
fileOutputByteCounter.increment(rfs.getFileStatus(outputPath).getLen());
}
@Override
public void close() throws IOException {}
protected class SpillThread extends Thread {
@Override
public void run() {
spillLock.lock();
spillThreadRunning = true;
try {
while (true) {
spillDone.signal();
while (!spillInProgress) {
spillReady.await();
}
try {
spillLock.unlock();
sortAndSpill();
} catch (Throwable t) {
LOG.warn("Got an exception in sortAndSpill", t);
sortSpillException = t;
} finally {
spillLock.lock();
if (bufend < bufstart) {
bufvoid = kvbuffer.length;
}
kvstart = kvend;
bufstart = bufend;
spillInProgress = false;
}
}
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
} finally {
spillLock.unlock();
spillThreadRunning = false;
}
}
}
private void checkSpillException() throws IOException {
final Throwable lspillException = sortSpillException;
if (lspillException != null) {
if (lspillException instanceof Error) {
final String logMsg =
"Task "
+ outputContext.getUniqueIdentifier()
+ " failed : "
+ ExceptionUtils.getStackTrace(lspillException);
outputContext.fatalError(lspillException, logMsg);
}
throw new IOException("Spill failed", lspillException);
}
}
private void startSpill() {
assert !spillInProgress;
kvend = (kvindex + NMETA) % kvmeta.capacity();
bufend = bufmark;
spillInProgress = true;
if (LOG.isInfoEnabled()) {
LOG.info("Spilling map output");
LOG.info("bufstart = " + bufstart + "; bufend = " + bufmark + "; bufvoid = " + bufvoid);
LOG.info(
"kvstart = "
+ kvstart
+ "("
+ (kvstart * 4)
+ "); kvend = "
+ kvend
+ "("
+ (kvend * 4)
+ "); length = "
+ (distanceTo(kvend, kvstart, kvmeta.capacity()) + 1)
+ "/"
+ maxRec);
}
spillReady.signal();
}
int getMetaStart() {
return kvend / NMETA;
}
int getMetaEnd() {
return 1
+ // kvend is a valid record
(kvstart >= kvend ? kvstart : kvmeta.capacity() + kvstart) / NMETA;
}
protected void sortAndSpill() throws IOException, InterruptedException {
final int mstart = getMetaStart();
final int mend = getMetaEnd();
sorter.sort(this, mstart, mend, nullProgressable);
spill(mstart, mend);
}
protected void spill(int mstart, int mend) throws IOException, InterruptedException {
// approximate the length of the output file to be the length of the
// buffer + header lengths for the partitions
final long size =
(bufend >= bufstart ? bufend - bufstart : (bufvoid - bufend) + bufstart)
+ partitions * APPROX_HEADER_LENGTH;
FSDataOutputStream out = null;
try {
// create spill file
final TezSpillRecord spillRec = new TezSpillRecord(partitions);
final Path filename = mapOutputFile.getSpillFileForWrite(numSpills, size);
out = rfs.create(filename);
int spindex = mstart;
final InMemValBytes value = createInMemValBytes();
for (int i = 0; i < partitions; ++i) {
IFile.Writer writer = null;
try {
long segmentStart = out.getPos();
writer = new Writer(conf, out, keyClass, valClass, codec, spilledRecordsCounter, null);
if (combiner == null) {
// spill directly
DataInputBuffer key = new DataInputBuffer();
while (spindex < mend && kvmeta.get(offsetFor(spindex) + PARTITION) == i) {
final int kvoff = offsetFor(spindex);
int keystart = kvmeta.get(kvoff + KEYSTART);
int valstart = kvmeta.get(kvoff + VALSTART);
key.reset(kvbuffer, keystart, valstart - keystart);
getVBytesForOffset(kvoff, value);
writer.append(key, value);
++spindex;
}
} else {
int spstart = spindex;
while (spindex < mend && kvmeta.get(offsetFor(spindex) + PARTITION) == i) {
++spindex;
}
// Note: we would like to avoid the combiner if we've fewer
// than some threshold of records for a partition
if (spstart != spindex) {
TezRawKeyValueIterator kvIter = new MRResultIterator(spstart, spindex);
if (LOG.isDebugEnabled()) {
LOG.debug("Running combine processor");
}
runCombineProcessor(kvIter, writer);
}
}
// close the writer
writer.close();
if (numSpills > 0) {
additionalSpillBytesWritten.increment(writer.getCompressedLength());
numAdditionalSpills.increment(1);
// Reset the value will be set during the final merge.
outputBytesWithOverheadCounter.setValue(0);
} else {
// Set this up for the first write only. Subsequent ones will be handled in the final
// merge.
outputBytesWithOverheadCounter.increment(writer.getRawLength());
}
// record offsets
final TezIndexRecord rec =
new TezIndexRecord(segmentStart, writer.getRawLength(), writer.getCompressedLength());
spillRec.putIndex(rec, i);
writer = null;
} finally {
if (null != writer) writer.close();
}
}
if (totalIndexCacheMemory >= indexCacheMemoryLimit) {
// create spill index file
Path indexFilename =
mapOutputFile.getSpillIndexFileForWrite(
numSpills, partitions * MAP_OUTPUT_INDEX_RECORD_LENGTH);
spillRec.writeToFile(indexFilename, conf);
} else {
indexCacheList.add(spillRec);
totalIndexCacheMemory += spillRec.size() * MAP_OUTPUT_INDEX_RECORD_LENGTH;
}
LOG.info("Finished spill " + numSpills);
++numSpills;
} finally {
if (out != null) out.close();
}
}
/**
* Handles the degenerate case where serialization fails to fit in the in-memory buffer, so we
* must spill the record from collect directly to a spill file. Consider this "losing".
*/
private void spillSingleRecord(final Object key, final Object value, int partition)
throws IOException {
long size = kvbuffer.length + partitions * APPROX_HEADER_LENGTH;
FSDataOutputStream out = null;
try {
// create spill file
final TezSpillRecord spillRec = new TezSpillRecord(partitions);
final Path filename = mapOutputFile.getSpillFileForWrite(numSpills, size);
out = rfs.create(filename);
// we don't run the combiner for a single record
for (int i = 0; i < partitions; ++i) {
IFile.Writer writer = null;
try {
long segmentStart = out.getPos();
// Create a new codec, don't care!
writer =
new IFile.Writer(conf, out, keyClass, valClass, codec, spilledRecordsCounter, null);
if (i == partition) {
final long recordStart = out.getPos();
writer.append(key, value);
// Note that our map byte count will not be accurate with
// compression
mapOutputByteCounter.increment(out.getPos() - recordStart);
}
writer.close();
if (numSpills > 0) {
additionalSpillBytesWritten.increment(writer.getCompressedLength());
numAdditionalSpills.increment(1);
outputBytesWithOverheadCounter.setValue(0);
} else {
// Set this up for the first write only. Subsequent ones will be handled in the final
// merge.
outputBytesWithOverheadCounter.increment(writer.getRawLength());
}
// record offsets
TezIndexRecord rec =
new TezIndexRecord(segmentStart, writer.getRawLength(), writer.getCompressedLength());
spillRec.putIndex(rec, i);
writer = null;
} catch (IOException e) {
if (null != writer) writer.close();
throw e;
}
}
if (totalIndexCacheMemory >= indexCacheMemoryLimit) {
// create spill index file
Path indexFilename =
mapOutputFile.getSpillIndexFileForWrite(
numSpills, partitions * MAP_OUTPUT_INDEX_RECORD_LENGTH);
spillRec.writeToFile(indexFilename, conf);
} else {
indexCacheList.add(spillRec);
totalIndexCacheMemory += spillRec.size() * MAP_OUTPUT_INDEX_RECORD_LENGTH;
}
++numSpills;
} finally {
if (out != null) out.close();
}
}
protected int getInMemVBytesLength(int kvoff) {
// get the keystart for the next serialized value to be the end
// of this value. If this is the last value in the buffer, use bufend
final int vallen = kvmeta.get(kvoff + VALLEN);
assert vallen >= 0;
return vallen;
}
/**
* Given an offset, populate vbytes with the associated set of deserialized value bytes. Should
* only be called during a spill.
