HBase MemStoreFlusher

HBase MemStore Flush由类org.apache.hadoop.hbase.regionserver.MemStoreFlusher实现，具体表现为HRegionServer中的一个实例变量cacheFlusher，类结构如下：

class MemStoreFlusher extends HasThread implements FlushRequester {
    ......
}

MemStoreFlusher实质是一个线程类。

HasThread可以理解为Thread的一个代码类:

/**
 * Abstract class which contains a Thread and delegates the common Thread
 * methods to that instance.
 * 
 * The purpose of this class is to workaround Sun JVM bug #6915621, in which
 * something internal to the JDK uses Thread.currentThread() as a monitor lock.
 * This can produce deadlocks like HBASE-4367, HBASE-4101, etc.
 */
public abstract class HasThread implements Runnable {

    private final Thread thread;

    public HasThread() {
        this.thread = new Thread(this);
    }

    public HasThread(String name) {
        this.thread = new Thread(this, name);
    }

    public Thread getThread() {
        return thread;
    }

    public abstract void run();

    // // Begin delegation to Thread

    public final String getName() {
        return thread.getName();
    }

    public void interrupt() {
        thread.interrupt();
    }

    public final boolean isAlive() {
        return thread.isAlive();
    }

    public boolean isInterrupted() {
        return thread.isInterrupted();
    }

    public final void setDaemon(boolean on) {
        thread.setDaemon(on);
    }

    public final void setName(String name) {
        thread.setName(name);
    }

    public final void setPriority(int newPriority) {
        thread.setPriority(newPriority);
    }

    public void setUncaughtExceptionHandler(UncaughtExceptionHandler eh) {
        thread.setUncaughtExceptionHandler(eh);
    }

    public void start() {
        thread.start();
    }

    public final void join() throws InterruptedException {
        thread.join();
    }

    public final void join(long millis, int nanos) throws InterruptedException {
        thread.join(millis, nanos);
    }

    public final void join(long millis) throws InterruptedException {
        thread.join(millis);
    }
    // // End delegation to Thread

}

FlushRequester是一个接口,仅包含一个方法:

/**
 * Request a flush.
 */
public interface FlushRequester {

    /**
     * Tell the listener the cache needs to be flushed.
     * 
     * @param region
     *            the HRegion requesting the cache flush
     */
    void requestFlush(HRegion region);

}

核心变量

// These two data members go together. Any entry in the one must have
// a corresponding entry in the other.
private final BlockingQueue<FlushQueueEntry> flushQueue = new DelayQueue<FlushQueueEntry>();

private final Map<HRegion, FlushRegionEntry> regionsInQueue = new HashMap<HRegion, FlushRegionEntry>();

flushQueue：DelayQueue队列，元素类型为FlushQueueEntry，代表某一Region的Flush请求，Flusher线程不断地从该队列中获取请求信息，完成Region的Flush操作；

regionsInQueue：维护HRegion实例与请求FlushRegionEntry之间的对应关系；

如注释中所说，如果某一个FlushQueueEntry实例存在于flushQueue中，那么它必然存在于regionsInQueue中，后者看似多余，其实不然，例如，验证某一Region是否已经发起过Flush请求。

private AtomicBoolean wakeupPending = new AtomicBoolean();

wakeupPending：主要与flushQueue结合使用，flushQueue是一种阻塞队列，当队列为空时，poll操作会将线程阻塞一段时间，某些情况下需要在flushQueue中加入一个“空元素”，以唤醒线程工作，但如果线程本次操作（后面会看到Flusher线程工作实质是一个循环操作）已经被加入“空”元素，则不需要重复加入。

private final long threadWakeFrequency;

threadWakeFrequency：用于flushQueue执行poll操作时，最多等待多长时间，配置项为hbase.server.thread.wakefrequency；

private final HRegionServer server;

server：当前HRegionServer实例；

private final ReentrantLock lock = new ReentrantLock();

private final Condition flushOccurred = lock.newCondition();

lock、flushOccurred：用于同步操作，类似于synchronized、wait、signal、signalAll；

protected final long globalMemStoreLimit;

protected final long globalMemStoreLimitLowMark;

private static final float DEFAULT_UPPER = 0.4f;

private static final float DEFAULT_LOWER = 0.35f;

private static final String UPPER_KEY = "hbase.regionserver.global.memstore.upperLimit";

private static final String LOWER_KEY = "hbase.regionserver.global.memstore.lowerLimit";

globalMemStoreLimit、globalMemStoreLimitLowMark：表示HRegionServer整个MemStore的上下限值，当整个MemStore的内存消耗值达到下限值时就会采取相应的措施；

private long blockingStoreFilesNumber;

private long blockingWaitTime;

blockingStoreFilesNumber：对某一Region执行Flush操作时，如果该Region中的某一Store中已有的StoreFile数目超过blockingStoreFilesNumber（hbase.hstore.compactionThreshold），则该Region的Flush操作会被最多延迟blockingWaitTime（hbase.hstore.blockingWaitTime）。

