hbase-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From git-site-r...@apache.org
Subject [36/51] [partial] hbase-site git commit: Published site at .
Date Sat, 09 Dec 2017 15:19:37 GMT
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/e23b49ba/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/HFileOutputFormat2.html
----------------------------------------------------------------------
diff --git a/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/HFileOutputFormat2.html b/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/HFileOutputFormat2.html
index 5c004ce..63ead06 100644
--- a/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/HFileOutputFormat2.html
+++ b/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/HFileOutputFormat2.html
@@ -342,584 +342,583 @@
 <span class="sourceLineNo">334</span>        }<a name="line.334"></a>
 <span class="sourceLineNo">335</span><a name="line.335"></a>
 <span class="sourceLineNo">336</span>        // we now have the proper WAL writer. full steam ahead<a name="line.336"></a>
-<span class="sourceLineNo">337</span>        // TODO : Currently in SettableTimeStamp but this will also move to ExtendedCell<a name="line.337"></a>
-<span class="sourceLineNo">338</span>        PrivateCellUtil.updateLatestStamp(cell, this.now);<a name="line.338"></a>
-<span class="sourceLineNo">339</span>        wl.writer.append(kv);<a name="line.339"></a>
-<span class="sourceLineNo">340</span>        wl.written += length;<a name="line.340"></a>
-<span class="sourceLineNo">341</span><a name="line.341"></a>
-<span class="sourceLineNo">342</span>        // Copy the row so we know when a row transition.<a name="line.342"></a>
-<span class="sourceLineNo">343</span>        this.previousRow = rowKey;<a name="line.343"></a>
-<span class="sourceLineNo">344</span>      }<a name="line.344"></a>
-<span class="sourceLineNo">345</span><a name="line.345"></a>
-<span class="sourceLineNo">346</span>      private void rollWriters(WriterLength writerLength) throws IOException {<a name="line.346"></a>
-<span class="sourceLineNo">347</span>        if (writerLength != null) {<a name="line.347"></a>
-<span class="sourceLineNo">348</span>          closeWriter(writerLength);<a name="line.348"></a>
-<span class="sourceLineNo">349</span>        } else {<a name="line.349"></a>
-<span class="sourceLineNo">350</span>          for (WriterLength wl : this.writers.values()) {<a name="line.350"></a>
-<span class="sourceLineNo">351</span>            closeWriter(wl);<a name="line.351"></a>
-<span class="sourceLineNo">352</span>          }<a name="line.352"></a>
-<span class="sourceLineNo">353</span>        }<a name="line.353"></a>
-<span class="sourceLineNo">354</span>        this.rollRequested = false;<a name="line.354"></a>
-<span class="sourceLineNo">355</span>      }<a name="line.355"></a>
-<span class="sourceLineNo">356</span><a name="line.356"></a>
-<span class="sourceLineNo">357</span>      private void closeWriter(WriterLength wl) throws IOException {<a name="line.357"></a>
-<span class="sourceLineNo">358</span>        if (wl.writer != null) {<a name="line.358"></a>
-<span class="sourceLineNo">359</span>          LOG.info(<a name="line.359"></a>
-<span class="sourceLineNo">360</span>              "Writer=" + wl.writer.getPath() + ((wl.written == 0)? "": ", wrote=" + wl.written));<a name="line.360"></a>
-<span class="sourceLineNo">361</span>          close(wl.writer);<a name="line.361"></a>
-<span class="sourceLineNo">362</span>        }<a name="line.362"></a>
-<span class="sourceLineNo">363</span>        wl.writer = null;<a name="line.363"></a>
-<span class="sourceLineNo">364</span>        wl.written = 0;<a name="line.364"></a>
-<span class="sourceLineNo">365</span>      }<a name="line.365"></a>
-<span class="sourceLineNo">366</span><a name="line.366"></a>
-<span class="sourceLineNo">367</span>      /*<a name="line.367"></a>
-<span class="sourceLineNo">368</span>       * Create a new StoreFile.Writer.<a name="line.368"></a>
-<span class="sourceLineNo">369</span>       * @param family<a name="line.369"></a>
-<span class="sourceLineNo">370</span>       * @return A WriterLength, containing a new StoreFile.Writer.<a name="line.370"></a>
-<span class="sourceLineNo">371</span>       * @throws IOException<a name="line.371"></a>
-<span class="sourceLineNo">372</span>       */<a name="line.372"></a>
-<span class="sourceLineNo">373</span>      @edu.umd.cs.findbugs.annotations.SuppressWarnings(value="BX_UNBOXING_IMMEDIATELY_REBOXED",<a name="line.373"></a>
-<span class="sourceLineNo">374</span>          justification="Not important")<a name="line.374"></a>
-<span class="sourceLineNo">375</span>      private WriterLength getNewWriter(byte[] tableName, byte[] family, Configuration<a name="line.375"></a>
-<span class="sourceLineNo">376</span>              conf, InetSocketAddress[] favoredNodes) throws IOException {<a name="line.376"></a>
-<span class="sourceLineNo">377</span>        byte[] tableAndFamily = getTableNameSuffixedWithFamily(tableName, family);<a name="line.377"></a>
-<span class="sourceLineNo">378</span>        Path familydir = new Path(outputDir, Bytes.toString(family));<a name="line.378"></a>
-<span class="sourceLineNo">379</span>        if (writeMultipleTables) {<a name="line.379"></a>
-<span class="sourceLineNo">380</span>          familydir = new Path(outputDir,<a name="line.380"></a>
-<span class="sourceLineNo">381</span>                  new Path(Bytes.toString(tableName), Bytes.toString(family)));<a name="line.381"></a>
-<span class="sourceLineNo">382</span>        }<a name="line.382"></a>
-<span class="sourceLineNo">383</span>        WriterLength wl = new WriterLength();<a name="line.383"></a>
-<span class="sourceLineNo">384</span>        Algorithm compression = compressionMap.get(tableAndFamily);<a name="line.384"></a>
-<span class="sourceLineNo">385</span>        compression = compression == null ? defaultCompression : compression;<a name="line.385"></a>
-<span class="sourceLineNo">386</span>        BloomType bloomType = bloomTypeMap.get(tableAndFamily);<a name="line.386"></a>