*/
int getVBytesForOffset(int kvoff, InMemValBytes vbytes) {
int vallen = getInMemVBytesLength(kvoff);
vbytes.reset(kvbuffer, kvmeta.get(kvoff + VALSTART), vallen);
return vallen;
}
/** Inner class wrapping valuebytes, used for appendRaw. */
static class InMemValBytes extends DataInputBuffer {
private byte[] buffer;
private int start;
private int length;
private final int bufvoid;
public InMemValBytes(int bufvoid) {
this.bufvoid = bufvoid;
}
public void reset(byte[] buffer, int start, int length) {
this.buffer = buffer;
this.start = start;
this.length = length;
if (start + length > bufvoid) {
this.buffer = new byte[this.length];
final int taillen = bufvoid - start;
System.arraycopy(buffer, start, this.buffer, 0, taillen);
System.arraycopy(buffer, 0, this.buffer, taillen, length - taillen);
this.start = 0;
}
super.reset(this.buffer, this.start, this.length);
}
}
InMemValBytes createInMemValBytes() {
return new InMemValBytes(bufvoid);
}
protected class MRResultIterator implements TezRawKeyValueIterator {
private final DataInputBuffer keybuf = new DataInputBuffer();
private final InMemValBytes vbytes = createInMemValBytes();
private final int end;
private int current;
public MRResultIterator(int start, int end) {
this.end = end;
current = start - 1;
}
public boolean next() throws IOException {
return ++current < end;
}
public DataInputBuffer getKey() throws IOException {
final int kvoff = offsetFor(current);
keybuf.reset(
kvbuffer,
kvmeta.get(kvoff + KEYSTART),
kvmeta.get(kvoff + VALSTART) - kvmeta.get(kvoff + KEYSTART));
return keybuf;
}
public DataInputBuffer getValue() throws IOException {
getVBytesForOffset(offsetFor(current), vbytes);
return vbytes;
}
public Progress getProgress() {
return null;
}
public void close() {}
}
private void mergeParts() throws IOException {
// get the approximate size of the final output/index files
long finalOutFileSize = 0;
long finalIndexFileSize = 0;
final Path[] filename = new Path[numSpills];
final String taskIdentifier = outputContext.getUniqueIdentifier();
for (int i = 0; i < numSpills; i++) {
filename[i] = mapOutputFile.getSpillFile(i);
finalOutFileSize += rfs.getFileStatus(filename[i]).getLen();
}
if (numSpills == 1) { // the spill is the final output
sameVolRename(filename[0], mapOutputFile.getOutputFileForWriteInVolume(filename[0]));
if (indexCacheList.size() == 0) {
sameVolRename(
mapOutputFile.getSpillIndexFile(0),
mapOutputFile.getOutputIndexFileForWriteInVolume(filename[0]));
} else {
indexCacheList
.get(0)
.writeToFile(mapOutputFile.getOutputIndexFileForWriteInVolume(filename[0]), conf);
}
return;
}
// read in paged indices
for (int i = indexCacheList.size(); i < numSpills; ++i) {
Path indexFileName = mapOutputFile.getSpillIndexFile(i);
indexCacheList.add(new TezSpillRecord(indexFileName, conf));
}
// make correction in the length to include the sequence file header
// lengths for each partition
finalOutFileSize += partitions * APPROX_HEADER_LENGTH;
finalIndexFileSize = partitions * MAP_OUTPUT_INDEX_RECORD_LENGTH;
Path finalOutputFile = mapOutputFile.getOutputFileForWrite(finalOutFileSize);
Path finalIndexFile = mapOutputFile.getOutputIndexFileForWrite(finalIndexFileSize);
// The output stream for the final single output file
FSDataOutputStream finalOut = rfs.create(finalOutputFile, true, 4096);
if (numSpills == 0) {
// TODO Change event generation to say there is no data rather than generating a dummy file
// create dummy files
TezSpillRecord sr = new TezSpillRecord(partitions);
try {
for (int i = 0; i < partitions; i++) {
long segmentStart = finalOut.getPos();
Writer writer = new Writer(conf, finalOut, keyClass, valClass, codec, null, null);
writer.close();
TezIndexRecord rec =
new TezIndexRecord(segmentStart, writer.getRawLength(), writer.getCompressedLength());
// Covers the case of multiple spills.
outputBytesWithOverheadCounter.increment(writer.getRawLength());
sr.putIndex(rec, i);
}
sr.writeToFile(finalIndexFile, conf);
} finally {
finalOut.close();
}
return;
} else {
final TezSpillRecord spillRec = new TezSpillRecord(partitions);
for (int parts = 0; parts < partitions; parts++) {
// create the segments to be merged
List<Segment> segmentList = new ArrayList<Segment>(numSpills);
for (int i = 0; i < numSpills; i++) {
TezIndexRecord indexRecord = indexCacheList.get(i).getIndex(parts);
Segment s =
new Segment(
conf,
rfs,
filename[i],
indexRecord.getStartOffset(),
indexRecord.getPartLength(),
codec,
ifileReadAhead,
ifileReadAheadLength,
ifileBufferSize,
true);
segmentList.add(i, s);
if (LOG.isDebugEnabled()) {
LOG.debug(
"TaskIdentifier="
+ taskIdentifier
+ " Partition="
+ parts
+ "Spill ="
+ i
+ "("
+ indexRecord.getStartOffset()
+ ","
+ indexRecord.getRawLength()
+ ", "
+ indexRecord.getPartLength()
+ ")");
}
}
int mergeFactor =
this.conf.getInt(
TezRuntimeConfiguration.TEZ_RUNTIME_IO_SORT_FACTOR,
TezRuntimeConfiguration.TEZ_RUNTIME_IO_SORT_FACTOR_DEFAULT);
// sort the segments only if there are intermediate merges
boolean sortSegments = segmentList.size() > mergeFactor;
// merge
TezRawKeyValueIterator kvIter =
TezMerger.merge(
conf,
rfs,
keyClass,
valClass,
codec,
segmentList,
mergeFactor,
new Path(taskIdentifier),
(RawComparator) ConfigUtils.getIntermediateOutputKeyComparator(conf),
nullProgressable,
sortSegments,
true,
null,
spilledRecordsCounter,
additionalSpillBytesRead,
null); // Not using any Progress in TezMerger. Should just work.