Flush请求

所有的Region Flush请求会被放到一个DelayedQueue中，因此放入该队列的元素必须实现Delayed接口：

interface FlushQueueEntry extends Delayed {
}

Flush请求会被分为两种类型：“空”请求与实质请求，“空”请求主要用于唤醒线程，实质请求即为Region Flush请求。

“空”请求：

/**
 * Token to insert into the flush queue that ensures that the flusher does
 * not sleep
 */
static class WakeupFlushThread implements FlushQueueEntry {

    @Override
    public long getDelay(TimeUnit unit) {
        return 0;
    }

    @Override
    public int compareTo(Delayed o) {
        return -1;
    }

}

“空”请求的作用主要是唤醒，不需要任何实质性的内容，且延迟时间被设为0，表示立即可从队列中获取。

实质请求：

/**
 * Datastructure used in the flush queue. Holds region and retry count.
 * Keeps tabs on how old this object is. Implements {@link Delayed}. On
 * construction, the delay is zero. When added to a delay queue, we'll come
 * out near immediately. Call {@link #requeue(long)} passing delay in
 * milliseconds before readding to delay queue if you want it to stay there
 * a while.
 */
static class FlushRegionEntry implements FlushQueueEntry {

    private final HRegion region;

    private final long createTime;

    private long whenToExpire;

    private int requeueCount = 0;

    FlushRegionEntry(final HRegion r) {
        this.region = r;

        this.createTime = System.currentTimeMillis();

        this.whenToExpire = this.createTime;
    }

    /**
     * @param maximumWait
     * @return True if we have been delayed > <code>maximumWait</code>
     *         milliseconds.
     */
    public boolean isMaximumWait(final long maximumWait) {
        return (System.currentTimeMillis() - this.createTime) > maximumWait;
    }

    /**
     * @return Count of times {@link #resetDelay()} was called; i.e this is
     *         number of times we've been requeued.
     */
    public int getRequeueCount() {
        return this.requeueCount;
    }

    /**
     * @param when
     *            When to expire, when to come up out of the queue. Specify
     *            in milliseconds. This method adds
     *            System.currentTimeMillis() to whatever you pass.
     * @return This.
     */
    public FlushRegionEntry requeue(final long when) {
        this.whenToExpire = System.currentTimeMillis() + when;

        this.requeueCount++;

        return this;
    }

    @Override
    public long getDelay(TimeUnit unit) {
        return unit.convert(this.whenToExpire - System.currentTimeMillis(),
                TimeUnit.MILLISECONDS);
    }

    @Override
    public int compareTo(Delayed other) {
        return Long.valueOf(
                getDelay(TimeUnit.MILLISECONDS)
                        - other.getDelay(TimeUnit.MILLISECONDS)).intValue();
        }

    @Override
    public String toString() {
        return "[flush region "
                + Bytes.toStringBinary(region.getRegionName()) + "]";
    }

}

region：表示发起Flush请求的HRegion实例；

createTime：表示Flush请求的创建时间；

whenToExpire：表示Flush请求的过期时间；

requeueCount：表示Flush请求的入队次数，因为有些Flush请求根据情况需要被延迟执行，所以需要重新入队。

构造函数

MemStoreFlusher的构造函数主要用于初始化上述这些变量，其中比较重要的是RegionServer整个MemStore内存消耗上下限值的计算：

long max = ManagementFactory.getMemoryMXBean().getHeapMemoryUsage()
                .getMax();

this.globalMemStoreLimit = globalMemStoreLimit(max, DEFAULT_UPPER,
        UPPER_KEY, conf);

long lower = globalMemStoreLimit(max, DEFAULT_LOWER, LOWER_KEY, conf);

if (lower > this.globalMemStoreLimit) {
    lower = this.globalMemStoreLimit;

