Java HStore类代码示例

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
private CompactJobQueue.CompactJob createCompactJob(final CompactionRequest request,
    final Path writtenPath, HStore store) throws IOException {
  // check reference file, not supported yet!
  boolean needToRebuild = false;
  for (StoreFile sf : request.getFiles()) {
    if (sf.getPath().getName().indexOf(".") != -1 || sf.isReference()) {
      needToRebuild = true;
      break;
    }
  }
  CompactJobQueue.CompactJob job;
  if (needToRebuild) {
    job = new CompactJobQueue.RebuildCompactJob(store, request, writtenPath);
  } else {
    job = new CompactJobQueue.NormalCompactJob(store, request, writtenPath);
  }
  return job;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
public LMDIndexWriter(HStore store, Path rawDataPath, TimeRangeTracker timeRangeTracker,
    String opType) {
  tableRelation = store.indexTableRelation;
  this.rawDataPath = rawDataPath;
  this.store = store;
  tracker = timeRangeTracker;
  this.opType = opType;
  lmdIndexParameters = store.getLMDIndexParameters();
  int size = 0;
  for (Map.Entry<byte[], TreeSet<byte[]>> entry : tableRelation.getIndexFamilyMap().entrySet()) {
    size += entry.getValue().size();
  }
  dimensions = size;
  int[] mins = new int[dimensions];
  Arrays.fill(mins, 0);
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
/**
 * Do a small get/scan against one store. This is required because store
 * has no actual methods of querying itself, and relies on StoreScanner.
 */
public static List<Cell> getFromStoreFile(HStore store,
                                              Get get) throws IOException {
  Scan scan = new Scan(get);
  InternalScanner scanner = (InternalScanner) store.getScanner(scan,
      scan.getFamilyMap().get(store.getFamily().getName()),
      // originally MultiVersionConcurrencyControl.resetThreadReadPoint() was called to set
      // readpoint 0.
      0);

  List<Cell> result = new ArrayList<Cell>();
  scanner.next(result);
  if (!result.isEmpty()) {
    // verify that we are on the row we want:
    Cell kv = result.get(0);
    if (!CellUtil.matchingRow(kv, get.getRow())) {
      result.clear();
    }
  }
  scanner.close();
  return result;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
/**
 * @param services Master services instance.
 * @param htd
 * @param parent
 * @param daughter
 * @param midkey
 * @param top True if we are to write a 'top' reference.
 * @return Path to reference we created.
 * @throws IOException
 */
private Path createReferences(final MasterServices services,
    final HTableDescriptor htd, final HRegionInfo parent,
    final HRegionInfo daughter, final byte [] midkey, final boolean top)
throws IOException {
  Path rootdir = services.getMasterFileSystem().getRootDir();
  Path tabledir = FSUtils.getTableDir(rootdir, parent.getTable());
  Path storedir = HStore.getStoreHomedir(tabledir, daughter,
    htd.getColumnFamilies()[0].getName());
  Reference ref =
    top? Reference.createTopReference(midkey): Reference.createBottomReference(midkey);
  long now = System.currentTimeMillis();
  // Reference name has this format: StoreFile#REF_NAME_PARSER
  Path p = new Path(storedir, Long.toString(now) + "." + parent.getEncodedName());
  FileSystem fs = services.getMasterFileSystem().getFileSystem();
  ref.write(fs, p);
  return p;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
@Override
public KeyValueScanner preStoreScannerOpen(
    final ObserverContext<RegionCoprocessorEnvironment> c, Store store, final Scan scan,
    final NavigableSet<byte[]> targetCols, KeyValueScanner s) throws IOException {
  TableName tn = store.getTableName();
  if (!tn.isSystemTable()) {
    Long newTtl = ttls.get(store.getTableName());
    Integer newVersions = versions.get(store.getTableName());
    ScanInfo oldSI = store.getScanInfo();
    HColumnDescriptor family = store.getFamily();
    ScanInfo scanInfo = new ScanInfo(TEST_UTIL.getConfiguration(),
        family.getName(), family.getMinVersions(),
        newVersions == null ? family.getMaxVersions() : newVersions,
        newTtl == null ? oldSI.getTtl() : newTtl, family.getKeepDeletedCells(),
        oldSI.getTimeToPurgeDeletes(), oldSI.getComparator());
    return new StoreScanner(store, scanInfo, scan, targetCols,
        ((HStore) store).getHRegion().getReadpoint(IsolationLevel.READ_COMMITTED));
  } else {
    return s;
  }
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
@Test
public void testPurgeExpiredFiles() throws Exception {
  Configuration conf = TEST_UTIL.getConfiguration();
  conf.setInt(HStore.BLOCKING_STOREFILES_KEY, 10000);