// write merged output to disk
long segmentStart = finalOut.getPos();
Writer writer =
new Writer(conf, finalOut, keyClass, valClass, codec, spilledRecordsCounter, null);
if (combiner == null || numSpills < minSpillsForCombine) {
TezMerger.writeFile(
kvIter,
writer,
nullProgressable,
TezRuntimeConfiguration.TEZ_RUNTIME_RECORDS_BEFORE_PROGRESS_DEFAULT);
} else {
runCombineProcessor(kvIter, writer);
}
writer.close();
// record offsets
final TezIndexRecord rec =
new TezIndexRecord(segmentStart, writer.getRawLength(), writer.getCompressedLength());
spillRec.putIndex(rec, parts);
}
spillRec.writeToFile(finalIndexFile, conf);
finalOut.close();
for (int i = 0; i < numSpills; i++) {
rfs.delete(filename[i], true);
}
}
}
}
// todo make thread safe and concurrent
public class DbImpl implements DB {
private final Options options;
private final File databaseDir;
private final TableCache tableCache;
private final DbLock dbLock;
private final VersionSet versions;
private final AtomicBoolean shuttingDown = new AtomicBoolean();
private final ReentrantLock mutex = new ReentrantLock();
private final Condition backgroundCondition = mutex.newCondition();
private final List<Long> pendingOutputs = newArrayList(); // todo
private LogWriter log;
private MemTable memTable;
private MemTable immutableMemTable;
private final InternalKeyComparator internalKeyComparator;
private volatile Throwable backgroundException;
private ExecutorService compactionExecutor;
private Future<?> backgroundCompaction;
private ManualCompaction manualCompaction;
public DbImpl(Options options, File databaseDir) throws IOException {
Preconditions.checkNotNull(options, "options is null");
Preconditions.checkNotNull(databaseDir, "databaseDir is null");
this.options = options;
if (this.options.compressionType() == CompressionType.ZLIB && !Zlib.available()) {
// There's little hope to continue.
this.options.compressionType(CompressionType.NONE);
}
if (this.options.compressionType() == CompressionType.SNAPPY && !Snappy.available()) {
// Disable snappy if it's not available.
this.options.compressionType(CompressionType.NONE);
}
this.databaseDir = databaseDir;
// use custom comparator if set
DBComparator comparator = options.comparator();
UserComparator userComparator;
if (comparator != null) {
userComparator = new CustomUserComparator(comparator);
} else {
userComparator = new BytewiseComparator();
}
internalKeyComparator = new InternalKeyComparator(userComparator);
memTable = new MemTable(internalKeyComparator);
immutableMemTable = null;
ThreadFactory compactionThreadFactory =
new ThreadFactoryBuilder()
.setNameFormat("leveldb-compaction-%s")
.setUncaughtExceptionHandler(
new UncaughtExceptionHandler() {
@Override
public void uncaughtException(Thread t, Throwable e) {
// todo need a real UncaughtExceptionHandler
System.out.printf("%s%n", t);
e.printStackTrace();
}
})
.build();
compactionExecutor = Executors.newSingleThreadExecutor(compactionThreadFactory);
// Reserve ten files or so for other uses and give the rest to TableCache.
int tableCacheSize = options.maxOpenFiles() - 10;
tableCache =
new TableCache(
databaseDir,
tableCacheSize,
new InternalUserComparator(internalKeyComparator),
options.verifyChecksums());
// create the version set
// create the database dir if it does not already exist
databaseDir.mkdirs();
Preconditions.checkArgument(
databaseDir.exists(),
"Database directory '%s' does not exist and could not be created",
databaseDir);
Preconditions.checkArgument(
databaseDir.isDirectory(), "Database directory '%s' is not a directory", databaseDir);
mutex.lock();
try {
// lock the database dir
dbLock = new DbLock(new File(databaseDir, Filename.lockFileName()));
// verify the "current" file
File currentFile = new File(databaseDir, Filename.currentFileName());
if (!currentFile.canRead()) {
Preconditions.checkArgument(
options.createIfMissing(),
"Database '%s' does not exist and the create if missing option is disabled",
databaseDir);
} else {
Preconditions.checkArgument(
!options.errorIfExists(),
"Database '%s' exists and the error if exists option is enabled",
databaseDir);
}
versions = new VersionSet(databaseDir, tableCache, internalKeyComparator);
// load (and recover) current version
versions.recover();
// Recover from all newer log files than the ones named in the
// descriptor (new log files may have been added by the previous
// incarnation without registering them in the descriptor).
//
// Note that PrevLogNumber() is no longer used, but we pay
// attention to it in case we are recovering a database
// produced by an older version of leveldb.
long minLogNumber = versions.getLogNumber();
long previousLogNumber = versions.getPrevLogNumber();
List<File> filenames = Filename.listFiles(databaseDir);
List<Long> logs = Lists.newArrayList();
for (File filename : filenames) {
FileInfo fileInfo = Filename.parseFileName(filename);
if (fileInfo != null
&& fileInfo.getFileType() == FileType.LOG
&& ((fileInfo.getFileNumber() >= minLogNumber)
|| (fileInfo.getFileNumber() == previousLogNumber))) {
logs.add(fileInfo.getFileNumber());
}
}
// Recover in the order in which the logs were generated
VersionEdit edit = new VersionEdit();
Collections.sort(logs);
for (Long fileNumber : logs) {
long maxSequence = recoverLogFile(fileNumber, edit);
if (versions.getLastSequence() < maxSequence) {
versions.setLastSequence(maxSequence);
}
}
// open transaction log
long logFileNumber = versions.getNextFileNumber();
this.log =
Logs.createLogWriter(
new File(databaseDir, Filename.logFileName(logFileNumber)), logFileNumber);
edit.setLogNumber(log.getFileNumber());
// apply recovered edits
versions.logAndApply(edit);
// cleanup unused files
deleteObsoleteFiles();
// schedule compactions
maybeScheduleCompaction();
} finally {
mutex.unlock();
}
}
public void close() {
if (shuttingDown.getAndSet(true)) {
return;
}
mutex.lock();
try {
while (backgroundCompaction != null) {
backgroundCondition.awaitUninterruptibly();
}
} finally {
mutex.unlock();
}
compactionExecutor.shutdown();
try {
compactionExecutor.awaitTermination(1, TimeUnit.DAYS);
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
}
try {
versions.destroy();
} catch (IOException ignored) {
}
try {
log.close();
} catch (IOException ignored) {
}
tableCache.close();
dbLock.release();
}
@Override
public String getProperty(String name) {
checkBackgroundException();
return null;
}
private void deleteObsoleteFiles() {
Preconditions.checkState(mutex.isHeldByCurrentThread());
// Make a set of all of the live files
List<Long> live = newArrayList(this.pendingOutputs);
for (FileMetaData fileMetaData : versions.getLiveFiles()) {
live.add(fileMetaData.getNumber());
}
for (File file : Filename.listFiles(databaseDir)) {
FileInfo fileInfo = Filename.parseFileName(file);