    LOG.info("Setting globalMemStoreLimitLowMark == globalMemStoreLimit "
            + "because supplied " + LOWER_KEY + " was > " + UPPER_KEY);
}

this.globalMemStoreLimitLowMark = lower;

方法globalMemStoreLimit的相关代码如下：

/**
 * Calculate size using passed <code>key</code> for configured percentage of
 * <code>max</code>.
 * 
 * @param max
 * @param defaultLimit
 * @param key
 * @param c
 * @return Limit.
 */
static long globalMemStoreLimit(final long max, final float defaultLimit,
        final String key, final Configuration c) {
    float limit = c.getFloat(key, defaultLimit);

    return getMemStoreLimit(max, limit, defaultLimit);
}

static long getMemStoreLimit(final long max, final float limit,
        final float defaultLimit) {
    float effectiveLimit = limit;

    if (limit >= 0.9f || limit < 0.1f) {
        LOG.warn("Setting global memstore limit to default of "
                + defaultLimit
                + " because supplied value outside allowed range of 0.1 -> 0.9");

        effectiveLimit = defaultLimit;
    }

    return (long) (max * effectiveLimit);
}

循环Flush操作

Flush请求的处理是在一个循环的操作中被处理的：

@Override
public void run() {
    while (!this.server.isStopped()) {
        FlushQueueEntry fqe = null;    

        try {
            ......    
        } catch (InterruptedException ex) {
            continue;
        } catch (ConcurrentModificationException ex) {
            continue;
        } catch (Exception ex) {
            LOG.error("Cache flusher failed for entry " + fqe, ex);

            if (!server.checkFileSystem()) {
                break;
            }
        }
    }

    this.regionsInQueue.clear();

    this.flushQueue.clear();

    // Signal anyone waiting, so they see the close flag
    lock.lock();

    try {
        flushOccurred.signalAll();
    } finally {
        lock.unlock();
    }

    LOG.info(getName() + " exiting");
}

只要该HRegionServer没有被请求停止，则该操作将一直被执行,不断地从请求队列中获取具体的请求fqe，然后执行Flush操作，具体的操作被包含在一个try、catch块中。如果该HRegionServer已经被请求停止，则会清空相应的数据结构及唤醒其它被阻塞的线程。

某一Flush操作

wakeupPending.set(false); // allow someone to wake us up again

fqe = flushQueue.poll(threadWakeFrequency,
        TimeUnit.MILLISECONDS);

从队列中获取一个Flush请求，如果此时队列为空则本线程会被阻塞直至超时，wakeupPending.set(false)则表示外界在某些条件下可以通过向队列中加入一个“空”请求（WakeupFlushThread）来唤醒被阻塞的线程。

如果从队列中获取数据的结果fqe为null或者为WakeupFlushThread实例时，则执行以下代码：

if (fqe == null || fqe instanceof WakeupFlushThread) {
    if (isAboveLowWaterMark()) {
        LOG.debug("Flush thread woke up because memory above low water="
                + StringUtils
                                        .humanReadableInt(this.globalMemStoreLimitLowMark));

        if (!flushOneForGlobalPressure()) {
            // Wasn't able to flush any region, but we're above
            // low water mark
            // This is unlikely to happen, but might happen when
            // closing the
            // entire server - another thread is flushing
            // regions. We'll just
            // sleep a little bit to avoid spinning, and then
            // pretend that
            // we flushed one, so anyone blocked will check
            // again
            lock.lock();

            try {
                Thread.sleep(1000);

                flushOccurred.signalAll();
            } finally {
                lock.unlock();
            }
        }

        // Enqueue another one of these tokens so we'll wake up
        // again
        wakeupFlushThread();
    }

    continue;
}

此时并没有获取到实质的Flush请求，主要判断当前RegionServer整个MemStore的内存消耗是否已达到下限临界值，如果已达到下限临界值，则为了缓解内存压力，需要选取某一个Region进行Flush操作。

判断内存消耗由方法isAboveHighWaterMark完成：

/**
 * Return true if we're above the high watermark
 */
private boolean isAboveLowWaterMark() {
    return server.getRegionServerAccounting().getGlobalMemstoreSize() >= globalMemStoreLimitLowMark;
}

如果isAboveLowWaterMark返回值为true，则表示此时RegionServer的整个MemStore内存消耗已达到下限临界值，需要选取一个Region进行Flush以缓解内存压力，由方法flushOneForGlobalPressure完成：

