HBase源码分析之HRegion上compact流程分析（二）

2016年03月03日 21:38:04 辰辰爸的技术博客阅读数：2767

继《HBase源码分析之HRegion上compact流程分析（一）》一文后，我们继续HRegion上compact流程分析，接下来要讲的是针对表中某个列簇下文件的合并，即HStore的compact()方法，代码如下：

/**
* Compact the StoreFiles. This method may take some time, so the calling
* thread must be able to block for long periods.
*
* 合并存储文件。该方法可能花费一些时间，
*
* During this time, the Store can work as usual, getting values from
* StoreFiles and writing new StoreFiles from the memstore.
* 在此期间，Store仍能像往常一样工作，从StoreFiles获取数据和从memstore写入新的StoreFiles
*
* Existing StoreFiles are not destroyed until the new compacted StoreFile is
* completely written-out to disk.
*
* The compactLock prevents multiple simultaneous compactions.
* The structureLock prevents us from interfering with other write operations.
*
* We don't want to hold the structureLock for the whole time, as a compact()
* can be lengthy and we want to allow cache-flushes during this period.
*
* Compaction event should be idempotent, since there is no IO Fencing for
* the region directory in hdfs. A region server might still try to complete the
* compaction after it lost the region. That is why the following events are carefully
* ordered for a compaction:
* 1. Compaction writes new files under region/.tmp directory (compaction output)
* 2. Compaction atomically moves the temporary file under region directory
* 3. Compaction appends a WAL edit containing the compaction input and output files.
* Forces sync on WAL.
* 4. Compaction deletes the input files from the region directory.
*
* Failure conditions are handled like this:
* - If RS fails before 2, compaction wont complete. Even if RS lives on and finishes
* the compaction later, it will only write the new data file to the region directory.
* Since we already have this data, this will be idempotent but we will have a redundant
* copy of the data.
* - If RS fails between 2 and 3, the region will have a redundant copy of the data. The
* RS that failed won't be able to finish snyc() for WAL because of lease recovery in WAL.
* - If RS fails after 3, the region region server who opens the region will pick up the
* the compaction marker from the WAL and replay it by removing the compaction input files.
* Failed RS can also attempt to delete those files, but the operation will be idempotent
*
* See HBASE-2231 for details.
*
* @param compaction compaction details obtained from requestCompaction()
* @throws IOException
* @return Storefile we compacted into or null if we failed or opted out early.
*/
@Override
public List<StoreFile> compact(CompactionContext compaction) throws IOException {
assert compaction != null;
List<StoreFile> sfs = null;
// 从合并上下文CompactionContext中获得合并请求CompactionRequest，即cr
CompactionRequest cr = compaction.getRequest();;
try {
// Do all sanity checking in here if we have a valid CompactionRequest
// because we need to clean up after it on the way out in a finally
// block below
//
// 获取compact开始时间compactionStartTime
long compactionStartTime = EnvironmentEdgeManager.currentTime();
// 确保合并请求request不为空，实际上getRequest已经判断并确保request不为空了，这里为什么还要再做判断和保证呢？先留个小小的疑问吧！
assert compaction.hasSelection();
// 从合并请求cr中获得需要合并的文件集合filesToCompact，集合中存储的都是存储文件StoreFile的实例
// 这个文件集合是在构造CompactionRequest请求，或者合并其他请求时，根据传入的参数或者其他请求中附带的文件集合来确定的，
// 即请求一旦生成，需要合并的文件集合filesToCompact就会存在
Collection<StoreFile> filesToCompact = cr.getFiles();
// 确保需要合并的文件集合filesToCompact不为空
assert !filesToCompact.isEmpty();
// 确保filesCompacting中包含所有的待合并文件filesToCompact
synchronized (filesCompacting) {
// sanity check: we're compacting files that this store knows about
// TODO: change this to LOG.error() after more debugging
Preconditions.checkArgument(filesCompacting.containsAll(filesToCompact));
}
// Ready to go. Have list of files to compact.
LOG.info("Starting compaction of " + filesToCompact.size() + " file(s) in "
+ this + " of " + this.getRegionInfo().getRegionNameAsString()
+ " into tmpdir=" + fs.getTempDir() + ", totalSize="
+ StringUtils.humanReadableInt(cr.getSize()));
// Commence the compaction.
// 开始合并，调用CompactionContext的compact()方法，获得合并后的新文件newFiles
List<Path> newFiles = compaction.compact();
// TODO: get rid of this!
// 根据参数hbase.hstore.compaction.complete确实是否要完整的完成compact
// 这里有意思，这么处理意味着，新旧文件同时存在，新文件没有被挪到指定位置且新文件的Reader被关闭，对外提供服务的还是旧文件，啥目的呢？快速应用于读？
if (!this.conf.getBoolean("hbase.hstore.compaction.complete", true)) {
LOG.warn("hbase.hstore.compaction.complete is set to false");
// 创建StoreFile列表sfs，大小为newFiles的大小
sfs = new ArrayList<StoreFile>(newFiles.size());
// 遍历新产生的合并后的文件newFiles，针对每个文件创建StoreFile和Reader，关闭StoreFile上的Reader，
// 并将创建的StoreFile添加至列表sfs
for (Path newFile : newFiles) {
// Create storefile around what we wrote with a reader on it.
StoreFile sf = createStoreFileAndReader(newFile);
// 关闭其上的Reader
sf.closeReader(true);
sfs.add(sf);
}
// 返回合并后的文件
return sfs;
}
// Do the steps necessary to complete the compaction.
// 执行必要的步骤以完成这个合并
// 移动已完成文件至正确的地方，创建StoreFile和Reader，返回StoreFile列表sfs
sfs = moveCompatedFilesIntoPlace(cr, newFiles);
// 在WAL中写入Compaction记录
writeCompactionWalRecord(filesToCompact, sfs);
// 替换StoreFiles：
// 1、去除掉所有的合并前，即已被合并的文件compactedFiles，将合并后的文件sfs加入到StoreFileManager的storefiles中去，
// storefiles为Store中目前全部提供服务的存储文件列表；
// 2、正在合并的文件列表filesCompacting中去除被合并的文件filesToCompact；
replaceStoreFiles(filesToCompact, sfs);
// 根据合并的类型，针对不同的计数器做累加，方便系统性能指标监控
if (cr.isMajor()) {// 如果是Major合并
// 计数器累加，包括条数和大小
majorCompactedCellsCount += getCompactionProgress().totalCompactingKVs;
majorCompactedCellsSize += getCompactionProgress().totalCompactedSize;
} else {// 如果不是Major合并
// 计数器累加，包括条数和大小
compactedCellsCount += getCompactionProgress().totalCompactingKVs;
compactedCellsSize += getCompactionProgress().totalCompactedSize;
}
// At this point the store will use new files for all new scanners.
// 至此，store将会为所有新的scanners使用新的文件
// 完成合并：归档旧文件（在文件系统中删除已被合并的文件compactedFiles，实际上是归档操作，将旧的文件从原位置移到归档目录下），关闭其上的Reader，并更新store大小
completeCompaction(filesToCompact, true); // Archive old files & update store size.
// 记录日志信息
logCompactionEndMessage(cr, sfs, compactionStartTime);
// 返回StoreFile列表sfs
return sfs;
} finally {
// 完成Compaction请求：Region汇报合并请求至终端、filesCompacting中删除请求中的所有待合并文件
finishCompactionRequest(cr);
}
}