-<span class="sourceLineNo">387</span>        bloomType = bloomType == null ? BloomType.NONE : bloomType;<a name="line.387"></a>
-<span class="sourceLineNo">388</span>        Integer blockSize = blockSizeMap.get(tableAndFamily);<a name="line.388"></a>
-<span class="sourceLineNo">389</span>        blockSize = blockSize == null ? HConstants.DEFAULT_BLOCKSIZE : blockSize;<a name="line.389"></a>
-<span class="sourceLineNo">390</span>        DataBlockEncoding encoding = overriddenEncoding;<a name="line.390"></a>
-<span class="sourceLineNo">391</span>        encoding = encoding == null ? datablockEncodingMap.get(tableAndFamily) : encoding;<a name="line.391"></a>
-<span class="sourceLineNo">392</span>        encoding = encoding == null ? DataBlockEncoding.NONE : encoding;<a name="line.392"></a>
-<span class="sourceLineNo">393</span>        Configuration tempConf = new Configuration(conf);<a name="line.393"></a>
-<span class="sourceLineNo">394</span>        tempConf.setFloat(HConstants.HFILE_BLOCK_CACHE_SIZE_KEY, 0.0f);<a name="line.394"></a>
-<span class="sourceLineNo">395</span>        HFileContextBuilder contextBuilder = new HFileContextBuilder()<a name="line.395"></a>
-<span class="sourceLineNo">396</span>                                    .withCompression(compression)<a name="line.396"></a>
-<span class="sourceLineNo">397</span>                                    .withChecksumType(HStore.getChecksumType(conf))<a name="line.397"></a>
-<span class="sourceLineNo">398</span>                                    .withBytesPerCheckSum(HStore.getBytesPerChecksum(conf))<a name="line.398"></a>
-<span class="sourceLineNo">399</span>                                    .withBlockSize(blockSize);<a name="line.399"></a>
-<span class="sourceLineNo">400</span><a name="line.400"></a>
-<span class="sourceLineNo">401</span>        if (HFile.getFormatVersion(conf) &gt;= HFile.MIN_FORMAT_VERSION_WITH_TAGS) {<a name="line.401"></a>
-<span class="sourceLineNo">402</span>          contextBuilder.withIncludesTags(true);<a name="line.402"></a>
-<span class="sourceLineNo">403</span>        }<a name="line.403"></a>
-<span class="sourceLineNo">404</span><a name="line.404"></a>
-<span class="sourceLineNo">405</span>        contextBuilder.withDataBlockEncoding(encoding);<a name="line.405"></a>
-<span class="sourceLineNo">406</span>        HFileContext hFileContext = contextBuilder.build();<a name="line.406"></a>
-<span class="sourceLineNo">407</span>        if (null == favoredNodes) {<a name="line.407"></a>
-<span class="sourceLineNo">408</span>          wl.writer =<a name="line.408"></a>
-<span class="sourceLineNo">409</span>              new StoreFileWriter.Builder(conf, new CacheConfig(tempConf), fs)<a name="line.409"></a>
-<span class="sourceLineNo">410</span>                  .withOutputDir(familydir).withBloomType(bloomType)<a name="line.410"></a>
-<span class="sourceLineNo">411</span>                  .withComparator(CellComparator.getInstance()).withFileContext(hFileContext).build();<a name="line.411"></a>
-<span class="sourceLineNo">412</span>        } else {<a name="line.412"></a>
-<span class="sourceLineNo">413</span>          wl.writer =<a name="line.413"></a>
-<span class="sourceLineNo">414</span>              new StoreFileWriter.Builder(conf, new CacheConfig(tempConf), new HFileSystem(fs))<a name="line.414"></a>
-<span class="sourceLineNo">415</span>                  .withOutputDir(familydir).withBloomType(bloomType)<a name="line.415"></a>
-<span class="sourceLineNo">416</span>                  .withComparator(CellComparator.getInstance()).withFileContext(hFileContext)<a name="line.416"></a>
-<span class="sourceLineNo">417</span>                  .withFavoredNodes(favoredNodes).build();<a name="line.417"></a>
-<span class="sourceLineNo">418</span>        }<a name="line.418"></a>
-<span class="sourceLineNo">419</span><a name="line.419"></a>
-<span class="sourceLineNo">420</span>        this.writers.put(tableAndFamily, wl);<a name="line.420"></a>
-<span class="sourceLineNo">421</span>        return wl;<a name="line.421"></a>
-<span class="sourceLineNo">422</span>      }<a name="line.422"></a>
-<span class="sourceLineNo">423</span><a name="line.423"></a>
-<span class="sourceLineNo">424</span>      private void close(final StoreFileWriter w) throws IOException {<a name="line.424"></a>
-<span class="sourceLineNo">425</span>        if (w != null) {<a name="line.425"></a>
-<span class="sourceLineNo">426</span>          w.appendFileInfo(BULKLOAD_TIME_KEY,<a name="line.426"></a>
-<span class="sourceLineNo">427</span>              Bytes.toBytes(System.currentTimeMillis()));<a name="line.427"></a>
-<span class="sourceLineNo">428</span>          w.appendFileInfo(BULKLOAD_TASK_KEY,<a name="line.428"></a>
-<span class="sourceLineNo">429</span>              Bytes.toBytes(context.getTaskAttemptID().toString()));<a name="line.429"></a>
-<span class="sourceLineNo">430</span>          w.appendFileInfo(MAJOR_COMPACTION_KEY,<a name="line.430"></a>
-<span class="sourceLineNo">431</span>              Bytes.toBytes(true));<a name="line.431"></a>
-<span class="sourceLineNo">432</span>          w.appendFileInfo(EXCLUDE_FROM_MINOR_COMPACTION_KEY,<a name="line.432"></a>
-<span class="sourceLineNo">433</span>              Bytes.toBytes(compactionExclude));<a name="line.433"></a>
-<span class="sourceLineNo">434</span>          w.appendTrackedTimestampsToMetadata();<a name="line.434"></a>
-<span class="sourceLineNo">435</span>          w.close();<a name="line.435"></a>
-<span class="sourceLineNo">436</span>        }<a name="line.436"></a>
-<span class="sourceLineNo">437</span>      }<a name="line.437"></a>
-<span class="sourceLineNo">438</span><a name="line.438"></a>
-<span class="sourceLineNo">439</span>      @Override<a name="line.439"></a>
-<span class="sourceLineNo">440</span>      public void close(TaskAttemptContext c)<a name="line.440"></a>
-<span class="sourceLineNo">441</span>      throws IOException, InterruptedException {<a name="line.441"></a>
-<span class="sourceLineNo">442</span>        for (WriterLength wl: this.writers.values()) {<a name="line.442"></a>