/**
 * The memstore across all regions has exceeded the low water mark. Pick one
 * region to flush and flush it synchronously (this is called from the flush
 * thread)
 * 
 * @return true if successful
 */
private boolean flushOneForGlobalPressure() {
    SortedMap<Long, HRegion> regionsBySize = server
            .getCopyOfOnlineRegionsSortedBySize();

    Set<HRegion> excludedRegions = new HashSet<HRegion>();

    boolean flushedOne = false;

    while (!flushedOne) {
        ......    
    }

    return true;
}

上述代码的主体思想是不断循环操作，直接成功选取某一Region完成Flush操作为止，在循环操作开始之前，已经依据Region大小获取到了该RegionServer上的所有Region：regionsBySize（SortedMap实现，依据Region大小作为排序依据，顺序为从大到小），如果选取的Region在执行Flush操作时发生了某些异常，导致Flush失败，则将其保存至excludedRegions，以使在下次选取过程中能够将其排除。

循环中的操作流程如下：

// Find the biggest region that doesn't have too many storefiles
// (might be null!)
HRegion bestFlushableRegion = getBiggestMemstoreRegion(
        regionsBySize, excludedRegions, true);

选取当前状态下最适合进行Flush操作的Region，该Region需要满足两个条件：

（1）Region没有包含超过一定数量的StoreFile；

（2）在满足（1）的所有Region中大小为最大值。

具体执行时代码如下：

private HRegion getBiggestMemstoreRegion(
        SortedMap<Long, HRegion> regionsBySize,
        Set<HRegion> excludedRegions, boolean checkStoreFileCount) {
    synchronized (regionsInQueue) {
        for (HRegion region : regionsBySize.values()) {
            //如果Region出现在excludedRegions中，则表示该Region是unflushable的。
            if (excludedRegions.contains(region)) {
                continue;
            }

            if (checkStoreFileCount && isTooManyStoreFiles(region)) {
                continue;
            }
                
            return region;
        }
    }
        
    return null;
}

private boolean isTooManyStoreFiles(HRegion region) {
    for (Store hstore : region.stores.values()) {
        if (hstore.getStorefilesCount() > this.blockingStoreFilesNumber) {
            return true;
        }
    }

    return false;
}

因为regionsBySize中的Region就是根据Region大小从大到小排列的，只要依次处理其中的Region即可，如果该Region即没有出现在excludedRegions，也没有包含过多的StoreFile（checkStoreFileCount为true），即该Region就是bestFlushableRegion。

为了防止bestFlushableRegion为null（如果目前所有的Region包含的StoreFile数目都大于临界值blockingStoreFilesNumber），我们需要选取一个目前最大的Region作为备选，即时它拥有的StoreFile数目大于临界值blockingStoreFilesNumber。

// Find the biggest region, total, even if it might have too many
// flushes.
HRegion bestAnyRegion = getBiggestMemstoreRegion(regionsBySize,
        excludedRegions, false);

if (bestAnyRegion == null) {
    LOG.error("Above memory mark but there are no flushable regions!");

    return false;
}

执行getBiggestMemstoreRegion方法时，checkStoreFileCount为false，表示这些选取不考虑Region包含StoreFile的数目。

如果我们无法获取一个bestAnyRegion（bestAnyRegion为null），表示目前虽然内存压力较大，但是我们无法选取出一个可进行Flush操作的Region，直接返回false即可。

无法选取出一个可进行Flush操作的Region的原因一般有两个：

（1）在循环选取的过程中，我们发现所有的Region进行Flush操作时都失败了（可能原因是HDFS失效），它们都会出现在excludedRegions中，因此，会导致上述方法执行时返回值为null；

（2）RegionServer开始执行关闭操作。

HRegion regionToFlush;

if (bestFlushableRegion != null
        && bestAnyRegion.memstoreSize.get() > 2 * bestFlushableRegion.memstoreSize
                .get()) {
    // Even if it's not supposed to be flushed, pick a region if
    // it's more than twice
    // as big as the best flushable one - otherwise when we're under
    // pressure we make
    // lots of little flushes and cause lots of compactions, etc,
    // which just makes
    // life worse!
    if (LOG.isDebugEnabled()) {
        LOG.debug("Under global heap pressure: "
                + "Region "
                + bestAnyRegion.getRegionNameAsString()
                + " has too many "
                + "store files, but is "
                + StringUtils
                        .humanReadableInt(bestAnyRegion.memstoreSize
                                .get())
                + " vs best flushable region's "
                + StringUtils
                        .humanReadableInt(bestFlushableRegion.memstoreSize
                                .get()) + ". Choosing the bigger.");
    }

    regionToFlush = bestAnyRegion;
} else {
    if (bestFlushableRegion == null) {
        regionToFlush = bestAnyRegion;
    } else {
        regionToFlush = bestFlushableRegion;
    }
}