if (fileInfo == null) continue;
long number = fileInfo.getFileNumber();
boolean keep = true;
switch (fileInfo.getFileType()) {
case LOG:
keep = ((number >= versions.getLogNumber()) || (number == versions.getPrevLogNumber()));
break;
case DESCRIPTOR:
// Keep my manifest file, and any newer incarnations'
// (in case there is a race that allows other incarnations)
keep = (number >= versions.getManifestFileNumber());
break;
case TABLE:
keep = live.contains(number);
break;
case TEMP:
// Any temp files that are currently being written to must
// be recorded in pending_outputs_, which is inserted into "live"
keep = live.contains(number);
break;
case CURRENT:
case DB_LOCK:
case INFO_LOG:
keep = true;
break;
}
if (!keep) {
if (fileInfo.getFileType() == FileType.TABLE) {
tableCache.evict(number);
}
// todo info logging system needed
// Log(options_.info_log, "Delete type=%d #%lld\n",
// int(type),
// static_cast < unsigned long long>(number));
file.delete();
}
}
}
public void flushMemTable() {
mutex.lock();
try {
// force compaction
makeRoomForWrite(true);
// todo bg_error code
while (immutableMemTable != null) {
backgroundCondition.awaitUninterruptibly();
}
} finally {
mutex.unlock();
}
}
public void compactRange(int level, Slice start, Slice end) {
Preconditions.checkArgument(level >= 0, "level is negative");
Preconditions.checkArgument(
level + 1 < NUM_LEVELS, "level is greater than or equal to %s", NUM_LEVELS);
Preconditions.checkNotNull(start, "start is null");
Preconditions.checkNotNull(end, "end is null");
mutex.lock();
try {
while (this.manualCompaction != null) {
backgroundCondition.awaitUninterruptibly();
}
ManualCompaction manualCompaction = new ManualCompaction(level, start, end);
this.manualCompaction = manualCompaction;
maybeScheduleCompaction();
while (this.manualCompaction == manualCompaction) {
backgroundCondition.awaitUninterruptibly();
}
} finally {
mutex.unlock();
}
}
private void maybeScheduleCompaction() {
Preconditions.checkState(mutex.isHeldByCurrentThread());
if (backgroundCompaction != null) {
// Already scheduled
} else if (shuttingDown.get()) {
// DB is being shutdown; no more background compactions
} else if (immutableMemTable == null
&& manualCompaction == null
&& !versions.needsCompaction()) {
// No work to be done
} else {
backgroundCompaction =
compactionExecutor.submit(
new Callable<Void>() {
@Override
public Void call() throws Exception {
try {
backgroundCall();
} catch (DatabaseShutdownException ignored) {
} catch (Throwable e) {
backgroundException = e;
}
return null;
}
});
}
}
public void checkBackgroundException() {
Throwable e = backgroundException;
if (e != null) {
throw new BackgroundProcessingException(e);
}
}
private void backgroundCall() throws IOException {
mutex.lock();
try {
if (backgroundCompaction == null) {
return;
}
try {
if (!shuttingDown.get()) {
backgroundCompaction();
}
} finally {
backgroundCompaction = null;
}
} finally {
try {
// Previous compaction may have produced too many files in a level,
// so reschedule another compaction if needed.
maybeScheduleCompaction();
} finally {
try {
backgroundCondition.signalAll();
} finally {
mutex.unlock();
}
}
}
}
private void backgroundCompaction() throws IOException {
Preconditions.checkState(mutex.isHeldByCurrentThread());
compactMemTableInternal();
Compaction compaction;
if (manualCompaction != null) {
compaction =
versions.compactRange(
manualCompaction.level,
new InternalKey(manualCompaction.begin, MAX_SEQUENCE_NUMBER, ValueType.VALUE),
new InternalKey(manualCompaction.end, 0, ValueType.DELETION));
} else {
compaction = versions.pickCompaction();
}
if (compaction == null) {
// no compaction
} else if (manualCompaction == null && compaction.isTrivialMove()) {
// Move file to next level
Preconditions.checkState(compaction.getLevelInputs().size() == 1);
FileMetaData fileMetaData = compaction.getLevelInputs().get(0);
compaction.getEdit().deleteFile(compaction.getLevel(), fileMetaData.getNumber());
compaction.getEdit().addFile(compaction.getLevel() + 1, fileMetaData);
versions.logAndApply(compaction.getEdit());
// log
} else {
CompactionState compactionState = new CompactionState(compaction);
doCompactionWork(compactionState);
cleanupCompaction(compactionState);
}
// manual compaction complete
if (manualCompaction != null) {
manualCompaction = null;
}
}
private void cleanupCompaction(CompactionState compactionState) {
Preconditions.checkState(mutex.isHeldByCurrentThread());
if (compactionState.builder != null) {
compactionState.builder.abandon();
} else {
Preconditions.checkArgument(compactionState.outfile == null);
}
for (FileMetaData output : compactionState.outputs) {
pendingOutputs.remove(output.getNumber());
}
}
private long recoverLogFile(long fileNumber, VersionEdit edit) throws IOException {
Preconditions.checkState(mutex.isHeldByCurrentThread());
File file = new File(databaseDir, Filename.logFileName(fileNumber));
FileChannel channel = new FileInputStream(file).getChannel();
try {
LogMonitor logMonitor = LogMonitors.logMonitor();
LogReader logReader = new LogReader(channel, logMonitor, true, 0);
// Log(options_.info_log, "Recovering log #%llu", (unsigned long long) log_number);
// Read all the records and add to a memtable
long maxSequence = 0;
MemTable memTable = null;
for (Slice record = logReader.readRecord(); record != null; record = logReader.readRecord()) {
SliceInput sliceInput = record.input();
// read header
if (sliceInput.available() < 12) {
logMonitor.corruption(sliceInput.available(), "log record too small");
continue;
}
long sequenceBegin = sliceInput.readLong();
int updateSize = sliceInput.readInt();
// read entries
WriteBatchImpl writeBatch = readWriteBatch(sliceInput, updateSize);
// apply entries to memTable
if (memTable == null) {
memTable = new MemTable(internalKeyComparator);
}
writeBatch.forEach(new InsertIntoHandler(memTable, sequenceBegin));
// update the maxSequence
long lastSequence = sequenceBegin + updateSize - 1;
if (lastSequence > maxSequence) {
maxSequence = lastSequence;
}
// flush mem table if necessary
if (memTable.approximateMemoryUsage() > options.writeBufferSize()) {
writeLevel0Table(memTable, edit, null);
memTable = null;
}
}
// flush mem table
if (memTable != null && !memTable.isEmpty()) {
writeLevel0Table(memTable, edit, null);
}
return maxSequence;
} finally {
channel.close();
}
}
@Override
public byte[] get(byte[] key) throws DBException {
return get(key, new ReadOptions());
}
@Override
public byte[] get(byte[] key, ReadOptions options) throws DBException {
checkBackgroundException();
LookupKey lookupKey;
mutex.lock();
try {
SnapshotImpl snapshot = getSnapshot(options);
lookupKey = new LookupKey(Slices.wrappedBuffer(key), snapshot.getLastSequence());
// First look in the memtable, then in the immutable memtable (if any).