-<span class="sourceLineNo">443</span>          close(wl.writer);<a name="line.443"></a>
-<span class="sourceLineNo">444</span>        }<a name="line.444"></a>
-<span class="sourceLineNo">445</span>      }<a name="line.445"></a>
-<span class="sourceLineNo">446</span>    };<a name="line.446"></a>
-<span class="sourceLineNo">447</span>  }<a name="line.447"></a>
-<span class="sourceLineNo">448</span><a name="line.448"></a>
-<span class="sourceLineNo">449</span>  /**<a name="line.449"></a>
-<span class="sourceLineNo">450</span>   * Configure block storage policy for CF after the directory is created.<a name="line.450"></a>
-<span class="sourceLineNo">451</span>   */<a name="line.451"></a>
-<span class="sourceLineNo">452</span>  static void configureStoragePolicy(final Configuration conf, final FileSystem fs,<a name="line.452"></a>
-<span class="sourceLineNo">453</span>      byte[] tableAndFamily, Path cfPath) {<a name="line.453"></a>
-<span class="sourceLineNo">454</span>    if (null == conf || null == fs || null == tableAndFamily || null == cfPath) {<a name="line.454"></a>
-<span class="sourceLineNo">455</span>      return;<a name="line.455"></a>
-<span class="sourceLineNo">456</span>    }<a name="line.456"></a>
-<span class="sourceLineNo">457</span><a name="line.457"></a>
-<span class="sourceLineNo">458</span>    String policy =<a name="line.458"></a>
-<span class="sourceLineNo">459</span>        conf.get(STORAGE_POLICY_PROPERTY_CF_PREFIX + Bytes.toString(tableAndFamily),<a name="line.459"></a>
-<span class="sourceLineNo">460</span>          conf.get(STORAGE_POLICY_PROPERTY));<a name="line.460"></a>
-<span class="sourceLineNo">461</span>    FSUtils.setStoragePolicy(fs, cfPath, policy);<a name="line.461"></a>
-<span class="sourceLineNo">462</span>  }<a name="line.462"></a>
-<span class="sourceLineNo">463</span><a name="line.463"></a>
-<span class="sourceLineNo">464</span>  /*<a name="line.464"></a>
-<span class="sourceLineNo">465</span>   * Data structure to hold a Writer and amount of data written on it.<a name="line.465"></a>
-<span class="sourceLineNo">466</span>   */<a name="line.466"></a>
-<span class="sourceLineNo">467</span>  static class WriterLength {<a name="line.467"></a>
-<span class="sourceLineNo">468</span>    long written = 0;<a name="line.468"></a>
-<span class="sourceLineNo">469</span>    StoreFileWriter writer = null;<a name="line.469"></a>
-<span class="sourceLineNo">470</span>  }<a name="line.470"></a>
-<span class="sourceLineNo">471</span><a name="line.471"></a>
-<span class="sourceLineNo">472</span>  /**<a name="line.472"></a>
-<span class="sourceLineNo">473</span>   * Return the start keys of all of the regions in this table,<a name="line.473"></a>
-<span class="sourceLineNo">474</span>   * as a list of ImmutableBytesWritable.<a name="line.474"></a>
-<span class="sourceLineNo">475</span>   */<a name="line.475"></a>
-<span class="sourceLineNo">476</span>  private static List&lt;ImmutableBytesWritable&gt; getRegionStartKeys(List&lt;RegionLocator&gt; regionLocators,<a name="line.476"></a>
-<span class="sourceLineNo">477</span>                                                                 boolean writeMultipleTables)<a name="line.477"></a>
-<span class="sourceLineNo">478</span>          throws IOException {<a name="line.478"></a>
-<span class="sourceLineNo">479</span><a name="line.479"></a>
-<span class="sourceLineNo">480</span>    ArrayList&lt;ImmutableBytesWritable&gt; ret = new ArrayList&lt;&gt;();<a name="line.480"></a>
-<span class="sourceLineNo">481</span>    for(RegionLocator regionLocator : regionLocators)<a name="line.481"></a>
-<span class="sourceLineNo">482</span>    {<a name="line.482"></a>
-<span class="sourceLineNo">483</span>      TableName tableName = regionLocator.getName();<a name="line.483"></a>
-<span class="sourceLineNo">484</span>      LOG.info("Looking up current regions for table " + tableName);<a name="line.484"></a>
-<span class="sourceLineNo">485</span>      byte[][] byteKeys = regionLocator.getStartKeys();<a name="line.485"></a>
-<span class="sourceLineNo">486</span>      for (byte[] byteKey : byteKeys) {<a name="line.486"></a>
-<span class="sourceLineNo">487</span>        byte[] fullKey = byteKey; //HFileOutputFormat2 use case<a name="line.487"></a>
-<span class="sourceLineNo">488</span>        if (writeMultipleTables)<a name="line.488"></a>
-<span class="sourceLineNo">489</span>        {<a name="line.489"></a>
-<span class="sourceLineNo">490</span>          //MultiTableHFileOutputFormat use case<a name="line.490"></a>
-<span class="sourceLineNo">491</span>          fullKey = combineTableNameSuffix(tableName.getName(), byteKey);<a name="line.491"></a>
-<span class="sourceLineNo">492</span>        }<a name="line.492"></a>
-<span class="sourceLineNo">493</span>        if (LOG.isDebugEnabled()) {<a name="line.493"></a>
-<span class="sourceLineNo">494</span>          LOG.debug("SplitPoint startkey for table [" + tableName + "]: [" + Bytes.toStringBinary<a name="line.494"></a>
-<span class="sourceLineNo">495</span>                  (fullKey) + "]");<a name="line.495"></a>
-<span class="sourceLineNo">496</span>        }<a name="line.496"></a>
-<span class="sourceLineNo">497</span>        ret.add(new ImmutableBytesWritable(fullKey));<a name="line.497"></a>
-<span class="sourceLineNo">498</span>      }<a name="line.498"></a>
-<span class="sourceLineNo">499</span>    }<a name="line.499"></a>
-<span class="sourceLineNo">500</span>    return ret;<a name="line.500"></a>
-<span class="sourceLineNo">501</span>  }<a name="line.501"></a>
-<span class="sourceLineNo">502</span><a name="line.502"></a>
-<span class="sourceLineNo">503</span>  /**<a name="line.503"></a>
-<span class="sourceLineNo">504</span>   * Write out a {@link SequenceFile} that can be read by<a name="line.504"></a>
-<span class="sourceLineNo">505</span>   * {@link TotalOrderPartitioner} that contains the split points in startKeys.<a name="line.505"></a>
-<span class="sourceLineNo">506</span>   */<a name="line.506"></a>
-<span class="sourceLineNo">507</span>  @SuppressWarnings("deprecation")<a name="line.507"></a>