根据bestFlushableRegion和bestAnyRegion的选取结果，决定最后的选取结果regionToFlush：

（1）虽然bestFlushableRegion不为null，但bestAnyRegion的MemStore大小比bestFlushableRegion的MemStore大小两倍还要在，此时regionToFlush = bestAnyRegion；

（2）否则，如果bestFlushableRegion为null，则regionToFlush = bestAnyRegion，否则regionToFlush = bestFlushableRegion。

至此，我们已经选取出了需要进行Flush操作的Region：regionToFlush，接下来对其进行Flush即可：

Preconditions.checkState(regionToFlush.memstoreSize.get() > 0);

LOG.info("Flush of region " + regionToFlush
        + " due to global heap pressure");

flushedOne = flushRegion(regionToFlush, true);

if (!flushedOne) {
    LOG.info("Excluding unflushable region " + regionToFlush
            + " - trying to find a different region to flush.");

    excludedRegions.add(regionToFlush);
}

如果该Region的Flush操作失败，即flushRegion的返回值为false，将其添加至excludedRegions中，并继续循环选取。

如果flushOneForGlobalPressure的返回值为false，则表示我们无法选取一个Region进行Flush，如注释所说，造成这种情况可能原因是RegionServer正处于关闭中，此时，会有其它线程来负责Region的Flush操作。我们仅仅需要休眠一会儿，假装我们完成了一个Region的Flush，然后就可以唤醒其它因内存压力而阻塞的线程了，使它们可以再次对内存消耗大小进行确认（后面会讲述为何有线程被阻塞）。

如果从队列中获取数据的结果fqe为FlushRegionEntry实例，则会直接执行以下代码：

FlushRegionEntry fre = (FlushRegionEntry) fqe;

if (!flushRegion(fre)) {
    break;
}

直接执行相应Region的Flush操作，如果发生错误（认为不可修复），则结束MemStoreFlusher线程的循环操作，执行清理工作。

MemStore与Put

在我们将大批量的数据定入HBase时，可能会由于内存的原因导致写入操作的Block，主要有以下两个方面的原因：

（1）reclaimMemStoreMemory

该方法是MemStoreFlusher的实例方法，在执行具体的Region batchMutate操作（完成写入操作）之前被调用，

HRegion region = getRegion(regionName);

if (!region.getRegionInfo().isMetaTable()) {
    /*
     * This method blocks callers until we're down to a safe
     * amount of memstore consumption.
     * 
     * ******************************************************
     */
    this.cacheFlusher.reclaimMemStoreMemory();
}

可见，一般地用户表都会在实际写入数据之前都会调用此方法，该方法可能会导致写入的阻塞。

reclaimMemStoreMemory分两种情况进行处理：isAboveHighWaterMark、isAboveLowWaterMark。

isAboveHighWaterMark：RegionServer整个MemStore的内存消耗值超过上限值

if (isAboveHighWaterMark()) {
    lock.lock();

    try {
        boolean blocked = false;

        long startTime = 0;

        while (isAboveHighWaterMark() && !server.isStopped()) {
            if (!blocked) {
                startTime = EnvironmentEdgeManager.currentTimeMillis();

                LOG.info("Blocking updates on "
                        + server.toString()
                        + ": the global memstore size "
                        + StringUtils.humanReadableInt(server
                                        .getRegionServerAccounting()
                                        .getGlobalMemstoreSize())
                        + " is >= than blocking "
                        + StringUtils
                                        .humanReadableInt(globalMemStoreLimit)
                        + " size");
            }

            blocked = true;

            wakeupFlushThread();

            try {
                // we should be able to wait forever, but we've seen a
                // bug where
                // we miss a notify, so put a 5 second bound on it at
                // least.
                flushOccurred.await(5, TimeUnit.SECONDS);
            } catch (InterruptedException ie) {
                Thread.currentThread().interrupt();
            }
        }

        if (blocked) {
            final long totalTime = EnvironmentEdgeManager
                    .currentTimeMillis() - startTime;

            if (totalTime > 0) {
                this.updatesBlockedMsHighWater.add(totalTime);
            }

