【Tinkerpop整理】以Spark为引擎全量检索图库数据流程源码解析-part3

【Tinkerpop整理】以Spark为引擎全量检索图库数据流程源码解析-part1
【Tinkerpop整理】以Spark为引擎全量检索图库数据流程源码解析-part2
【Tinkerpop整理】以Spark为引擎全量检索图库数据流程源码解析-part3

11:Backend

/**
 * These are the names for the edge store and property index databases, respectively.
 * The edge store contains all edges and properties. The property index contains an
 * inverted index from attribute value to vertex.
 * <p>
 * These names are fixed and should NEVER be changed. Changing these strings can
 * disrupt storage adapters that rely on these names for specific configurations.
 * In the past, the store name for the ID table, janusgraph_ids, was also marked here,
 * but to clear the upgrade path from Titan to JanusGraph, we had to pull it into
 * configuration.
 */
public static final String EDGESTORE_NAME = "edgestore";
public static final String INDEXSTORE_NAME = "graphindex";

public static final String METRICS_STOREMANAGER_NAME = "storeManager";
public static final String METRICS_MERGED_STORE = "stores";
public static final String METRICS_MERGED_CACHE = "caches";
public static final String METRICS_CACHE_SUFFIX = ".cache";
public static final String LOCK_STORE_SUFFIX = "_lock_";

public static final String SYSTEM_TX_LOG_NAME = "txlog";
public static final String SYSTEM_MGMT_LOG_NAME = "systemlog";

public static final double EDGESTORE_CACHE_PERCENT = 0.8;
public static final double INDEXSTORE_CACHE_PERCENT = 0.2;

private static final long ETERNAL_CACHE_EXPIRATION = 1000L *3600*24*365*200; //200 years

public static final int THREAD_POOL_SIZE_SCALE_FACTOR = 2;

private final KeyColumnValueStoreManager storeManager;
private final KeyColumnValueStoreManager storeManagerLocking;
private final StoreFeatures storeFeatures;

private KCVSCache edgeStore;
private KCVSCache indexStore;
private KCVSCache txLogStore;
private IDAuthority idAuthority;
private KCVSConfiguration systemConfig;
private KCVSConfiguration userConfig;
private boolean hasAttemptedClose;

private final StandardScanner scanner;

private final KCVSLogManager managementLogManager;
private final KCVSLogManager txLogManager;
private final LogManager userLogManager;


private final Map<String, IndexProvider> indexes;

private final int bufferSize;
private final Duration maxWriteTime;
private final Duration maxReadTime;
private final boolean cacheEnabled;
private final ExecutorService threadPool;

private final Function<String, Locker> lockerCreator;
private final ConcurrentHashMap<String, Locker> lockers = new ConcurrentHashMap<>();

private final Configuration configuration;

public Backend(Configuration configuration) {
    this.configuration = configuration;
    //获取
    //此处获取的对象是：HbaseStoreManeger，数据存储管理对象
    //KeyColumnValueStoreManager provides the persistence context to the graph database storage backend.
    KeyColumnValueStoreManager manager = getStorageManager(configuration);
    if (configuration.get(BASIC_METRICS)) {
        storeManager = new MetricInstrumentedStoreManager(manager,METRICS_STOREMANAGER_NAME,configuration.get(METRICS_MERGE_STORES),METRICS_MERGED_STORE);
    } else {
        storeManager = manager;
    }
    //获取索引对象，如果是es对象，则获取org.janusgraph.diskstorage.es.ElasticSearchIndex 对象；
    //如果该ES开启kerberos认证，则会出现认证不过的偶现问题；
    indexes = getIndexes(configuration);
    storeFeatures = storeManager.getFeatures();

    managementLogManager = getKCVSLogManager(MANAGEMENT_LOG);
    txLogManager = getKCVSLogManager(TRANSACTION_LOG);
    userLogManager = getLogManager(USER_LOG);


    cacheEnabled = !configuration.get(STORAGE_BATCH) && configuration.get(DB_CACHE);