下面，我们来概述下整个流程：

1、首先，从合并上下文CompactionContext中获得合并请求CompactionRequest，即cr；

2、获取compact开始时间compactionStartTime；

3、确保合并请求request不为空：

实际上getRequest已经判断并确保request不为空了，这里为什么还要再做判断和保证呢？先留个小小的疑问吧！

4、从合并请求cr中获得需要合并的文件集合filesToCompact：

集合中存储的都是存储文件StoreFile的实例，这个文件集合是在构造CompactionRequest请求，或者合并其他请求时，根据传入的参数或者其他请求中附带的文件集合来确定的，即请求一旦生成，需要合并的文件集合filesToCompact就会存在。

5、确保需要合并的文件集合filesToCompact不为空；

6、确保filesCompacting中包含所有的待合并文件filesToCompact：

那么这个filesCompacting中的文件是何时添加的呢？

7、开始合并，调用CompactionContext的compact()方法，获得合并后的新文件newFiles：

这一步是核心流程，它会持有通过scanner访问待合并文件，然后将数据全部写入新文件，后续文章会着重分析。

8、根据参数hbase.hstore.compaction.complete确实是否要完整的完成compact，默认为true：

8.1、如果配置的是false，则：

8.1.1、创建StoreFile列表sfs，大小为newFiles的大小；

8.1.2、遍历新产生的合并后的文件newFiles，针对每个文件创建StoreFile和Reader，关闭StoreFile上的Reader，并将创建的StoreFile添加至列表sfs；