LookupResult lookupResult = memTable.get(lookupKey);
if (lookupResult != null) {
Slice value = lookupResult.getValue();
if (value == null) {
return null;
}
return value.getBytes();
}
if (immutableMemTable != null) {
lookupResult = immutableMemTable.get(lookupKey);
if (lookupResult != null) {
Slice value = lookupResult.getValue();
if (value == null) {
return null;
}
return value.getBytes();
}
}
} finally {
mutex.unlock();
}
// Not in memTables; try live files in level order
LookupResult lookupResult = versions.get(lookupKey);
// schedule compaction if necessary
mutex.lock();
try {
if (versions.needsCompaction()) {
maybeScheduleCompaction();
}
} finally {
mutex.unlock();
}
if (lookupResult != null) {
Slice value = lookupResult.getValue();
if (value != null) {
return value.getBytes();
}
}
return null;
}
@Override
public void put(byte[] key, byte[] value) throws DBException {
put(key, value, new WriteOptions());
}
@Override
public Snapshot put(byte[] key, byte[] value, WriteOptions options) throws DBException {
return writeInternal(new WriteBatchImpl().put(key, value), options);
}
@Override
public void delete(byte[] key) throws DBException {
writeInternal(new WriteBatchImpl().delete(key), new WriteOptions());
}
@Override
public Snapshot delete(byte[] key, WriteOptions options) throws DBException {
return writeInternal(new WriteBatchImpl().delete(key), options);
}
@Override
public void write(WriteBatch updates) throws DBException {
writeInternal((WriteBatchImpl) updates, new WriteOptions());
}
@Override
public Snapshot write(WriteBatch updates, WriteOptions options) throws DBException {
return writeInternal((WriteBatchImpl) updates, options);
}
public Snapshot writeInternal(WriteBatchImpl updates, WriteOptions options) throws DBException {
checkBackgroundException();
mutex.lock();
try {
long sequenceEnd;
if (updates.size() != 0) {
makeRoomForWrite(false);
// Get sequence numbers for this change set
final long sequenceBegin = versions.getLastSequence() + 1;
sequenceEnd = sequenceBegin + updates.size() - 1;
// Reserve this sequence in the version set
versions.setLastSequence(sequenceEnd);
// Log write
Slice record = writeWriteBatch(updates, sequenceBegin);
try {
log.addRecord(record, options.sync());
} catch (IOException e) {
throw Throwables.propagate(e);
}
// Update memtable
updates.forEach(new InsertIntoHandler(memTable, sequenceBegin));
} else {
sequenceEnd = versions.getLastSequence();
}
if (options.snapshot()) {
return new SnapshotImpl(versions.getCurrent(), sequenceEnd);
} else {
return null;
}
} finally {
mutex.unlock();
}
}
@Override
public WriteBatch createWriteBatch() {
checkBackgroundException();
return new WriteBatchImpl();
}
@Override
public SeekingIteratorAdapter iterator() {
return iterator(new ReadOptions());
}
public SeekingIteratorAdapter iterator(ReadOptions options) {
checkBackgroundException();
mutex.lock();
try {
DbIterator rawIterator = internalIterator();
// filter any entries not visible in our snapshot
SnapshotImpl snapshot = getSnapshot(options);
SnapshotSeekingIterator snapshotIterator =
new SnapshotSeekingIterator(
rawIterator, snapshot, internalKeyComparator.getUserComparator());
return new SeekingIteratorAdapter(snapshotIterator);
} finally {
mutex.unlock();
}
}
SeekingIterable<InternalKey, Slice> internalIterable() {
return new SeekingIterable<InternalKey, Slice>() {
@Override
public DbIterator iterator() {
return internalIterator();
}
};
}
DbIterator internalIterator() {
mutex.lock();
try {
// merge together the memTable, immutableMemTable, and tables in version set
MemTableIterator iterator = null;
if (immutableMemTable != null) {
iterator = immutableMemTable.iterator();
}
Version current = versions.getCurrent();
return new DbIterator(
memTable.iterator(),
iterator,
current.getLevel0Files(),
current.getLevelIterators(),
internalKeyComparator);
} finally {
mutex.unlock();
}
}
public Snapshot getSnapshot() {
checkBackgroundException();
mutex.lock();
try {
return new SnapshotImpl(versions.getCurrent(), versions.getLastSequence());
} finally {
mutex.unlock();
}
}
private SnapshotImpl getSnapshot(ReadOptions options) {
SnapshotImpl snapshot;
if (options.snapshot() != null) {
snapshot = (SnapshotImpl) options.snapshot();
} else {
snapshot = new SnapshotImpl(versions.getCurrent(), versions.getLastSequence());
snapshot.close(); // To avoid holding the snapshot active..
}
return snapshot;
}
private void makeRoomForWrite(boolean force) {
Preconditions.checkState(mutex.isHeldByCurrentThread());
boolean allowDelay = !force;
while (true) {
// todo background processing system need work
// if (!bg_error_.ok()) {
// // Yield previous error
// s = bg_error_;
// break;
// } else
if (allowDelay && versions.numberOfFilesInLevel(0) > L0_SLOWDOWN_WRITES_TRIGGER) {
// We are getting close to hitting a hard limit on the number of
// L0 files. Rather than delaying a single write by several
// seconds when we hit the hard limit, start delaying each
// individual write by 1ms to reduce latency variance. Also,
// this delay hands over some CPU to the compaction thread in
// case it is sharing the same core as the writer.
try {
mutex.unlock();
Thread.sleep(1);
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
throw new RuntimeException(e);
} finally {
mutex.lock();
}
// Do not delay a single write more than once
allowDelay = false;
} else if (!force && memTable.approximateMemoryUsage() <= options.writeBufferSize()) {
// There is room in current memtable
break;
} else if (immutableMemTable != null) {
// We have filled up the current memtable, but the previous
// one is still being compacted, so we wait.
backgroundCondition.awaitUninterruptibly();
} else if (versions.numberOfFilesInLevel(0) >= L0_STOP_WRITES_TRIGGER) {
// There are too many level-0 files.
// Log(options_.info_log, "waiting...\n");
backgroundCondition.awaitUninterruptibly();
} else {
// Attempt to switch to a new memtable and trigger compaction of old
Preconditions.checkState(versions.getPrevLogNumber() == 0);
// close the existing log
try {
log.close();
} catch (IOException e) {
throw new RuntimeException("Unable to close log file " + log.getFile(), e);
}
// open a new log
long logNumber = versions.getNextFileNumber();
try {
this.log =
Logs.createLogWriter(
new File(databaseDir, Filename.logFileName(logNumber)), logNumber);
} catch (IOException e) {
throw new RuntimeException(
"Unable to open new log file "
+ new File(databaseDir, Filename.logFileName(logNumber)).getAbsoluteFile(),
e);
}
// create a new mem table
immutableMemTable = memTable;
memTable = new MemTable(internalKeyComparator);
// Do not force another compaction there is space available
force = false;
maybeScheduleCompaction();
}
}
}
public void compactMemTable() throws IOException {
mutex.lock();
try {
compactMemTableInternal();
} finally {
mutex.unlock();
}
}
private void compactMemTableInternal() throws IOException {
Preconditions.checkState(mutex.isHeldByCurrentThread());
if (immutableMemTable == null) {
return;
}
try {
// Save the contents of the memtable as a new Table
VersionEdit edit = new VersionEdit();
Version base = versions.getCurrent();
writeLevel0Table(immutableMemTable, edit, base);
if (shuttingDown.get()) {
throw new DatabaseShutdownException("Database shutdown during memtable compaction");
}
// Replace immutable memtable with the generated Table
edit.setPreviousLogNumber(0);
edit.setLogNumber(log.getFileNumber()); // Earlier logs no longer needed
versions.logAndApply(edit);
immutableMemTable = null;
deleteObsoleteFiles();
} finally {
backgroundCondition.signalAll();
}
}
private void writeLevel0Table(MemTable mem, VersionEdit edit, Version base) throws IOException {
Preconditions.checkState(mutex.isHeldByCurrentThread());
// skip empty mem table
if (mem.isEmpty()) {
return;
}
// write the memtable to a new sstable
long fileNumber = versions.getNextFileNumber();
pendingOutputs.add(fileNumber);
mutex.unlock();
FileMetaData meta;
try {
meta = buildTable(mem, fileNumber);
} finally {
mutex.lock();
}
pendingOutputs.remove(fileNumber);
// Note that if file size is zero, the file has been deleted and
// should not be added to the manifest.