  TEST_UTIL.startMiniCluster(1);
  try {
    Store store = prepareData();
    assertEquals(10, store.getStorefilesCount());
    TEST_UTIL.getHBaseAdmin().majorCompact(tableName);
    while (store.getStorefilesCount() > 1) {
      Thread.sleep(100);
    }
    assertTrue(store.getStorefilesCount() == 1);
  } finally {
    TEST_UTIL.shutdownMiniCluster();
  }
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
private long testCompactionWithoutThroughputLimit() throws Exception {
  Configuration conf = TEST_UTIL.getConfiguration();
  conf.set(StoreEngine.STORE_ENGINE_CLASS_KEY, DefaultStoreEngine.class.getName());
  conf.setInt(CompactionConfiguration.HBASE_HSTORE_COMPACTION_MIN_KEY, 100);
  conf.setInt(CompactionConfiguration.HBASE_HSTORE_COMPACTION_MAX_KEY, 200);
  conf.setInt(HStore.BLOCKING_STOREFILES_KEY, 10000);
  conf.set(CompactionThroughputControllerFactory.HBASE_THROUGHPUT_CONTROLLER_KEY,
    NoLimitCompactionThroughputController.class.getName());
  TEST_UTIL.startMiniCluster(1);
  try {
    Store store = prepareData();
    assertEquals(10, store.getStorefilesCount());
    long startTime = System.currentTimeMillis();
    TEST_UTIL.getHBaseAdmin().majorCompact(tableName);
    while (store.getStorefilesCount() != 1) {
      Thread.sleep(20);
    }
    return System.currentTimeMillis() - startTime;
  } finally {
    TEST_UTIL.shutdownMiniCluster();
  }
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
private HTableDescriptor createHtd(boolean isStripe) throws Exception {
  HTableDescriptor htd = new HTableDescriptor(TABLE_NAME);
  htd.addFamily(new HColumnDescriptor(COLUMN_FAMILY));
  String noSplitsPolicy = DisabledRegionSplitPolicy.class.getName();
  htd.setConfiguration(HConstants.HBASE_REGION_SPLIT_POLICY_KEY, noSplitsPolicy);
  if (isStripe) {
    htd.setConfiguration(StoreEngine.STORE_ENGINE_CLASS_KEY, StripeStoreEngine.class.getName());
    if (initialStripeCount != null) {
      htd.setConfiguration(
          StripeStoreConfig.INITIAL_STRIPE_COUNT_KEY, initialStripeCount.toString());
      htd.setConfiguration(
          HStore.BLOCKING_STOREFILES_KEY, Long.toString(10 * initialStripeCount));
    } else {
      htd.setConfiguration(HStore.BLOCKING_STOREFILES_KEY, "500");
    }
    if (splitSize != null) {
      htd.setConfiguration(StripeStoreConfig.SIZE_TO_SPLIT_KEY, splitSize.toString());
    }
    if (splitParts != null) {
      htd.setConfiguration(StripeStoreConfig.SPLIT_PARTS_KEY, splitParts.toString());
    }
  } else {
    htd.setConfiguration(HStore.BLOCKING_STOREFILES_KEY, "10"); // default
  }
  return htd;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
/**
 * Do a small get/scan against one store. This is required because store
 * has no actual methods of querying itself, and relies on StoreScanner.
 */
public static List<Cell> getFromStoreFile(HStore store,
                                              Get get) throws IOException {
  Scan scan = new Scan(get);
  InternalScanner scanner = (InternalScanner) store.getScanner(scan,
      scan.getFamilyMap().get(store.getFamily().getName()),
      // originally MultiVersionConsistencyControl.resetThreadReadPoint() was called to set
      // readpoint 0.
      0);

  List<Cell> result = new ArrayList<Cell>();
  scanner.next(result);
  if (!result.isEmpty()) {
    // verify that we are on the row we want:
    Cell kv = result.get(0);
    if (!CellUtil.matchingRow(kv, get.getRow())) {
      result.clear();
    }
  }
  scanner.close();
  return result;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
@Override
public KeyValueScanner preStoreScannerOpen(
    final ObserverContext<RegionCoprocessorEnvironment> c, Store store, final Scan scan,
    final NavigableSet<byte[]> targetCols, KeyValueScanner s) throws IOException {
  TableName tn = store.getTableName();
  if (!tn.isSystemTable()) {
    Long newTtl = ttls.get(store.getTableName());
    Integer newVersions = versions.get(store.getTableName());
    ScanInfo oldSI = store.getScanInfo();
    HColumnDescriptor family = store.getFamily();
    ScanInfo scanInfo = new ScanInfo(family.getName(), family.getMinVersions(),
        newVersions == null ? family.getMaxVersions() : newVersions,
        newTtl == null ? oldSI.getTtl() : newTtl, family.getKeepDeletedCells(),
        oldSI.getTimeToPurgeDeletes(), oldSI.getComparator());
    return new StoreScanner(store, scanInfo, scan, targetCols,
        ((HStore) store).getHRegion().getReadpoint(IsolationLevel.READ_COMMITTED));
  } else {
    return s;
  }
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
/**
 * Gets the number of del cell in the del files
 * @param paths the del file paths
 * @return the cell size
 */
private int countDelCellsInDelFiles(List<Path> paths) throws IOException {
  List<HStoreFile> sfs = new ArrayList<>();
  int size = 0;
  for (Path path : paths) {
    HStoreFile sf = new HStoreFile(fs, path, conf, cacheConf, BloomType.NONE, true);
    sfs.add(sf);
  }
  List<KeyValueScanner> scanners = new ArrayList<>(StoreFileScanner.getScannersForStoreFiles(sfs,
    false, true, false, false, HConstants.LATEST_TIMESTAMP));
  long timeToPurgeDeletes = Math.max(conf.getLong("hbase.hstore.time.to.purge.deletes", 0), 0);
  long ttl = HStore.determineTTLFromFamily(hcd);
  ScanInfo scanInfo = new ScanInfo(conf, hcd, ttl, timeToPurgeDeletes, CellComparatorImpl.COMPARATOR);
  StoreScanner scanner = new StoreScanner(scanInfo, ScanType.COMPACT_RETAIN_DELETES, scanners);
  List<Cell> results = new ArrayList<>();
  boolean hasMore = true;