-<span class="sourceLineNo">508</span>  private static void writePartitions(Configuration conf, Path partitionsPath,<a name="line.508"></a>
-<span class="sourceLineNo">509</span>      List&lt;ImmutableBytesWritable&gt; startKeys, boolean writeMultipleTables) throws IOException {<a name="line.509"></a>
-<span class="sourceLineNo">510</span>    LOG.info("Writing partition information to " + partitionsPath);<a name="line.510"></a>
-<span class="sourceLineNo">511</span>    if (startKeys.isEmpty()) {<a name="line.511"></a>
-<span class="sourceLineNo">512</span>      throw new IllegalArgumentException("No regions passed");<a name="line.512"></a>
-<span class="sourceLineNo">513</span>    }<a name="line.513"></a>
-<span class="sourceLineNo">514</span><a name="line.514"></a>
-<span class="sourceLineNo">515</span>    // We're generating a list of split points, and we don't ever<a name="line.515"></a>
-<span class="sourceLineNo">516</span>    // have keys &lt; the first region (which has an empty start key)<a name="line.516"></a>
-<span class="sourceLineNo">517</span>    // so we need to remove it. Otherwise we would end up with an<a name="line.517"></a>
-<span class="sourceLineNo">518</span>    // empty reducer with index 0<a name="line.518"></a>
-<span class="sourceLineNo">519</span>    TreeSet&lt;ImmutableBytesWritable&gt; sorted = new TreeSet&lt;&gt;(startKeys);<a name="line.519"></a>
-<span class="sourceLineNo">520</span>    ImmutableBytesWritable first = sorted.first();<a name="line.520"></a>
-<span class="sourceLineNo">521</span>    if (writeMultipleTables) {<a name="line.521"></a>
-<span class="sourceLineNo">522</span>      first = new ImmutableBytesWritable(MultiTableHFileOutputFormat.getSuffix(sorted.first<a name="line.522"></a>
-<span class="sourceLineNo">523</span>              ().get()));<a name="line.523"></a>
-<span class="sourceLineNo">524</span>    }<a name="line.524"></a>
-<span class="sourceLineNo">525</span>    if (!first.equals(HConstants.EMPTY_BYTE_ARRAY)) {<a name="line.525"></a>
-<span class="sourceLineNo">526</span>      throw new IllegalArgumentException(<a name="line.526"></a>
-<span class="sourceLineNo">527</span>          "First region of table should have empty start key. Instead has: "<a name="line.527"></a>
-<span class="sourceLineNo">528</span>          + Bytes.toStringBinary(first.get()));<a name="line.528"></a>
-<span class="sourceLineNo">529</span>    }<a name="line.529"></a>
-<span class="sourceLineNo">530</span>    sorted.remove(sorted.first());<a name="line.530"></a>
-<span class="sourceLineNo">531</span><a name="line.531"></a>
-<span class="sourceLineNo">532</span>    // Write the actual file<a name="line.532"></a>
-<span class="sourceLineNo">533</span>    FileSystem fs = partitionsPath.getFileSystem(conf);<a name="line.533"></a>
-<span class="sourceLineNo">534</span>    SequenceFile.Writer writer = SequenceFile.createWriter(<a name="line.534"></a>
-<span class="sourceLineNo">535</span>      fs, conf, partitionsPath, ImmutableBytesWritable.class,<a name="line.535"></a>
-<span class="sourceLineNo">536</span>      NullWritable.class);<a name="line.536"></a>
-<span class="sourceLineNo">537</span><a name="line.537"></a>
-<span class="sourceLineNo">538</span>    try {<a name="line.538"></a>
-<span class="sourceLineNo">539</span>      for (ImmutableBytesWritable startKey : sorted) {<a name="line.539"></a>
-<span class="sourceLineNo">540</span>        writer.append(startKey, NullWritable.get());<a name="line.540"></a>
-<span class="sourceLineNo">541</span>      }<a name="line.541"></a>
-<span class="sourceLineNo">542</span>    } finally {<a name="line.542"></a>
-<span class="sourceLineNo">543</span>      writer.close();<a name="line.543"></a>
-<span class="sourceLineNo">544</span>    }<a name="line.544"></a>
-<span class="sourceLineNo">545</span>  }<a name="line.545"></a>
-<span class="sourceLineNo">546</span><a name="line.546"></a>
-<span class="sourceLineNo">547</span>  /**<a name="line.547"></a>
-<span class="sourceLineNo">548</span>   * Configure a MapReduce Job to perform an incremental load into the given<a name="line.548"></a>
-<span class="sourceLineNo">549</span>   * table. This<a name="line.549"></a>
-<span class="sourceLineNo">550</span>   * &lt;ul&gt;<a name="line.550"></a>
-<span class="sourceLineNo">551</span>   *   &lt;li&gt;Inspects the table to configure a total order partitioner&lt;/li&gt;<a name="line.551"></a>
-<span class="sourceLineNo">552</span>   *   &lt;li&gt;Uploads the partitions file to the cluster and adds it to the DistributedCache&lt;/li&gt;<a name="line.552"></a>
-<span class="sourceLineNo">553</span>   *   &lt;li&gt;Sets the number of reduce tasks to match the current number of regions&lt;/li&gt;<a name="line.553"></a>
-<span class="sourceLineNo">554</span>   *   &lt;li&gt;Sets the output key/value class to match HFileOutputFormat2's requirements&lt;/li&gt;<a name="line.554"></a>
-<span class="sourceLineNo">555</span>   *   &lt;li&gt;Sets the reducer up to perform the appropriate sorting (either KeyValueSortReducer or<a name="line.555"></a>
-<span class="sourceLineNo">556</span>   *     PutSortReducer)&lt;/li&gt;<a name="line.556"></a>
-<span class="sourceLineNo">557</span>   * &lt;/ul&gt;<a name="line.557"></a>
-<span class="sourceLineNo">558</span>   * The user should be sure to set the map output value class to either KeyValue or Put before<a name="line.558"></a>
-<span class="sourceLineNo">559</span>   * running this function.<a name="line.559"></a>
-<span class="sourceLineNo">560</span>   */<a name="line.560"></a>
-<span class="sourceLineNo">561</span>  public static void configureIncrementalLoad(Job job, Table table, RegionLocator regionLocator)<a name="line.561"></a>
-<span class="sourceLineNo">562</span>      throws IOException {<a name="line.562"></a>
-<span class="sourceLineNo">563</span>    configureIncrementalLoad(job, table.getDescriptor(), regionLocator);<a name="line.563"></a>
-<span class="sourceLineNo">564</span>  }<a name="line.564"></a>
-<span class="sourceLineNo">565</span><a name="line.565"></a>