int level = 0;
if (meta != null && meta.getFileSize() > 0) {
Slice minUserKey = meta.getSmallest().getUserKey();
Slice maxUserKey = meta.getLargest().getUserKey();
if (base != null) {
level = base.pickLevelForMemTableOutput(minUserKey, maxUserKey);
}
edit.addFile(level, meta);
}
}
private FileMetaData buildTable(SeekingIterable<InternalKey, Slice> data, long fileNumber)
throws IOException {
File file = new File(databaseDir, Filename.tableFileName(fileNumber));
try {
InternalKey smallest = null;
InternalKey largest = null;
FileChannel channel = new FileOutputStream(file).getChannel();
try {
TableBuilder tableBuilder =
new TableBuilder(options, channel, new InternalUserComparator(internalKeyComparator));
for (Entry<InternalKey, Slice> entry : data) {
// update keys
InternalKey key = entry.getKey();
if (smallest == null) {
smallest = key;
}
largest = key;
tableBuilder.add(key.encode(), entry.getValue());
}
tableBuilder.finish();
} finally {
try {
channel.force(true);
} finally {
channel.close();
}
}
if (smallest == null) {
return null;
}
FileMetaData fileMetaData = new FileMetaData(fileNumber, file.length(), smallest, largest);
// verify table can be opened
tableCache.newIterator(fileMetaData);
pendingOutputs.remove(fileNumber);
return fileMetaData;
} catch (IOException e) {
file.delete();
throw e;
}
}
private void doCompactionWork(CompactionState compactionState) throws IOException {
Preconditions.checkState(mutex.isHeldByCurrentThread());
Preconditions.checkArgument(
versions.numberOfBytesInLevel(compactionState.getCompaction().getLevel()) > 0);
Preconditions.checkArgument(compactionState.builder == null);
Preconditions.checkArgument(compactionState.outfile == null);
// todo track snapshots
compactionState.smallestSnapshot = versions.getLastSequence();
// Release mutex while we're actually doing the compaction work
mutex.unlock();
try {
MergingIterator iterator = versions.makeInputIterator(compactionState.compaction);
Slice currentUserKey = null;
boolean hasCurrentUserKey = false;
long lastSequenceForKey = MAX_SEQUENCE_NUMBER;
while (iterator.hasNext() && !shuttingDown.get()) {
// always give priority to compacting the current mem table
mutex.lock();
try {
compactMemTableInternal();
} finally {
mutex.unlock();
}
InternalKey key = iterator.peek().getKey();
if (compactionState.compaction.shouldStopBefore(key) && compactionState.builder != null) {
finishCompactionOutputFile(compactionState);
}
// Handle key/value, add to state, etc.
boolean drop = false;
// todo if key doesn't parse (it is corrupted),
if (false /*!ParseInternalKey(key, &ikey)*/) {
// do not hide error keys
currentUserKey = null;
hasCurrentUserKey = false;
lastSequenceForKey = MAX_SEQUENCE_NUMBER;
} else {
if (!hasCurrentUserKey
|| internalKeyComparator.getUserComparator().compare(key.getUserKey(), currentUserKey)
!= 0) {
// First occurrence of this user key
currentUserKey = key.getUserKey();
hasCurrentUserKey = true;
lastSequenceForKey = MAX_SEQUENCE_NUMBER;
}
if (lastSequenceForKey <= compactionState.smallestSnapshot) {
// Hidden by an newer entry for same user key
drop = true; // (A)
} else if (key.getValueType() == ValueType.DELETION
&& key.getSequenceNumber() <= compactionState.smallestSnapshot
&& compactionState.compaction.isBaseLevelForKey(key.getUserKey())) {
// For this user key:
// (1) there is no data in higher levels
// (2) data in lower levels will have larger sequence numbers
// (3) data in layers that are being compacted here and have
// smaller sequence numbers will be dropped in the next
// few iterations of this loop (by rule (A) above).
// Therefore this deletion marker is obsolete and can be dropped.
drop = true;
}
lastSequenceForKey = key.getSequenceNumber();
}
if (!drop) {
// Open output file if necessary
if (compactionState.builder == null) {
openCompactionOutputFile(compactionState);
}
if (compactionState.builder.getEntryCount() == 0) {
compactionState.currentSmallest = key;
}
compactionState.currentLargest = key;
compactionState.builder.add(key.encode(), iterator.peek().getValue());
// Close output file if it is big enough
if (compactionState.builder.getFileSize()
>= compactionState.compaction.getMaxOutputFileSize()) {
finishCompactionOutputFile(compactionState);
}
}
iterator.next();
}
if (shuttingDown.get()) {
throw new DatabaseShutdownException("DB shutdown during compaction");
}
if (compactionState.builder != null) {
finishCompactionOutputFile(compactionState);
}
} finally {
mutex.lock();
}
// todo port CompactionStats code
installCompactionResults(compactionState);
}
private void openCompactionOutputFile(CompactionState compactionState)
throws FileNotFoundException {
Preconditions.checkNotNull(compactionState, "compactionState is null");
Preconditions.checkArgument(
compactionState.builder == null, "compactionState builder is not null");
mutex.lock();
try {
long fileNumber = versions.getNextFileNumber();
pendingOutputs.add(fileNumber);
compactionState.currentFileNumber = fileNumber;
compactionState.currentFileSize = 0;
compactionState.currentSmallest = null;
compactionState.currentLargest = null;
File file = new File(databaseDir, Filename.tableFileName(fileNumber));
compactionState.outfile = new FileOutputStream(file).getChannel();
compactionState.builder =
new TableBuilder(
options, compactionState.outfile, new InternalUserComparator(internalKeyComparator));
} finally {
mutex.unlock();
}
}
private void finishCompactionOutputFile(CompactionState compactionState) throws IOException {
Preconditions.checkNotNull(compactionState, "compactionState is null");
Preconditions.checkArgument(compactionState.outfile != null);
Preconditions.checkArgument(compactionState.builder != null);
long outputNumber = compactionState.currentFileNumber;
Preconditions.checkArgument(outputNumber != 0);
long currentEntries = compactionState.builder.getEntryCount();
compactionState.builder.finish();
long currentBytes = compactionState.builder.getFileSize();
compactionState.currentFileSize = currentBytes;
compactionState.totalBytes += currentBytes;
FileMetaData currentFileMetaData =
new FileMetaData(
compactionState.currentFileNumber,
compactionState.currentFileSize,
compactionState.currentSmallest,
compactionState.currentLargest);
compactionState.outputs.add(currentFileMetaData);
compactionState.builder = null;
compactionState.outfile.force(true);
compactionState.outfile.close();
compactionState.outfile = null;
if (currentEntries > 0) {
// Verify that the table is usable
tableCache.newIterator(outputNumber);
}
}
private void installCompactionResults(CompactionState compact) throws IOException {
Preconditions.checkState(mutex.isHeldByCurrentThread());
// Add compaction outputs
compact.compaction.addInputDeletions(compact.compaction.getEdit());
int level = compact.compaction.getLevel();
for (FileMetaData output : compact.outputs) {
compact.compaction.getEdit().addFile(level + 1, output);
pendingOutputs.remove(output.getNumber());
}
try {
versions.logAndApply(compact.compaction.getEdit());
deleteObsoleteFiles();
} catch (IOException e) {
// Compaction failed for some reason. Simply discard the work and try again later.