    int bufferSizeTmp = configuration.get(BUFFER_SIZE);
    Preconditions.checkArgument(bufferSizeTmp > 0, "Buffer size must be positive");
    if (!storeFeatures.hasBatchMutation()) {
        bufferSize = Integer.MAX_VALUE;
    } else bufferSize = bufferSizeTmp;

    maxWriteTime = configuration.get(STORAGE_WRITE_WAITTIME);
    maxReadTime = configuration.get(STORAGE_READ_WAITTIME);

    if (!storeFeatures.hasLocking()) {
        Preconditions.checkArgument(storeFeatures.isKeyConsistent(),"Store needs to support some form of locking");
        storeManagerLocking = new ExpectedValueCheckingStoreManager(storeManager,LOCK_STORE_SUFFIX,this,maxReadTime);
    } else {
        storeManagerLocking = storeManager;
    }

    if (configuration.get(PARALLEL_BACKEND_OPS)) {
        int poolSize = Runtime.getRuntime().availableProcessors() * THREAD_POOL_SIZE_SCALE_FACTOR;
        threadPool = Executors.newFixedThreadPool(poolSize);
        log.info("Initiated backend operations thread pool of size {}", poolSize);
    } else {
        threadPool = null;
    }

    final String lockBackendName = configuration.get(LOCK_BACKEND);
    if (REGISTERED_LOCKERS.containsKey(lockBackendName)) {
        lockerCreator = REGISTERED_LOCKERS.get(lockBackendName);
    } else {
        throw new JanusGraphConfigurationException("Unknown lock backend \"" +
                lockBackendName + "\".  Known lock backends: " +
                Joiner.on(", ").join(REGISTERED_LOCKERS.keySet()) + ".");
    }
    // Never used for backends that have innate transaction support, but we
    // want to maintain the non-null invariant regardless; it will default
    // to consistent-key implementation if none is specified
    Preconditions.checkNotNull(lockerCreator);

    scanner = new StandardScanner(storeManager);
}

12，BypassMergeSortShuffleWriter

@Override
public void write(Iterator<Product2<K, V>> records) throws IOException {
  assert (partitionWriters == null);
  //判断是否有数据
  if (!records.hasNext()) {
    partitionLengths = new long[numPartitions];
    shuffleBlockResolver.writeIndexFileAndCommit(shuffleId, mapId, partitionLengths, null);
    mapStatus = MapStatus$.MODULE$.apply(blockManager.shuffleServerId(), partitionLengths);
    return;
  }
  final SerializerInstance serInstance = serializer.newInstance();
  final long openStartTime = System.nanoTime();
  //创建磁盘写入对象
  partitionWriters = new DiskBlockObjectWriter[numPartitions];
  //创建文件块写入对象；每个 ShuffleBlock 被称为 FileSegment，存储shuffle产生的数据
  partitionWriterSegments = new FileSegment[numPartitions];
  for (int i = 0; i < numPartitions; i++) {
    final Tuple2<TempShuffleBlockId, File> tempShuffleBlockIdPlusFile =
      blockManager.diskBlockManager().createTempShuffleBlock();
      //获取数据写入文件路径：本地磁盘
    final File file = tempShuffleBlockIdPlusFile._2();
    //获取block块id，UUID保证每个block块的唯一性
    final BlockId blockId = tempShuffleBlockIdPlusFile._1();
    partitionWriters[i] =
      blockManager.getDiskWriter(blockId, file, serInstance, fileBufferSize, writeMetrics);
  }
  // Creating the file to write to and creating a disk writer both involve interacting with
  // the disk, and can take a long time in aggregate when we open many files, so should be
  // included in the shuffle write time.
  writeMetrics.incWriteTime(System.nanoTime() - openStartTime);

  while (records.hasNext()) {
      //遍历待写入的每一条数据；此处：[v[id],v[]]
    final Product2<K, V> record = records.next();
    final K key = record._1();
    partitionWriters[partitioner.getPartition(key)].write(key, record._2());
  }