8.1.3、返回合并后的文件列表sfs；

8.2、如果配置的是true，则：

8.2.1、移动已完成文件至正确的地方，创建StoreFile和Reader，返回StoreFile列表sfs；

8.2.2、在WAL中写入Compaction记录；

8.2.3、替换StoreFiles：包括去除掉所有的合并前，即已被合并的文件compactedFiles，将合并后的文件sfs加入到StoreFileManager的storefiles中去，storefiles为Store中目前全部提供服务的存储文件列表，还有正在合并的文件列表filesCompacting中去除被合并的文件filesToCompact。

8.2.4、根据合并的类型，针对不同的计数器做累加，方便系统性能指标监控；

8.2.5、完成合并：归档旧文件（在文件系统中删除已被合并的文件compactedFiles，实际上是归档操作，将旧的文件从原位置移到归档目录下），关闭其上的Reader，并更新store大小；

8.2.6、记录日志信息；

8.2.7、完成Compaction请求：Region汇报合并请求至终端、filesCompacting中删除请求中的所有待合并文件；

8.2.8、返回StoreFile列表sfs。

至此，整个流程详述完毕。接下来，我们针对其中的部分细节，再做详细描述。

首先，真正执行合并的CompactionContext的compact()方法我们暂时不讲，只需要知道它会持有通过scanner访问待合并文件，然后将数据全部写入新文件，并得到这些新文件的集合newFiles即可，我们会在后续文章详细介绍。

接下来，在获得合并后的新文件newFiles之后，我们会根据一个参数来确定后续处理流程，这个参数就是hbase.hstore.compaction.complete，由它来确定是否完整的结束一次合并操作，这完整与非完整的主要区别，或者说实质性区别就是：由谁来继续对外提供数据读取服务。

先来看下非完整性结束，它会为合并后的每个文件创建StoreFile和Reader实例，同时关闭新文件上的Reader，也就意味着扔继续由旧文件提供数据读取服务，而新文件与旧文件同时存在，旧文件位置不变，涉及到列簇CF下的目前所有可用storefiles列表不变，存储的仍是旧文件的StoreFile对象；

而对于完整性结束来说，它会移动已完成文件至正确的地方，创建StoreFile和Reader，返回StoreFile列表sfs，然后在WAL中写入Compaction记录，并替换掉storefiles，根据合并的类型，针对不同的计数器做累加，方便系统性能指标监控，归档旧文件（在文件系统中删除已被合并的文件compactedFiles，实际上是归档操作，将旧的文件从原位置移到归档目录下），关闭其上的Reader，并更新store大小，完成Compaction请求：Region汇报合并请求至终端、filesCompacting中删除请求中的所有待合并文件等等，很多复杂的操作。不要着急，我们就其中复杂的地方，一个个的解释：

1、移动已完成文件至正确的地方，创建StoreFile和Reader，返回StoreFile列表sfs

这个是通过moveCompatedFilesIntoPlace()方法实现的，代码如下：

private List<StoreFile> moveCompatedFilesIntoPlace(
CompactionRequest cr, List<Path> newFiles) throws IOException {
// 创建StoreFile列表sfs
List<StoreFile> sfs = new ArrayList<StoreFile>(newFiles.size());
// 遍历newFiles
for (Path newFile : newFiles) {
assert newFile != null;
// 将新文件newFile挪至正确地点，并创建StoreFile和Reader
StoreFile sf = moveFileIntoPlace(newFile);
if (this.getCoprocessorHost() != null) {
this.getCoprocessorHost().postCompact(this, sf, cr);
}
assert sf != null;
sfs.add(sf);
}
return sfs;
}

首先呢，创建StoreFile列表sfs，遍历合并后的文件newFiles，将新文件newFile挪至正确地点，并创建StoreFile和Reader。而文件位置改变，则是通过moveFileIntoPlace()方法实现的，它的代码如下：

// Package-visible for tests
StoreFile moveFileIntoPlace(final Path newFile) throws IOException {
// 检测新文件
validateStoreFile(newFile);
// Move the file into the right spot
// 移动文件至正确的地点
Path destPath = fs.commitStoreFile(getColumnFamilyName(), newFile);
// 创建StoreFile和Reader
return createStoreFileAndReader(destPath);
}