-<span class="sourceLineNo">566</span>  /**<a name="line.566"></a>
-<span class="sourceLineNo">567</span>   * Configure a MapReduce Job to perform an incremental load into the given<a name="line.567"></a>
-<span class="sourceLineNo">568</span>   * table. This<a name="line.568"></a>
-<span class="sourceLineNo">569</span>   * &lt;ul&gt;<a name="line.569"></a>
-<span class="sourceLineNo">570</span>   *   &lt;li&gt;Inspects the table to configure a total order partitioner&lt;/li&gt;<a name="line.570"></a>
-<span class="sourceLineNo">571</span>   *   &lt;li&gt;Uploads the partitions file to the cluster and adds it to the DistributedCache&lt;/li&gt;<a name="line.571"></a>
-<span class="sourceLineNo">572</span>   *   &lt;li&gt;Sets the number of reduce tasks to match the current number of regions&lt;/li&gt;<a name="line.572"></a>
-<span class="sourceLineNo">573</span>   *   &lt;li&gt;Sets the output key/value class to match HFileOutputFormat2's requirements&lt;/li&gt;<a name="line.573"></a>
-<span class="sourceLineNo">574</span>   *   &lt;li&gt;Sets the reducer up to perform the appropriate sorting (either KeyValueSortReducer or<a name="line.574"></a>
-<span class="sourceLineNo">575</span>   *     PutSortReducer)&lt;/li&gt;<a name="line.575"></a>
-<span class="sourceLineNo">576</span>   * &lt;/ul&gt;<a name="line.576"></a>
-<span class="sourceLineNo">577</span>   * The user should be sure to set the map output value class to either KeyValue or Put before<a name="line.577"></a>
-<span class="sourceLineNo">578</span>   * running this function.<a name="line.578"></a>
-<span class="sourceLineNo">579</span>   */<a name="line.579"></a>
-<span class="sourceLineNo">580</span>  public static void configureIncrementalLoad(Job job, TableDescriptor tableDescriptor,<a name="line.580"></a>
-<span class="sourceLineNo">581</span>      RegionLocator regionLocator) throws IOException {<a name="line.581"></a>
-<span class="sourceLineNo">582</span>    ArrayList&lt;TableInfo&gt; singleTableInfo = new ArrayList&lt;&gt;();<a name="line.582"></a>
-<span class="sourceLineNo">583</span>    singleTableInfo.add(new TableInfo(tableDescriptor, regionLocator));<a name="line.583"></a>
-<span class="sourceLineNo">584</span>    configureIncrementalLoad(job, singleTableInfo, HFileOutputFormat2.class);<a name="line.584"></a>
-<span class="sourceLineNo">585</span>  }<a name="line.585"></a>
-<span class="sourceLineNo">586</span><a name="line.586"></a>
-<span class="sourceLineNo">587</span>  static void configureIncrementalLoad(Job job, List&lt;TableInfo&gt; multiTableInfo,<a name="line.587"></a>
-<span class="sourceLineNo">588</span>      Class&lt;? extends OutputFormat&lt;?, ?&gt;&gt; cls) throws IOException {<a name="line.588"></a>
-<span class="sourceLineNo">589</span>    Configuration conf = job.getConfiguration();<a name="line.589"></a>
-<span class="sourceLineNo">590</span>    job.setOutputKeyClass(ImmutableBytesWritable.class);<a name="line.590"></a>
-<span class="sourceLineNo">591</span>    job.setOutputValueClass(MapReduceCell.class);<a name="line.591"></a>
-<span class="sourceLineNo">592</span>    job.setOutputFormatClass(cls);<a name="line.592"></a>
-<span class="sourceLineNo">593</span><a name="line.593"></a>
-<span class="sourceLineNo">594</span>    if (multiTableInfo.stream().distinct().count() != multiTableInfo.size()) {<a name="line.594"></a>
-<span class="sourceLineNo">595</span>      throw new IllegalArgumentException("Duplicate entries found in TableInfo argument");<a name="line.595"></a>
-<span class="sourceLineNo">596</span>    }<a name="line.596"></a>
-<span class="sourceLineNo">597</span>    boolean writeMultipleTables = false;<a name="line.597"></a>
-<span class="sourceLineNo">598</span>    if (MultiTableHFileOutputFormat.class.equals(cls)) {<a name="line.598"></a>
-<span class="sourceLineNo">599</span>      writeMultipleTables = true;<a name="line.599"></a>
-<span class="sourceLineNo">600</span>      conf.setBoolean(MULTI_TABLE_HFILEOUTPUTFORMAT_CONF_KEY, true);<a name="line.600"></a>
-<span class="sourceLineNo">601</span>    }<a name="line.601"></a>
-<span class="sourceLineNo">602</span>    // Based on the configured map output class, set the correct reducer to properly<a name="line.602"></a>
-<span class="sourceLineNo">603</span>    // sort the incoming values.<a name="line.603"></a>
-<span class="sourceLineNo">604</span>    // TODO it would be nice to pick one or the other of these formats.<a name="line.604"></a>
-<span class="sourceLineNo">605</span>    if (KeyValue.class.equals(job.getMapOutputValueClass())<a name="line.605"></a>
-<span class="sourceLineNo">606</span>        || MapReduceCell.class.equals(job.getMapOutputValueClass())) {<a name="line.606"></a>
-<span class="sourceLineNo">607</span>      job.setReducerClass(CellSortReducer.class);<a name="line.607"></a>
-<span class="sourceLineNo">608</span>    } else if (Put.class.equals(job.getMapOutputValueClass())) {<a name="line.608"></a>
-<span class="sourceLineNo">609</span>      job.setReducerClass(PutSortReducer.class);<a name="line.609"></a>
-<span class="sourceLineNo">610</span>    } else if (Text.class.equals(job.getMapOutputValueClass())) {<a name="line.610"></a>
-<span class="sourceLineNo">611</span>      job.setReducerClass(TextSortReducer.class);<a name="line.611"></a>
-<span class="sourceLineNo">612</span>    } else {<a name="line.612"></a>
-<span class="sourceLineNo">613</span>      LOG.warn("Unknown map output value type:" + job.getMapOutputValueClass());<a name="line.613"></a>
-<span class="sourceLineNo">614</span>    }<a name="line.614"></a>
-<span class="sourceLineNo">615</span><a name="line.615"></a>
-<span class="sourceLineNo">616</span>    conf.setStrings("io.serializations", conf.get("io.serializations"),<a name="line.616"></a>
-<span class="sourceLineNo">617</span>        MutationSerialization.class.getName(), ResultSerialization.class.getName(),<a name="line.617"></a>
-<span class="sourceLineNo">618</span>        CellSerialization.class.getName());<a name="line.618"></a>
-<span class="sourceLineNo">619</span><a name="line.619"></a>