  for (int i = 0; i < numPartitions; i++) {
    final DiskBlockObjectWriter writer = partitionWriters[i];
    partitionWriterSegments[i] = writer.commitAndGet();
    writer.close();
  }
    //写入到磁盘文件路径：(e.g:/data/5/hadoop/yarn/local/usercache/graph/appcache/)
  File output = shuffleBlockResolver.getDataFile(shuffleId, mapId);
  File tmp = Utils.tempFileWith(output);
  try {
      //数据写入磁盘
    partitionLengths = writePartitionedFile(tmp);
    //将block文件index写入，driver端的blockmanager；index记载文件数据，其实pos，长度：length
    //数据文件区段：文件路径，其实位置，长度
    /**
    writeIndexFileAndCommit
    Write an index file with the offsets of each block, plus a final offset at the end for the
    end of the output file. This will be used by getBlockData to figure out where each block
    begins and ends.
    */
    shuffleBlockResolver.writeIndexFileAndCommit(shuffleId, mapId, partitionLengths, tmp);
  } finally {
    if (tmp.exists() && !tmp.delete()) {
      logger.error("Error while deleting temp file {}", tmp.getAbsolutePath());
    }
  }
  mapStatus = MapStatus$.MODULE$.apply(blockManager.shuffleServerId(), partitionLengths);
}

13：DiskBlockObjectWriter

/**
 * Flush the partial writes and commit them as a single atomic block.
 * A commit may write additional bytes to frame the atomic block.
 *
 * @return file segment with previous offset and length committed on this call.
 */
def commitAndGet(): FileSegment = {
  if (streamOpen) {
    // NOTE: Because Kryo doesn't flush the underlying stream we explicitly flush both the
    //       serializer stream and the lower level stream.
    objOut.flush()
    bs.flush()
    objOut.close()
    streamOpen = false

    if (syncWrites) {
      // Force outstanding writes to disk and track how long it takes
      val start = System.nanoTime()
      fos.getFD.sync()
      writeMetrics.incWriteTime(System.nanoTime() - start)
    }
    //该数据文件中，数据的其实位置
    val pos = channel.position()
    //数据文件区段：文件路径，其实位置，长度
    val fileSegment = new FileSegment(file, committedPosition, pos - committedPosition)
    committedPosition = pos
    // In certain compression codecs, more bytes are written after streams are closed
    writeMetrics.incBytesWritten(committedPosition - reportedPosition)
    reportedPosition = committedPosition
    numRecordsWritten = 0
    fileSegment
  } else {
    new FileSegment(file, committedPosition, 0)
  }
}

14：IndexShuffleBlockResolver

/**
 * Write an index file with the offsets of each block, plus a final offset at the end for the
 * end of the output file. This will be used by getBlockData to figure out where each block
 * begins and ends.
 *
 * It will commit the data and index file as an atomic operation, use the existing ones, or
 * replace them with new ones.
 *
 * Note: the `lengths` will be updated to match the existing index file if use the existing ones.
 */
def writeIndexFileAndCommit(
    shuffleId: Int,
    mapId: Int,
    lengths: Array[Long],
    dataTmp: File): Unit = {
  val indexFile = getIndexFile(shuffleId, mapId)
  val indexTmp = Utils.tempFileWith(indexFile)
  try {
    val dataFile = getDataFile(shuffleId, mapId)
    // There is only one IndexShuffleBlockResolver per executor, this synchronization make sure
    // the following check and rename are atomic.
    synchronized {
      val existingLengths = checkIndexAndDataFile(indexFile, dataFile, lengths.length)
      if (existingLengths != null) {
        // Another attempt for the same task has already written our map outputs successfully,
        // so just use the existing partition lengths and delete our temporary map outputs.
        System.arraycopy(existingLengths, 0, lengths, 0, lengths.length)
        if (dataTmp != null && dataTmp.exists()) {
          dataTmp.delete()
        }
      } else {
        // This is the first successful attempt in writing the map outputs for this task,
        // so override any existing index and data files with the ones we wrote.
        val out = new DataOutputStream(new BufferedOutputStream(new FileOutputStream(indexTmp)))
        Utils.tryWithSafeFinally {
          // We take in lengths of each block, need to convert it to offsets.
          var offset = 0L
          out.writeLong(offset)
          for (length <- lengths) {
            offset += length
            out.writeLong(offset)
          }
        } {
          out.close()
        }