我们发现，移动文件实际上是通过HStore的成员变量fs的commitStoreFile()方法来完成的。这个fs是HRegionFileSystem类型的变量，HRegionFileSystem是HRegion上文件系统的一个抽象，它实现了各种文件等的实际物理操作。我们来看下它的commitStoreFile()方法:

/**
* Move the file from a build/temp location to the main family store directory.
* @param familyName Family that will gain the file
* @param buildPath {@link Path} to the file to commit.
* @param seqNum Sequence Number to append to the file name (less then 0 if no sequence number)
* @param generateNewName False if you want to keep the buildPath name
* @return The new {@link Path} of the committed file
* @throws IOException
*/
private Path commitStoreFile(final String familyName, final Path buildPath,
final long seqNum, final boolean generateNewName) throws IOException {
// 根据列簇名familyName获取存储路径storeDir
Path storeDir = getStoreDir(familyName);
// 如果在文件系统fs中不存在路径的情况下创建它时失败则抛出异常
if(!fs.exists(storeDir) && !createDir(storeDir))
throw new IOException("Failed creating " + storeDir);
String name = buildPath.getName();
if (generateNewName) {
name = generateUniqueName((seqNum < 0) ? null : "_SeqId_" + seqNum + "_");
}
Path dstPath = new Path(storeDir, name);
if (!fs.exists(buildPath)) {
throw new FileNotFoundException(buildPath.toString());
}
LOG.debug("Committing store file " + buildPath + " as " + dstPath);
// buildPath exists, therefore not doing an exists() check.
if (!rename(buildPath, dstPath)) {
throw new IOException("Failed rename of " + buildPath + " to " + dstPath);
}
return dstPath;
}

非常简单，根据列簇名familyName获取存储路径storeDir，检测并在必要时创建storeDir，根据buildPath来获取文件名name，然后利用storeDir和name来构造目标路径storeDir，通过rename()方法实现文件从buildPath至dstPath的移动即可。

而创建StoreFile和Reader的方法最终调用的是createStoreFileAndReader()方法，代码如下：

private StoreFile createStoreFileAndReader(final StoreFileInfo info)
throws IOException {
info.setRegionCoprocessorHost(this.region.getCoprocessorHost());
StoreFile storeFile = new StoreFile(this.getFileSystem(), info, this.conf, this.cacheConf,
this.family.getBloomFilterType());
storeFile.createReader();
return storeFile;
}

StoreFile是一个存储数据文件。Stores通常含有一个或多个StoreFile，而Reader是其内部类，由Reader来提供文件数据的读取服务。

2、在WAL中写入Compaction记录

这个过程是通过writeCompactionWalRecord()方法来完成的，代码如下：

/**
* Writes the compaction WAL record.
* 在WAL中写入合并记录
*
* @param filesCompacted Files compacted (input).
* @param newFiles Files from compaction.
*/
private void writeCompactionWalRecord(Collection<StoreFile> filesCompacted,
Collection<StoreFile> newFiles) throws IOException {
// 如果region中的WAL为空，则直接返回
if (region.getWAL() == null) return;
// 将被合并的文件路径添加至inputPaths列表
List<Path> inputPaths = new ArrayList<Path>(filesCompacted.size());
for (StoreFile f : filesCompacted) {
inputPaths.add(f.getPath());
}
// 将合并后的文件路径添加至inputPaths列表
List<Path> outputPaths = new ArrayList<Path>(newFiles.size());
for (StoreFile f : newFiles) {
outputPaths.add(f.getPath());
}
// 获取HRegionInfo，即info
HRegionInfo info = this.region.getRegionInfo();
// 构造compaction的描述信息CompactionDescriptor
CompactionDescriptor compactionDescriptor = ProtobufUtil.toCompactionDescriptor(info,
family.getName(), inputPaths, outputPaths, fs.getStoreDir(getFamily().getNameAsString()));
// 利用WALUtil工具类的writeCompactionMarker()方法，在WAL中写入一个合并标记
WALUtil.writeCompactionMarker(region.getWAL(), this.region.getTableDesc(),
this.region.getRegionInfo(), compactionDescriptor, this.region.getSequenceId());
}