            LOG.info("Unblocking updates for server "
                    + server.toString());
        }
    } finally {
        lock.unlock();
    }
}

当写入数据之前，如果我们发现当内存的消耗已经超过上限值时，会有一个循环等待的过程，直到内存的消耗值低于上限值为止，在每次等待操作之前都会通过wakeupFlushThread方法在Flush请求队列放入一个空元素，以激活MemStoreFlusher线程进行工作（可能会选取某一Region进行Flush），其中，上限值的判断如下所示：

/**
 * Return true if global memory usage is above the high watermark
 */
private boolean isAboveHighWaterMark() {
    return server.getRegionServerAccounting().getGlobalMemstoreSize() >= globalMemStoreLimit;
}

isAboveLowWaterMark：RegionServer的整个MemStore的内存消耗值仅超过下限值

else if (isAboveLowWaterMark()) {
    wakeupFlushThread();
}

此时，不需要阻塞写入操作，仅仅需要在Flush请求队列中加入一个“空”元素，促使MemStoreFlusher工作即可。

（2）checkResources

/**
 * Perform a batch of mutations. It supports only Put and Delete mutations
 * and will ignore other types passed.
 * 
 * @param mutationsAndLocks
 *            the list of mutations paired with their requested lock IDs.
 * @return an array of OperationStatus which internally contains the
 *         OperationStatusCode and the exceptionMessage if any.
 * @throws IOException
 */
public OperationStatus[] batchMutate(
        Pair<Mutation, Integer>[] mutationsAndLocks) throws IOException {
    BatchOperationInProgress<Pair<Mutation, Integer>> batchOp = new BatchOperationInProgress<Pair<Mutation, Integer>>(
            mutationsAndLocks);

    boolean initialized = false;

    while (!batchOp.isDone()) {
        checkReadOnly();
        
        // Check if resources to support an update, may be blocked.
        checkResources();
         ......
            
    }

    return batchOp.retCodeDetails;
}

在Region batchMutate中,每次循环写入数据之前都会进行checkResources的操作,该操作可能会导致本次地写入操作被阻塞。

/*
 * Check if resources to support an update.
 * 
 * Here we synchronize on HRegion, a broad scoped lock. Its appropriate
 * given we're figuring in here whether this region is able to take on
 * writes. This is only method with a synchronize (at time of writing), this
 * and the synchronize on 'this' inside in internalFlushCache to send the
 * notify.
 */
private void checkResources() throws RegionTooBusyException,
        InterruptedIOException {
    // If catalog region, do not impose resource constraints or block
    // updates.
    if (this.getRegionInfo().isMetaRegion()) {
        return;
    }

    boolean blocked = false;

    long startTime = 0;

    while (this.memstoreSize.get() > this.blockingMemStoreSize) {
        requestFlush();

        if (!blocked) {
            startTime = EnvironmentEdgeManager.currentTimeMillis();

            LOG.info("Blocking updates for '"
                    + Thread.currentThread().getName()
                    + "' on region "
                    + Bytes.toStringBinary(getRegionName())
                    + ": memstore size "
                    + StringUtils.humanReadableInt(this.memstoreSize.get())
                    + " is >= than blocking "
                    + StringUtils
                                .humanReadableInt(this.blockingMemStoreSize)
                    + " size");
        }

        long now = EnvironmentEdgeManager.currentTimeMillis();

        long timeToWait = startTime + busyWaitDuration - now;

        if (timeToWait <= 0L) {
            final long totalTime = now - startTime;

            this.updatesBlockedMs.add(totalTime);

            LOG.info("Failed to unblock updates for region " + this + " '"
                    + Thread.currentThread().getName() + "' in "
                    + totalTime + "ms. The region is still busy.");

            throw new RegionTooBusyException("region is flushing");
        }

        blocked = true;

        synchronized (this) {
            try {
                wait(Math.min(timeToWait, threadWakeFrequency));
            } catch (InterruptedException ie) {
                final long totalTime = EnvironmentEdgeManager
                        .currentTimeMillis() - startTime;

                if (totalTime > 0) {
                    this.updatesBlockedMs.add(totalTime);
                }