// Discard any files we may have created during this failed compaction
for (FileMetaData output : compact.outputs) {
File file = new File(databaseDir, Filename.tableFileName(output.getNumber()));
file.delete();
}
compact.outputs.clear();
}
}
int numberOfFilesInLevel(int level) {
return versions.getCurrent().numberOfFilesInLevel(level);
}
@Override
public long[] getApproximateSizes(Range... ranges) {
Preconditions.checkNotNull(ranges, "ranges is null");
long[] sizes = new long[ranges.length];
for (int i = 0; i < ranges.length; i++) {
Range range = ranges[i];
sizes[i] = getApproximateSizes(range);
}
return sizes;
}
public long getApproximateSizes(Range range) {
Version v = versions.getCurrent();
InternalKey startKey =
new InternalKey(
Slices.wrappedBuffer(range.start()),
SequenceNumber.MAX_SEQUENCE_NUMBER,
ValueType.VALUE);
InternalKey limitKey =
new InternalKey(
Slices.wrappedBuffer(range.limit()),
SequenceNumber.MAX_SEQUENCE_NUMBER,
ValueType.VALUE);
long startOffset = v.getApproximateOffsetOf(startKey);
long limitOffset = v.getApproximateOffsetOf(limitKey);
return (limitOffset >= startOffset ? limitOffset - startOffset : 0);
}
public long getMaxNextLevelOverlappingBytes() {
return versions.getMaxNextLevelOverlappingBytes();
}
private static class CompactionState {
private final Compaction compaction;
private final List<FileMetaData> outputs = newArrayList();
private long smallestSnapshot;
// State kept for output being generated
private FileChannel outfile;
private TableBuilder builder;
// Current file being generated
private long currentFileNumber;
private long currentFileSize;
private InternalKey currentSmallest;
private InternalKey currentLargest;
private long totalBytes;
private CompactionState(Compaction compaction) {
this.compaction = compaction;
}
public Compaction getCompaction() {
return compaction;
}
}
private static class ManualCompaction {
private final int level;
private final Slice begin;
private final Slice end;
private ManualCompaction(int level, Slice begin, Slice end) {
this.level = level;
this.begin = begin;
this.end = end;
}
}
private WriteBatchImpl readWriteBatch(SliceInput record, int updateSize) throws IOException {
WriteBatchImpl writeBatch = new WriteBatchImpl();
int entries = 0;
while (record.isReadable()) {
entries++;
ValueType valueType = ValueType.getValueTypeByPersistentId(record.readByte());
if (valueType == VALUE) {
Slice key = readLengthPrefixedBytes(record);
Slice value = readLengthPrefixedBytes(record);
writeBatch.put(key, value);
} else if (valueType == DELETION) {
Slice key = readLengthPrefixedBytes(record);
writeBatch.delete(key);
} else {
throw new IllegalStateException("Unexpected value type " + valueType);
}
}
if (entries != updateSize) {
throw new IOException(
String.format(
"Expected %d entries in log record but found %s entries", updateSize, entries));
}
return writeBatch;
}
private Slice writeWriteBatch(WriteBatchImpl updates, long sequenceBegin) {
Slice record = Slices.allocate(SIZE_OF_LONG + SIZE_OF_INT + updates.getApproximateSize());
final SliceOutput sliceOutput = record.output();
sliceOutput.writeLong(sequenceBegin);
sliceOutput.writeInt(updates.size());
updates.forEach(
new Handler() {
@Override
public void put(Slice key, Slice value) {
sliceOutput.writeByte(VALUE.getPersistentId());
writeLengthPrefixedBytes(sliceOutput, key);
writeLengthPrefixedBytes(sliceOutput, value);
}
@Override
public void delete(Slice key) {
sliceOutput.writeByte(DELETION.getPersistentId());
writeLengthPrefixedBytes(sliceOutput, key);
}
});
return record.slice(0, sliceOutput.size());
}
private static class InsertIntoHandler implements Handler {
private long sequence;
private final MemTable memTable;
public InsertIntoHandler(MemTable memTable, long sequenceBegin) {
this.memTable = memTable;
this.sequence = sequenceBegin;
}
@Override
public void put(Slice key, Slice value) {
memTable.add(sequence++, VALUE, key, value);
}
@Override
public void delete(Slice key) {
memTable.add(sequence++, DELETION, key, Slices.EMPTY_SLICE);
}
}
public static class DatabaseShutdownException extends DBException {
public DatabaseShutdownException() {}
public DatabaseShutdownException(String message) {
super(message);
}
}
public static class BackgroundProcessingException extends DBException {
public BackgroundProcessingException(Throwable cause) {
super(cause);
}
}
private Object suspensionMutex = new Object();
private int suspensionCounter = 0;
@Override
public void suspendCompactions() throws InterruptedException {
compactionExecutor.execute(
new Runnable() {
@Override
public void run() {
try {
synchronized (suspensionMutex) {
suspensionCounter++;
suspensionMutex.notifyAll();
while (suspensionCounter > 0 && !compactionExecutor.isShutdown()) {
suspensionMutex.wait(500);
}
}
} catch (InterruptedException e) {
}
}
});
synchronized (suspensionMutex) {
while (suspensionCounter < 1) {
suspensionMutex.wait();
}
}
}
@Override
public void resumeCompactions() {
synchronized (suspensionMutex) {
suspensionCounter--;
suspensionMutex.notifyAll();
}
}
@Override
public void compactRange(byte[] begin, byte[] end) throws DBException {
throw new UnsupportedOperationException("Not yet implemented");
}
}
/**
* @author oifa yulian modified copy of linked blocking queue currently goes to garbage - node
* elements which are based on 2 pointers : item and next and should signal boolean value under
* offer function
*/
public class ConcurrentCyclicFIFO<E> {
static class Node<E> {
volatile E item;
Node<E> next;
Node(E x) {
item = x;
}
}
/** Current number of elements */
private final AtomicInteger count = new AtomicInteger(0);
/** Head of linked list */
private transient Node<E> head;
/** Tail of linked list */
private transient Node<E> last;
/** Lock held by take, poll, etc */
private final ReentrantLock takeLock = new ReentrantLock();
/** Wait queue for waiting takes */
private final Condition notEmpty = takeLock.newCondition();
/** Lock held by put, offer, etc */
private final ReentrantLock putLock = new ReentrantLock();
/**
* Signals a waiting take. Called only from put/offer (which do not otherwise ordinarily lock
* takeLock.)
*/
private void signalNotEmpty() {
final ReentrantLock takeLock = this.takeLock;
takeLock.lock();
try {
notEmpty.signal();
} finally {
takeLock.unlock();
}
}
/**
* Creates a node and links it at end of queue.