-<span class="sourceLineNo">620</span>    if (conf.getBoolean(LOCALITY_SENSITIVE_CONF_KEY, DEFAULT_LOCALITY_SENSITIVE)) {<a name="line.620"></a>
-<span class="sourceLineNo">621</span>      LOG.info("bulkload locality sensitive enabled");<a name="line.621"></a>
-<span class="sourceLineNo">622</span>    }<a name="line.622"></a>
-<span class="sourceLineNo">623</span><a name="line.623"></a>
-<span class="sourceLineNo">624</span>    /* Now get the region start keys for every table required */<a name="line.624"></a>
-<span class="sourceLineNo">625</span>    List&lt;String&gt; allTableNames = new ArrayList&lt;&gt;(multiTableInfo.size());<a name="line.625"></a>
-<span class="sourceLineNo">626</span>    List&lt;RegionLocator&gt; regionLocators = new ArrayList&lt;&gt;( multiTableInfo.size());<a name="line.626"></a>
-<span class="sourceLineNo">627</span>    List&lt;TableDescriptor&gt; tableDescriptors = new ArrayList&lt;&gt;( multiTableInfo.size());<a name="line.627"></a>
-<span class="sourceLineNo">628</span><a name="line.628"></a>
-<span class="sourceLineNo">629</span>    for( TableInfo tableInfo : multiTableInfo )<a name="line.629"></a>
-<span class="sourceLineNo">630</span>    {<a name="line.630"></a>
-<span class="sourceLineNo">631</span>      regionLocators.add(tableInfo.getRegionLocator());<a name="line.631"></a>
-<span class="sourceLineNo">632</span>      allTableNames.add(tableInfo.getRegionLocator().getName().getNameAsString());<a name="line.632"></a>
-<span class="sourceLineNo">633</span>      tableDescriptors.add(tableInfo.getTableDescriptor());<a name="line.633"></a>
-<span class="sourceLineNo">634</span>    }<a name="line.634"></a>
-<span class="sourceLineNo">635</span>    // Record tablenames for creating writer by favored nodes, and decoding compression, block size and other attributes of columnfamily per table<a name="line.635"></a>
-<span class="sourceLineNo">636</span>    conf.set(OUTPUT_TABLE_NAME_CONF_KEY, StringUtils.join(allTableNames, Bytes<a name="line.636"></a>
-<span class="sourceLineNo">637</span>            .toString(tableSeparator)));<a name="line.637"></a>
-<span class="sourceLineNo">638</span>    List&lt;ImmutableBytesWritable&gt; startKeys = getRegionStartKeys(regionLocators, writeMultipleTables);<a name="line.638"></a>
-<span class="sourceLineNo">639</span>    // Use table's region boundaries for TOP split points.<a name="line.639"></a>
-<span class="sourceLineNo">640</span>    LOG.info("Configuring " + startKeys.size() + " reduce partitions " +<a name="line.640"></a>
-<span class="sourceLineNo">641</span>        "to match current region count for all tables");<a name="line.641"></a>
-<span class="sourceLineNo">642</span>    job.setNumReduceTasks(startKeys.size());<a name="line.642"></a>
-<span class="sourceLineNo">643</span><a name="line.643"></a>
-<span class="sourceLineNo">644</span>    configurePartitioner(job, startKeys, writeMultipleTables);<a name="line.644"></a>
-<span class="sourceLineNo">645</span>    // Set compression algorithms based on column families<a name="line.645"></a>
-<span class="sourceLineNo">646</span><a name="line.646"></a>
-<span class="sourceLineNo">647</span>    conf.set(COMPRESSION_FAMILIES_CONF_KEY, serializeColumnFamilyAttribute(compressionDetails,<a name="line.647"></a>
-<span class="sourceLineNo">648</span>            tableDescriptors));<a name="line.648"></a>
-<span class="sourceLineNo">649</span>    conf.set(BLOCK_SIZE_FAMILIES_CONF_KEY, serializeColumnFamilyAttribute(blockSizeDetails,<a name="line.649"></a>
-<span class="sourceLineNo">650</span>            tableDescriptors));<a name="line.650"></a>
-<span class="sourceLineNo">651</span>    conf.set(BLOOM_TYPE_FAMILIES_CONF_KEY, serializeColumnFamilyAttribute(bloomTypeDetails,<a name="line.651"></a>
-<span class="sourceLineNo">652</span>            tableDescriptors));<a name="line.652"></a>
-<span class="sourceLineNo">653</span>    conf.set(DATABLOCK_ENCODING_FAMILIES_CONF_KEY,<a name="line.653"></a>
-<span class="sourceLineNo">654</span>            serializeColumnFamilyAttribute(dataBlockEncodingDetails, tableDescriptors));<a name="line.654"></a>
-<span class="sourceLineNo">655</span><a name="line.655"></a>
-<span class="sourceLineNo">656</span>    TableMapReduceUtil.addDependencyJars(job);<a name="line.656"></a>
-<span class="sourceLineNo">657</span>    TableMapReduceUtil.initCredentials(job);<a name="line.657"></a>
-<span class="sourceLineNo">658</span>    LOG.info("Incremental output configured for tables: " + StringUtils.join(allTableNames, ","));<a name="line.658"></a>
-<span class="sourceLineNo">659</span>  }<a name="line.659"></a>
-<span class="sourceLineNo">660</span><a name="line.660"></a>
-<span class="sourceLineNo">661</span>  public static void configureIncrementalLoadMap(Job job, TableDescriptor tableDescriptor) throws<a name="line.661"></a>
-<span class="sourceLineNo">662</span>      IOException {<a name="line.662"></a>
-<span class="sourceLineNo">663</span>    Configuration conf = job.getConfiguration();<a name="line.663"></a>
-<span class="sourceLineNo">664</span><a name="line.664"></a>
-<span class="sourceLineNo">665</span>    job.setOutputKeyClass(ImmutableBytesWritable.class);<a name="line.665"></a>
-<span class="sourceLineNo">666</span>    job.setOutputValueClass(MapReduceCell.class);<a name="line.666"></a>
-<span class="sourceLineNo">667</span>    job.setOutputFormatClass(HFileOutputFormat2.class);<a name="line.667"></a>
-<span class="sourceLineNo">668</span><a name="line.668"></a>
-<span class="sourceLineNo">669</span>    ArrayList&lt;TableDescriptor&gt; singleTableDescriptor = new ArrayList&lt;&gt;(1);<a name="line.669"></a>
-<span class="sourceLineNo">670</span>    singleTableDescriptor.add(tableDescriptor);<a name="line.670"></a>
-<span class="sourceLineNo">671</span><a name="line.671"></a>
-<span class="sourceLineNo">672</span>    conf.set(OUTPUT_TABLE_NAME_CONF_KEY, tableDescriptor.getTableName().getNameAsString());<a name="line.672"></a>
-<span class="sourceLineNo">673</span>    // Set compression algorithms based on column families<a name="line.673"></a>