                LOG.info("Interrupted while waiting to unblock updates for region "
                        + this
                        + " '"
                        + Thread.currentThread().getName()
                        + "'");

                InterruptedIOException iie = new InterruptedIOException();

                iie.initCause(ie);

                throw iie;
            }
        }
    }

    if (blocked) {
        // Add in the blocked time if appropriate
        final long totalTime = EnvironmentEdgeManager.currentTimeMillis()
                - startTime;

        if (totalTime > 0) {
            this.updatesBlockedMs.add(totalTime);
        }

        LOG.info("Unblocking updates for region " + this + " '"
                + Thread.currentThread().getName() + "'");
    }
}

由上述代码可知，阻塞条件为

this.memstoreSize.get() > this.blockingMemStoreSize

如果上述条件成立，本次写入操作会被阻塞直到该Region MemStore的内存消耗值低于要求值为止。

其中，memstoreSize表示即将被写入数据的Region的MemStore的当前大小，blockingMemStoreSize由下述代码计算而来：

long flushSize = this.htableDescriptor.getMemStoreFlushSize();

if (flushSize <= 0) {
    flushSize = conf.getLong(HConstants.HREGION_MEMSTORE_FLUSH_SIZE,
            HTableDescriptor.DEFAULT_MEMSTORE_FLUSH_SIZE);
}

this.memstoreFlushSize = flushSize;

this.blockingMemStoreSize = this.memstoreFlushSize
        * conf.getLong("hbase.hregion.memstore.block.multiplier", 2);

可以看出，blockingMemStoreSize为memstoreFlushSize的整数倍。

MemStoreFlusher flushRegion

当MemStoreFlusher线程在Flush队列中取出要进行Flush操作的请求元素（FlushRegionEntry）时，都是通过下面的方法来完成Flush的。

/*
 * A flushRegion that checks store file count. If too many, puts the flush
 * on delay queue to retry later.
 * 
 * @param fqe
 * 
 * @return true if the region was successfully flushed, false otherwise. If
 * false, there will be accompanying log messages explaining why the log was
 * not flushed.
 */
private boolean flushRegion(final FlushRegionEntry fqe) {
    HRegion region = fqe.region;

    if (!fqe.region.getRegionInfo().isMetaRegion()
            && isTooManyStoreFiles(region)) {
        if (fqe.isMaximumWait(this.blockingWaitTime)) {
            LOG.info("Waited "
                    + (System.currentTimeMillis() - fqe.createTime)
                    + "ms on a compaction to clean up 'too many store files'; waited "
                    + "long enough... proceeding with flush of "
                    + region.getRegionNameAsString());
        } else {
            // If this is first time we've been put off, then emit a log
            // message.
            if (fqe.getRequeueCount() <= 0) {
                // Note: We don't impose blockingStoreFiles constraint on
                // meta regions
                LOG.warn("Region " + region.getRegionNameAsString()
                        + " has too many "
                        + "store files; delaying flush up to "
                        + this.blockingWaitTime + "ms");

                if (!this.server.compactSplitThread.requestSplit(region)) {
                    try {
                        this.server.compactSplitThread.requestCompaction(
                                region, getName());
                    } catch (IOException e) {
                            LOG.error(
                                "Cache flush failed"
                                        + (region != null ? (" for region " + Bytes
                                                    .toStringBinary(region
                                                            .getRegionName()))
                                                    : ""),
                                    RemoteExceptionHandler.checkIOException(e));
                        }
                    }
                }

            // Put back on the queue. Have it come back out of the queue
            // after a delay of this.blockingWaitTime / 100 ms.
            this.flushQueue.add(fqe.requeue(this.blockingWaitTime / 100));

            // Tell a lie, it's not flushed but it's ok
            return true;
        }
    }

    return flushRegion(region, false);
}

上述代码根据具体情况，可能会在执行具体的flushRegion操作之前，采取一些特殊的动作。

如果当前Region所属的表是用户表，且该Region中包含过多的StoreFile，则会下述判断：

（1）该Flush请求已达到最大等待时间，认为此时必须进行处理，仅仅打印一些信息即可（因此请求队列的实现为一个DealyedQueue，每一个队列元素都会根据自己的“过期时间”进行排序）；

（2）该Flush请求尚未达到最大等待时间，认为因为该Region已经包含过多的StoreFile，应该延迟本次的Flush请求，而且在延迟操作之前，如果是第一次被延迟，则会根据情况判断是否发起Split或Compact请求；