*
* @param x the item
*/
private void insert(Node<E> x) {
last = last.next = x;
}
/**
* Removes a node from head of queue,
*
* @return the node
*/
private Node<E> extract() {
Node<E> current = head;
head = head.next;
current.item = head.item;
head.item = null;
return current;
}
public ConcurrentCyclicFIFO() {
last = head = new Node<E>(null);
}
public int size() {
return count.get();
}
public boolean offer(E e) {
if (e == null) throw new NullPointerException();
final AtomicInteger count = this.count;
boolean shouldSignal = false;
final ReentrantLock putLock = this.putLock;
putLock.lock();
try {
insert(new Node(e));
shouldSignal = (count.getAndIncrement() == 0);
} finally {
putLock.unlock();
}
if (shouldSignal) signalNotEmpty();
return !shouldSignal;
}
public E take() throws InterruptedException {
Node<E> x;
final AtomicInteger count = this.count;
final ReentrantLock takeLock = this.takeLock;
takeLock.lockInterruptibly();
try {
try {
while (count.get() == 0) notEmpty.await();
} catch (InterruptedException ie) {
notEmpty.signal(); // propagate to a non-interrupted thread
throw ie;
}
x = extract();
if (count.getAndDecrement() > 1) notEmpty.signal();
} finally {
takeLock.unlock();
}
E result = x.item;
// temporary clearence
x.item = null;
x.next = null;
return result;
}
public E poll() {
final AtomicInteger count = this.count;
if (count.get() == 0) return null;
Node<E> x = null;
final ReentrantLock takeLock = this.takeLock;
takeLock.lock();
try {
if (count.get() > 0) {
x = extract();
if (count.getAndDecrement() > 1) notEmpty.signal();
}
} finally {
takeLock.unlock();
}
if (x != null) {
E result = x.item;
// temporary clearence
x.item = null;
x.next = null;
return result;
}
return null;
}
public void clear() {
putLock.lock();
takeLock.lock();
try {
head.next = null;
assert head.item == null;
last = head;
count.set(0);
} finally {
takeLock.unlock();
putLock.unlock();
}
}
}
public OTEServerDiscoveryImpl() {
lock = new ReentrantLock();
condition = lock.newCondition();
store = new OTEServerStoreImpl(lock, condition);
notification = new OteServerNotification(store);
}
/**
* Basic test of the ability to add to a buffer with a fixed capacity queue and to drain the
* elements from the queue including tests of the non-blocking aspects of the API.
*
* @throws TimeoutException
* @throws ExecutionException
* @throws InterruptedException
*/
public void test_blockingBuffer()
throws InterruptedException, ExecutionException, TimeoutException {
final Integer e0 = new Integer(0);
final Integer e1 = new Integer(1);
final Integer e2 = new Integer(2);
final int queueCapacity = 3;
final BlockingQueue<Integer[]> queue = new ArrayBlockingQueue<Integer[]>(queueCapacity);
final int chunkSize = 4;
final long chunkTimeout = 1000;
final TimeUnit chunkTimeoutUnit = TimeUnit.MILLISECONDS;
/*
* The test timeout is just a smidge longer than the chunk timeout.
*
* Note: use Long.MAX_VALUE iff debugging.
*/
// final long testTimeout = Long.MAX_VALUE;
final long testTimeout = chunkTimeout + 20;
final boolean ordered = false;
final BlockingBuffer<Integer[]> buffer =
new BlockingBuffer<Integer[]>(queue, chunkSize, chunkTimeout, chunkTimeoutUnit, ordered);
// buffer is empty.
assertTrue(buffer.isOpen());
assertTrue(buffer.isEmpty());
assertEquals("chunkCount", 0L, buffer.getChunksAddedCount());
assertEquals("elementCount", 0L, buffer.getElementsAddedCount());
final IAsynchronousIterator<Integer[]> itr = buffer.iterator();
// nothing available from the iterator (non-blocking test).
assertFalse(itr.hasNext(1, TimeUnit.NANOSECONDS));
assertNull(itr.next(1, TimeUnit.NANOSECONDS));
// add an element to the buffer - should not block.
buffer.add(new Integer[] {e0});
// should be one element and one chunk accepted by the buffer.
assertTrue(buffer.isOpen());
assertFalse(buffer.isEmpty());
assertEquals("chunkCount", 1L, buffer.getChunksAddedCount());
assertEquals("elementCount", 1L, buffer.getElementsAddedCount());
// something should be available now (non-blocking).
assertTrue(itr.hasNext(1, TimeUnit.NANOSECONDS));
// something should be available now (blocking).
assertTrue(itr.hasNext());
// add another element to the buffer - should not block.
buffer.add(new Integer[] {e1});
// should be two elements and two chunks accepted into the buffer
assertTrue(buffer.isOpen());
assertFalse(buffer.isEmpty());
assertEquals("chunkCount", 2L, buffer.getChunksAddedCount());
assertEquals("elementCount", 2L, buffer.getElementsAddedCount());
final ReentrantLock lock = new ReentrantLock();
final Condition cond = lock.newCondition();
final AtomicBoolean proceedFlag = new AtomicBoolean(false);
// future of task writing a 3rd element on the buffer.
final Future<?> producerFuture =
service.submit(
new Callable<Void>() {
public Void call() throws Exception {
lock.lockInterruptibly();
try {
if (!proceedFlag.get()) {
cond.await();
}
/*
* add another element - should block until we take an
* element using the iterator.
*/
buffer.add(new Integer[] {e2});
/*
* itr.hasNext() will block until the buffer is closed.
*/
buffer.close();
} finally {
lock.unlock();
}
// done.
return null;
}
});
// future of task draining the buffer.
final Future<?> consumerFuture =
service.submit(
new Callable<Void>() {
public Void call() throws Exception {
try {
lock.lockInterruptibly();
try {
assertTrue(itr.hasNext());
// take the first chunk - two elements.
if (log.isInfoEnabled()) log.info("Awaiting first chunk");
assertSameArray(new Integer[] {e0, e1}, itr.next(50, TimeUnit.MILLISECONDS));
if (log.isInfoEnabled()) log.info("Have first chunk");
/*
* Verify that we obtained the first chunk before the
* buffer was closed. Otherwise next() blocked
* attempting to compile a full chunk until the producer
* timeout, at which point the producer closed the
* buffer and next() noticed the closed buffer and
* returned.
*/
assertTrue(buffer.isOpen());
assertFalse("buffer was closed.", itr.isExhausted());
/*
* Verify that nothing is available from the iterator
* (non-blocking test).
*/
assertFalse(itr.hasNext(1, TimeUnit.NANOSECONDS));
assertNull(itr.next(1, TimeUnit.NANOSECONDS));
// Signal the producer that it should continue.
proceedFlag.set(true);
cond.signal();
} finally {
lock.unlock();
}
// should block until we close the buffer.
assertTrue(itr.hasNext());
// last chunk
assertSameArray(new Integer[] {e2}, itr.next());
// should be immediately false.
assertFalse(itr.hasNext(1, TimeUnit.NANOSECONDS));
// should be immediately null.
assertNull(itr.next(1, TimeUnit.NANOSECONDS));
// The synchronous API should also report an exhausted
// itr.
assertFalse(itr.hasNext());
try {
itr.next();
fail("Expecting: " + NoSuchElementException.class);
} catch (NoSuchElementException ex) {
if (log.isInfoEnabled()) log.info("Ignoring expected exception: " + ex);
}
return null;
} catch (Throwable t) {
log.error("Consumer failed or blocked: " + t, t);
throw new Exception(t);
}
}
});
// wait a little bit for the producer future.
producerFuture.get(testTimeout, chunkTimeoutUnit);
// wait a little bit for the consumer future.
consumerFuture.get(testTimeout, chunkTimeoutUnit);
}