        if (indexFile.exists()) {
          indexFile.delete()
        }
        if (dataFile.exists()) {
          dataFile.delete()
        }
        if (!indexTmp.renameTo(indexFile)) {
          throw new IOException("fail to rename file " + indexTmp + " to " + indexFile)
        }
        if (dataTmp != null && dataTmp.exists() && !dataTmp.renameTo(dataFile)) {
          throw new IOException("fail to rename file " + dataTmp + " to " + dataFile)
        }
      }
    }
  } finally {
    if (indexTmp.exists() && !indexTmp.delete()) {
      logError(s"Failed to delete temporary index file at ${indexTmp.getAbsolutePath}")
    }
  }
}

15：Executor:

//执行task任务，并将task任务信息写入监控，更新任务状态
override def run(): Unit = {
    val startTimeMs = System.currentTimeMillis()
    def elapsedTimeMs = System.currentTimeMillis() - startTimeMs
    def timeoutExceeded(): Boolean = killTimeoutMs > 0 && elapsedTimeMs > killTimeoutMs
    try {
      // Only attempt to kill the task once. If interruptThread = false then a second kill
      // attempt would be a no-op and if interruptThread = true then it may not be safe or
      // effective to interrupt multiple times:
      taskRunner.kill(interruptThread = interruptThread, reason = reason)
      // Monitor the killed task until it exits. The synchronization logic here is complicated
      // because we don't want to synchronize on the taskRunner while possibly taking a thread
      // dump, but we also need to be careful to avoid races between checking whether the task
      // has finished and wait()ing for it to finish.
      var finished: Boolean = false
      while (!finished && !timeoutExceeded()) {
        taskRunner.synchronized {
          // We need to synchronize on the TaskRunner while checking whether the task has
          // finished in order to avoid a race where the task is marked as finished right after
          // we check and before we call wait().
          if (taskRunner.isFinished) {
            finished = true
          } else {
            taskRunner.wait(killPollingIntervalMs)
          }
        }
        if (taskRunner.isFinished) {
          finished = true
        } else {
          logWarning(s"Killed task $taskId is still running after $elapsedTimeMs ms")
          if (takeThreadDump) {
            try {
              Utils.getThreadDumpForThread(taskRunner.getThreadId).foreach { thread =>
                if (thread.threadName == taskRunner.threadName) {
                  logWarning(s"Thread dump from task $taskId:\n${thread.stackTrace}")
                }
              }
            } catch {
              case NonFatal(e) =>
                logWarning("Exception thrown while obtaining thread dump: ", e)
            }
          }
        }
      }

      if (!taskRunner.isFinished && timeoutExceeded()) {
        if (isLocal) {
          logError(s"Killed task $taskId could not be stopped within $killTimeoutMs ms; " +
            "not killing JVM because we are running in local mode.")
        } else {
          // In non-local-mode, the exception thrown here will bubble up to the uncaught exception
          // handler and cause the executor JVM to exit.
          throw new SparkException(
            s"Killing executor JVM because killed task $taskId could not be stopped within " +
              s"$killTimeoutMs ms.")
        }
      }
    } finally {
      // Clean up entries in the taskReaperForTask map.
      taskReaperForTask.synchronized {
        taskReaperForTask.get(taskId).foreach { taskReaperInMap =>
          if (taskReaperInMap eq this) {
            taskReaperForTask.remove(taskId)
          } else {
            // This must have been a TaskReaper where interruptThread == false where a subsequent
            // killTask() call for the same task had interruptThread == true and overwrote the
            // map entry.
          }
        }
      }
    }
  }
}