逻辑比较简单：

1、将被合并的文件路径添加至inputPaths列表；

2、将合并后的文件路径添加至outputPaths列表；

3、获取HRegionInfo，即info；

4、构造compaction的描述信息CompactionDescriptor；

5、利用WALUtil工具类的writeCompactionMarker()方法，在WAL中写入一个合并标记。

首先说下这个compaction的描述信息CompactionDescriptor，其中包含了表名TableName、Region名EncodedRegionName、列簇名FamilyName、存储Home路径StoreHomeDir、合并的输入CompactionInput、合并的输出CompactionOutput等关键信息，完整的描述了合并的全部详细信息。其构造代码如下：

public static CompactionDescriptor toCompactionDescriptor(HRegionInfo info, byte[] family,
List<Path> inputPaths, List<Path> outputPaths, Path storeDir) {
// compaction descriptor contains relative paths.
// input / output paths are relative to the store dir
// store dir is relative to region dir
CompactionDescriptor.Builder builder = CompactionDescriptor.newBuilder()
.setTableName(ByteStringer.wrap(info.getTableName()))
.setEncodedRegionName(ByteStringer.wrap(info.getEncodedNameAsBytes()))
.setFamilyName(ByteStringer.wrap(family))
.setStoreHomeDir(storeDir.getName()); //make relative
for (Path inputPath : inputPaths) {
builder.addCompactionInput(inputPath.getName()); //relative path
}
for (Path outputPath : outputPaths) {
builder.addCompactionOutput(outputPath.getName());
}
builder.setRegionName(ByteStringer.wrap(info.getRegionName()));
return builder.build();
}

最后，利用WALUtil工具类的writeCompactionMarker()方法，在WAL中写入一个合并标记，我们来看下代码：

/**
* Write the marker that a compaction has succeeded and is about to be committed.
* This provides info to the HMaster to allow it to recover the compaction if
* this regionserver dies in the middle (This part is not yet implemented). It also prevents
* the compaction from finishing if this regionserver has already lost its lease on the log.
* @param sequenceId Used by WAL to get sequence Id for the waledit.
*/
public static void writeCompactionMarker(WAL log, HTableDescriptor htd, HRegionInfo info,
final CompactionDescriptor c, AtomicLong sequenceId) throws IOException {
// 从合并信息CompactionDescriptor中获取表名tn
TableName tn = TableName.valueOf(c.getTableName().toByteArray());
// we use HLogKey here instead of WALKey directly to support legacy coprocessors.
// 根据region的名字、表明tn，创建一个WALKey
WALKey key = new HLogKey(info.getEncodedNameAsBytes(), tn);
// WAL中添加一条记录，包括表的描述信息HTableDescriptor、WALKey、Compaction信息WALEdit、序列号sequenceId
// Compaction信息WALEdit是根据WALEdit的createCompaction()方法，由HRegionInfo、CompactionDescriptor获取的
//
log.append(htd, info, key, WALEdit.createCompaction(info, c), sequenceId, false, null);
// 同步日志
log.sync();
if (LOG.isTraceEnabled()) {
LOG.trace("Appended compaction marker " + TextFormat.shortDebugString(c));
}
}

它实际上在WAL中append了一条记录，包括表的描述信息HTableDescriptor、WALKey、Compaction信息WALEdit、序列号sequenceId，而Compaction信息WALEdit是根据WALEdit的createCompaction()方法，由HRegionInfo、CompactionDescriptor构造的。代码如下：

/**
* Create a compacion WALEdit
* @param c
* @return A WALEdit that has <code>c</code> serialized as its value
*/
public static WALEdit createCompaction(final HRegionInfo hri, final CompactionDescriptor c) {
// 将CompactionDescriptor转化成byte []
byte [] pbbytes = c.toByteArray();
// 构造KeyValue，包括Region的startKey、“METAFAMILY”字符串、
// "HBASE::COMPACTION"字符串、当前时间和合并描述CompactionDescriptor的二进制形式
KeyValue kv = new KeyValue(getRowForRegion(hri), METAFAMILY, COMPACTION,
EnvironmentEdgeManager.currentTime(), pbbytes);
// 将KeyValue添加至WALEdit，并返回WALEdit实例
return new WALEdit().add(kv); //replication scope null so that this won't be replicated
}