-<span class="sourceLineNo">674</span>    conf.set(COMPRESSION_FAMILIES_CONF_KEY,<a name="line.674"></a>
-<span class="sourceLineNo">675</span>        serializeColumnFamilyAttribute(compressionDetails, singleTableDescriptor));<a name="line.675"></a>
-<span class="sourceLineNo">676</span>    conf.set(BLOCK_SIZE_FAMILIES_CONF_KEY,<a name="line.676"></a>
-<span class="sourceLineNo">677</span>        serializeColumnFamilyAttribute(blockSizeDetails, singleTableDescriptor));<a name="line.677"></a>
-<span class="sourceLineNo">678</span>    conf.set(BLOOM_TYPE_FAMILIES_CONF_KEY,<a name="line.678"></a>
-<span class="sourceLineNo">679</span>        serializeColumnFamilyAttribute(bloomTypeDetails, singleTableDescriptor));<a name="line.679"></a>
-<span class="sourceLineNo">680</span>    conf.set(DATABLOCK_ENCODING_FAMILIES_CONF_KEY,<a name="line.680"></a>
-<span class="sourceLineNo">681</span>        serializeColumnFamilyAttribute(dataBlockEncodingDetails, singleTableDescriptor));<a name="line.681"></a>
-<span class="sourceLineNo">682</span><a name="line.682"></a>
-<span class="sourceLineNo">683</span>    TableMapReduceUtil.addDependencyJars(job);<a name="line.683"></a>
-<span class="sourceLineNo">684</span>    TableMapReduceUtil.initCredentials(job);<a name="line.684"></a>
-<span class="sourceLineNo">685</span>    LOG.info("Incremental table " + tableDescriptor.getTableName() + " output configured.");<a name="line.685"></a>
-<span class="sourceLineNo">686</span>  }<a name="line.686"></a>
-<span class="sourceLineNo">687</span><a name="line.687"></a>
-<span class="sourceLineNo">688</span>  /**<a name="line.688"></a>
-<span class="sourceLineNo">689</span>   * Runs inside the task to deserialize column family to compression algorithm<a name="line.689"></a>
-<span class="sourceLineNo">690</span>   * map from the configuration.<a name="line.690"></a>
-<span class="sourceLineNo">691</span>   *<a name="line.691"></a>
-<span class="sourceLineNo">692</span>   * @param conf to read the serialized values from<a name="line.692"></a>
-<span class="sourceLineNo">693</span>   * @return a map from column family to the configured compression algorithm<a name="line.693"></a>
-<span class="sourceLineNo">694</span>   */<a name="line.694"></a>
-<span class="sourceLineNo">695</span>  @VisibleForTesting<a name="line.695"></a>
-<span class="sourceLineNo">696</span>  static Map&lt;byte[], Algorithm&gt; createFamilyCompressionMap(Configuration<a name="line.696"></a>
-<span class="sourceLineNo">697</span>      conf) {<a name="line.697"></a>
-<span class="sourceLineNo">698</span>    Map&lt;byte[], String&gt; stringMap = createFamilyConfValueMap(conf,<a name="line.698"></a>
-<span class="sourceLineNo">699</span>        COMPRESSION_FAMILIES_CONF_KEY);<a name="line.699"></a>
-<span class="sourceLineNo">700</span>    Map&lt;byte[], Algorithm&gt; compressionMap = new TreeMap&lt;&gt;(Bytes.BYTES_COMPARATOR);<a name="line.700"></a>
-<span class="sourceLineNo">701</span>    for (Map.Entry&lt;byte[], String&gt; e : stringMap.entrySet()) {<a name="line.701"></a>
-<span class="sourceLineNo">702</span>      Algorithm algorithm = HFileWriterImpl.compressionByName(e.getValue());<a name="line.702"></a>
-<span class="sourceLineNo">703</span>      compressionMap.put(e.getKey(), algorithm);<a name="line.703"></a>
-<span class="sourceLineNo">704</span>    }<a name="line.704"></a>
-<span class="sourceLineNo">705</span>    return compressionMap;<a name="line.705"></a>
-<span class="sourceLineNo">706</span>  }<a name="line.706"></a>
-<span class="sourceLineNo">707</span><a name="line.707"></a>
-<span class="sourceLineNo">708</span>  /**<a name="line.708"></a>
-<span class="sourceLineNo">709</span>   * Runs inside the task to deserialize column family to bloom filter type<a name="line.709"></a>
-<span class="sourceLineNo">710</span>   * map from the configuration.<a name="line.710"></a>
-<span class="sourceLineNo">711</span>   *<a name="line.711"></a>
-<span class="sourceLineNo">712</span>   * @param conf to read the serialized values from<a name="line.712"></a>
-<span class="sourceLineNo">713</span>   * @return a map from column family to the the configured bloom filter type<a name="line.713"></a>
-<span class="sourceLineNo">714</span>   */<a name="line.714"></a>
-<span class="sourceLineNo">715</span>  @VisibleForTesting<a name="line.715"></a>
-<span class="sourceLineNo">716</span>  static Map&lt;byte[], BloomType&gt; createFamilyBloomTypeMap(Configuration conf) {<a name="line.716"></a>
-<span class="sourceLineNo">717</span>    Map&lt;byte[], String&gt; stringMap = createFamilyConfValueMap(conf,<a name="line.717"></a>
-<span class="sourceLineNo">718</span>        BLOOM_TYPE_FAMILIES_CONF_KEY);<a name="line.718"></a>
-<span class="sourceLineNo">719</span>    Map&lt;byte[], BloomType&gt; bloomTypeMap = new TreeMap&lt;&gt;(Bytes.BYTES_COMPARATOR);<a name="line.719"></a>
-<span class="sourceLineNo">720</span>    for (Map.Entry&lt;byte[], String&gt; e : stringMap.entrySet()) {<a name="line.720"></a>
-<span class="sourceLineNo">721</span>      BloomType bloomType = BloomType.valueOf(e.getValue());<a name="line.721"></a>
-<span class="sourceLineNo">722</span>      bloomTypeMap.put(e.getKey(), bloomType);<a name="line.722"></a>
-<span class="sourceLineNo">723</span>    }<a name="line.723"></a>
-<span class="sourceLineNo">724</span>    return bloomTypeMap;<a name="line.724"></a>
-<span class="sourceLineNo">725</span>  }<a name="line.725"></a>
-<span class="sourceLineNo">726</span><a name="line.726"></a>
-<span class="sourceLineNo">727</span>  /**<a name="line.727"></a>
-<span class="sourceLineNo">728</span>   * Runs inside the task to deserialize column family to block size<a name="line.728"></a>
-<span class="sourceLineNo">729</span>   * map from the configuration.<a name="line.729"></a>
-<span class="sourceLineNo">730</span>   *<a name="line.730"></a>
-<span class="sourceLineNo">731</span>   * @param conf to read the serialized values from<a name="line.731"></a>
-<span class="sourceLineNo">732</span>   * @return a map from column family to the configured block size<a name="line.732"></a>