  while (hasMore) {
    hasMore = scanner.next(results);
    size += results.size();
    results.clear();
  }
  scanner.close();
  return size;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
@Override
protected List<KeyValueScanner> selectScannersFrom(HStore store,
    List<? extends KeyValueScanner> allScanners) {
  List<KeyValueScanner> scanners = super.selectScannersFrom(store, allScanners);
  List<KeyValueScanner> newScanners = new ArrayList<>(scanners.size());
  for (KeyValueScanner scanner : scanners) {
    newScanners.add(new DelegatingKeyValueScanner(scanner) {
      @Override
      public boolean reseek(Cell key) throws IOException {
        if (ON.get()) {
          REQ_COUNT.incrementAndGet();
          if (!THROW_ONCE.get() || REQ_COUNT.get() == 1) {
            if (IS_DO_NOT_RETRY.get()) {
              throw new DoNotRetryIOException("Injected exception");
            } else {
              throw new IOException("Injected exception");
            }
          }
        }
        return super.reseek(key);
      }
    });
  }
  return newScanners;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
/**
 * Do a small get/scan against one store. This is required because store
 * has no actual methods of querying itself, and relies on StoreScanner.
 */
public static List<Cell> getFromStoreFile(HStore store,
                                              Get get) throws IOException {
  Scan scan = new Scan(get);
  InternalScanner scanner = (InternalScanner) store.getScanner(scan,
      scan.getFamilyMap().get(store.getColumnFamilyDescriptor().getName()),
      // originally MultiVersionConcurrencyControl.resetThreadReadPoint() was called to set
      // readpoint 0.
      0);

  List<Cell> result = new ArrayList<>();
  scanner.next(result);
  if (!result.isEmpty()) {
    // verify that we are on the row we want:
    Cell kv = result.get(0);
    if (!CellUtil.matchingRows(kv, get.getRow())) {
      result.clear();
    }
  }
  scanner.close();
  return result;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
private Path createReferences(final MasterServices services,
    final TableDescriptor td, final HRegionInfo parent,
    final HRegionInfo daughter, final byte [] midkey, final boolean top)
throws IOException {
  Path rootdir = services.getMasterFileSystem().getRootDir();
  Path tabledir = FSUtils.getTableDir(rootdir, parent.getTable());
  Path storedir = HStore.getStoreHomedir(tabledir, daughter,
    td.getColumnFamilies()[0].getName());
  Reference ref =
    top? Reference.createTopReference(midkey): Reference.createBottomReference(midkey);
  long now = System.currentTimeMillis();
  // Reference name has this format: StoreFile#REF_NAME_PARSER
  Path p = new Path(storedir, Long.toString(now) + "." + parent.getEncodedName());
  FileSystem fs = services.getMasterFileSystem().getFileSystem();
  ref.write(fs, p);
  return p;
}
 

import org.apache.hadoop.hbase.regionserver.HStore; //导入依赖的package包/类
private HStore prepareData() throws IOException {
  Admin admin = TEST_UTIL.getAdmin();
  TableDescriptor desc = TableDescriptorBuilder.newBuilder(tableName)
      .setValue(DefaultStoreEngine.DEFAULT_COMPACTION_POLICY_CLASS_KEY,
        FIFOCompactionPolicy.class.getName())
      .setValue(HConstants.HBASE_REGION_SPLIT_POLICY_KEY,
        DisabledRegionSplitPolicy.class.getName())
      .addColumnFamily(ColumnFamilyDescriptorBuilder.newBuilder(family).setTimeToLive(1).build())
      .build();
  admin.createTable(desc);
  Table table = TEST_UTIL.getConnection().getTable(tableName);
  TimeOffsetEnvironmentEdge edge =
      (TimeOffsetEnvironmentEdge) EnvironmentEdgeManager.getDelegate();
  for (int i = 0; i < 10; i++) {
    for (int j = 0; j < 10; j++) {
      byte[] value = new byte[128 * 1024];
      ThreadLocalRandom.current().nextBytes(value);
      table.put(new Put(Bytes.toBytes(i * 10 + j)).addColumn(family, qualifier, value));
    }
    admin.flush(tableName);
    edge.increment(1001);
  }
  return getStoreWithName(tableName);
}