代码注释比较详细，不再赘述。

3、替换StoreFiles，其中包括亮点：

（1）去除掉所有的合并前，即已被合并的文件compactedFiles，将合并后的文件sfs加入到StoreFileManager的storefiles中去，storefiles为Store中目前全部提供服务的存储文件列表；

（2）正在合并的文件列表filesCompacting中去除被合并的文件filesToCompact；

具体代码replaceStoreFiles()方法如下：

@VisibleForTesting
void replaceStoreFiles(final Collection<StoreFile> compactedFiles,
final Collection<StoreFile> result) throws IOException {
// 加锁，上读写锁ReentrantReadWriteLock的写锁，意味着这是一把互斥锁
this.lock.writeLock().lock();
try {
// 通过StoreFileManager的addCompactionResults()方法，将被合并的文件
// 去除掉所有的合并前，即已被合并的文件compactedFiles
// 将合并后的文件加入到StoreFileManager的storefiles中去，storefiles为Store中目前全部提供服务的存储文件列表
this.storeEngine.getStoreFileManager().addCompactionResults(compactedFiles, result);
// 正在合并的文件列表filesCompacting中去除被合并的文件
filesCompacting.removeAll(compactedFiles); // safe bc: lock.writeLock();
} finally {
// 解锁
this.lock.writeLock().unlock();
}
}

4、完成合并：归档旧文件（在文件系统中删除已被合并的文件compactedFiles，实际上是归档操作，将旧的文件从原位置移到归档目录下），关闭其上的Reader，并更新store大小。completeCompaction()代码如下：

/**
* It works by processing a compaction that's been written to disk.
*
* It is usually invoked at the end of a compaction, but might also be
* invoked at HStore startup, if the prior execution died midway through.
*
* Moving the compacted TreeMap into place means:
* <pre>
* 1) Unload all replaced StoreFile, close and collect list to delete.
* 2) Compute new store size
* </pre>
*
* @param compactedFiles list of files that were compacted
*/
@VisibleForTesting
protected void completeCompaction(final Collection<StoreFile> compactedFiles, boolean removeFiles)
throws IOException {
try {
// Do not delete old store files until we have sent out notification of
// change in case old files are still being accessed by outstanding scanners.
// Don't do this under writeLock; see HBASE-4485 for a possible deadlock
// scenario that could have happened if continue to hold the lock.
// 通知Reader观察者
notifyChangedReadersObservers();
// At this point the store will use new files for all scanners.
// let the archive util decide if we should archive or delete the files
LOG.debug("Removing store files after compaction...");
// 遍历已被合并的文件completeCompaction，关闭其上的Reader
for (StoreFile compactedFile : compactedFiles) {
compactedFile.closeReader(true);
}
// 在文件系统中删除已被合并的文件compactedFiles，实际上是归档操作，将旧的文件从原位置移到归档目录下
if (removeFiles) {
this.fs.removeStoreFiles(this.getColumnFamilyName(), compactedFiles);
}
} catch (IOException e) {
e = RemoteExceptionHandler.checkIOException(e);
LOG.error("Failed removing compacted files in " + this +
". Files we were trying to remove are " + compactedFiles.toString() +
"; some of them may have been already removed", e);
}
// 4. Compute new store size
// 计算新的store大小
this.storeSize = 0L;
this.totalUncompressedBytes = 0L;
// 遍历StoreFiles，计算storeSize、totalUncompressedBytes等大小
for (StoreFile hsf : this.storeEngine.getStoreFileManager().getStorefiles()) {
StoreFile.Reader r = hsf.getReader();
if (r == null) {
LOG.warn("StoreFile " + hsf + " has a null Reader");
continue;
}
this.storeSize += r.length();
this.totalUncompressedBytes += r.getTotalUncompressedBytes();
}
}

其他代码注释中都有，这里，我们要单独说下HRegionFileSystem的removeStoreFiles()方法，如下：

/**
* Closes and archives the specified store files from the specified family.
* @param familyName Family that contains the store files
* @param storeFiles set of store files to remove
* @throws IOException if the archiving fails
*/
public void removeStoreFiles(final String familyName, final Collection<StoreFile> storeFiles)
throws IOException {
HFileArchiver.archiveStoreFiles(this.conf, this.fs, this.regionInfoForFs,
this.tableDir, Bytes.toBytes(familyName), storeFiles);
}

它最终是通过HFileArchiver的archiveStoreFiles()方法来完成的，代码如下：

/**
* Remove the store files, either by archiving them or outright deletion
* @param conf {@link Configuration} to examine to determine the archive directory
* @param fs the filesystem where the store files live
* @param regionInfo {@link HRegionInfo} of the region hosting the store files
* @param family the family hosting the store files
* @param compactedFiles files to be disposed of. No further reading of these files should be
* attempted; otherwise likely to cause an {@link IOException}
* @throws IOException if the files could not be correctly disposed.
*/
public static void archiveStoreFiles(Configuration conf, FileSystem fs, HRegionInfo regionInfo,
Path tableDir, byte[] family, Collection<StoreFile> compactedFiles) throws IOException {
// sometimes in testing, we don't have rss, so we need to check for that
if (fs == null) {
LOG.warn("Passed filesystem is null, so just deleting the files without archiving for region:"
+ Bytes.toString(regionInfo.getRegionName()) + ", family:" + Bytes.toString(family));
deleteStoreFilesWithoutArchiving(compactedFiles);
return;
}
// short circuit if we don't have any files to delete
// 判断被合并文件列表compactedFiles的大小，如果为0，立即返回
if (compactedFiles.size() == 0) {
LOG.debug("No store files to dispose, done!");
return;
}
// build the archive path
if (regionInfo == null || family == null) throw new IOException(
"Need to have a region and a family to archive from.");
// 获取归档存储路径
Path storeArchiveDir = HFileArchiveUtil.getStoreArchivePath(conf, regionInfo, tableDir, family);
// make sure we don't archive if we can't and that the archive dir exists
// 创建路径
if (!fs.mkdirs(storeArchiveDir)) {
throw new IOException("Could not make archive directory (" + storeArchiveDir + ") for store:"
+ Bytes.toString(family) + ", deleting compacted files instead.");
}
// otherwise we attempt to archive the store files
if (LOG.isDebugEnabled()) LOG.debug("Archiving compacted store files.");
// Wrap the storefile into a File
StoreToFile getStorePath = new StoreToFile(fs);
Collection<File> storeFiles = Collections2.transform(compactedFiles, getStorePath);
// do the actual archive
// 通过resolveAndArchive()执行归档
if (!resolveAndArchive(fs, storeArchiveDir, storeFiles)) {
throw new IOException("Failed to archive/delete all the files for region:"
+ Bytes.toString(regionInfo.getRegionName()) + ", family:" + Bytes.toString(family)
+ " into " + storeArchiveDir + ". Something is probably awry on the filesystem.");
}
}

层层调用啊，接着来吧，继续看关键代码：

// 如果是文件
if (file.isFile()) {
// attempt to archive the file
if (!resolveAndArchiveFile(baseArchiveDir, file, startTime)) {
LOG.warn("Couldn't archive " + file + " into backup directory: " + baseArchiveDir);
failures.add(file);
}
}

而这个resolveAndArchiveFile()方法不是简单的删除文件，而是通过rename()方法将旧的存储文件挪至了归档路径下，代码如下：

// move the archive file to the stamped backup
Path backedupArchiveFile = new Path(archiveDir, filename + SEPARATOR + archiveStartTime);
if (!fs.rename(archiveFile, backedupArchiveFile)) {
LOG.error("Could not rename archive file to backup: " + backedupArchiveFile
+ ", deleting existing file in favor of newer.");
// try to delete the exisiting file, if we can't rename it
if (!fs.delete(archiveFile, false)) {
throw new IOException("Couldn't delete existing archive file (" + archiveFile
+ ") or rename it to the backup file (" + backedupArchiveFile
+ ") to make room for similarly named file.");
}
}

5、完成Compaction请求：Region汇报合并请求至终端、filesCompacting中删除请求中的所有待合并文件

这部分是由方法finishCompactionRequest()完成的，代码如下：

private void finishCompactionRequest(CompactionRequest cr) {
// Region汇报合并请求至终端
this.region.reportCompactionRequestEnd(cr.isMajor(), cr.getFiles().size(), cr.getSize());
//
if (cr.isOffPeak()) {
offPeakCompactionTracker.set(false);
cr.setOffPeak(false);
}
// filesCompacting中删除请求中的所有待合并文件
synchronized (filesCompacting) {
filesCompacting.removeAll(cr.getFiles());
}
}

读者可自行分析，不再赘述。

好了，就先到这里吧，且待下回分解！

HBase源码分析之HRegion上compact流程分析（二）

猜你喜欢