-<span class="sourceLineNo">733</span>   */<a name="line.733"></a>
-<span class="sourceLineNo">734</span>  @VisibleForTesting<a name="line.734"></a>
-<span class="sourceLineNo">735</span>  static Map&lt;byte[], Integer&gt; createFamilyBlockSizeMap(Configuration conf) {<a name="line.735"></a>
-<span class="sourceLineNo">736</span>    Map&lt;byte[], String&gt; stringMap = createFamilyConfValueMap(conf,<a name="line.736"></a>
-<span class="sourceLineNo">737</span>        BLOCK_SIZE_FAMILIES_CONF_KEY);<a name="line.737"></a>
-<span class="sourceLineNo">738</span>    Map&lt;byte[], Integer&gt; blockSizeMap = new TreeMap&lt;&gt;(Bytes.BYTES_COMPARATOR);<a name="line.738"></a>
-<span class="sourceLineNo">739</span>    for (Map.Entry&lt;byte[], String&gt; e : stringMap.entrySet()) {<a name="line.739"></a>
-<span class="sourceLineNo">740</span>      Integer blockSize = Integer.parseInt(e.getValue());<a name="line.740"></a>
-<span class="sourceLineNo">741</span>      blockSizeMap.put(e.getKey(), blockSize);<a name="line.741"></a>
-<span class="sourceLineNo">742</span>    }<a name="line.742"></a>
-<span class="sourceLineNo">743</span>    return blockSizeMap;<a name="line.743"></a>
-<span class="sourceLineNo">744</span>  }<a name="line.744"></a>
-<span class="sourceLineNo">745</span><a name="line.745"></a>
-<span class="sourceLineNo">746</span>  /**<a name="line.746"></a>
-<span class="sourceLineNo">747</span>   * Runs inside the task to deserialize column family to data block encoding<a name="line.747"></a>
-<span class="sourceLineNo">748</span>   * type map from the configuration.<a name="line.748"></a>
-<span class="sourceLineNo">749</span>   *<a name="line.749"></a>
-<span class="sourceLineNo">750</span>   * @param conf to read the serialized values from<a name="line.750"></a>
-<span class="sourceLineNo">751</span>   * @return a map from column family to HFileDataBlockEncoder for the<a name="line.751"></a>
-<span class="sourceLineNo">752</span>   *         configured data block type for the family<a name="line.752"></a>
-<span class="sourceLineNo">753</span>   */<a name="line.753"></a>
-<span class="sourceLineNo">754</span>  @VisibleForTesting<a name="line.754"></a>
-<span class="sourceLineNo">755</span>  static Map&lt;byte[], DataBlockEncoding&gt; createFamilyDataBlockEncodingMap(<a name="line.755"></a>
-<span class="sourceLineNo">756</span>      Configuration conf) {<a name="line.756"></a>
-<span class="sourceLineNo">757</span>    Map&lt;byte[], String&gt; stringMap = createFamilyConfValueMap(conf,<a name="line.757"></a>
-<span class="sourceLineNo">758</span>        DATABLOCK_ENCODING_FAMILIES_CONF_KEY);<a name="line.758"></a>
-<span class="sourceLineNo">759</span>    Map&lt;byte[], DataBlockEncoding&gt; encoderMap = new TreeMap&lt;&gt;(Bytes.BYTES_COMPARATOR);<a name="line.759"></a>
-<span class="sourceLineNo">760</span>    for (Map.Entry&lt;byte[], String&gt; e : stringMap.entrySet()) {<a name="line.760"></a>
-<span class="sourceLineNo">761</span>      encoderMap.put(e.getKey(), DataBlockEncoding.valueOf((e.getValue())));<a name="line.761"></a>
-<span class="sourceLineNo">762</span>    }<a name="line.762"></a>
-<span class="sourceLineNo">763</span>    return encoderMap;<a name="line.763"></a>
-<span class="sourceLineNo">764</span>  }<a name="line.764"></a>
+<span class="sourceLineNo">337</span>        PrivateCellUtil.updateLatestStamp(cell, this.now);<a name="line.337"></a>
+<span class="sourceLineNo">338</span>        wl.writer.append(kv);<a name="line.338"></a>
+<span class="sourceLineNo">339</span>        wl.written += length;<a name="line.339"></a>
+<span class="sourceLineNo">340</span><a name="line.340"></a>
+<span class="sourceLineNo">341</span>        // Copy the row so we know when a row transition.<a name="line.341"></a>
+<span class="sourceLineNo">342</span>        this.previousRow = rowKey;<a name="line.342"></a>
+<span class="sourceLineNo">343</span>      }<a name="line.343"></a>
+<span class="sourceLineNo">344</span><a name="line.344"></a>
+<span class="sourceLineNo">345</span>      private void rollWriters(WriterLength writerLength) throws IOException {<a name="line.345"></a>
+<span class="sourceLineNo">346</span>        if (writerLength != null) {<a name="line.346"></a>
+<span class="sourceLineNo">347</span>          closeWriter(writerLength);<a name="line.347"></a>
+<span class="sourceLineNo">348</span>        } else {<a name="line.348"></a>
+<span class="sourceLineNo">349</span>          for (WriterLength wl : this.writers.values()) {<a name="line.349"></a>
+<span class="sourceLineNo">350</span>            closeWriter(wl);<a name="line.350"></a>
+<span class="sourceLineNo">351</span>          }<a name="line.351"></a>
+<span class="sourceLineNo">352</span>        }<a name="line.352"></a>
+<span class="sourceLineNo">353</span>        this.rollRequested = false;<a name="line.353"></a>
+<span class="sourceLineNo">354</span>      }<a name="line.354"></a>
+<span class="sourceLineNo">355</span><a name="line.355"></a>
+<span class="sourceLineNo">356</span>      private void closeWriter(WriterLength wl) throws IOException {<a name="line.356"></a>
+<span class="sourceLineNo">357</span>        if (wl.writer != null) {<a name="line.357"></a>
+<span class="sourceLineNo">358</span>          LOG.info(<a name="line.358"></a>
+<span class="sourceLineNo">359</span>              "Writer=" + wl.writer.getPath() + ((wl.written == 0)? "": ", wrote=" + wl.written));<a name="line.359"></a>
+<span class="sourceLineNo">360</span>          close(wl.writer);<a name="line.360"></a>
+<span class="sourceLineNo">361</span>        }<a name="line.361"></a>
+<span class="sourceLineNo">362</span>        wl.writer = null;<a name="line.362"></a>
+<span class="sourceLineNo">363</span>        wl.written = 0;<a name="line.363"></a>
+<span class="sourceLineNo">364</span>      }<a name="line.364"></a>
+<span class="sourceLineNo">365</span><a name="line.365"></a>
+<span class="sourceLineNo">366</span>      /*<a name="line.366"></a>
+<span class="sourceLineNo">367</span>       * Create a new StoreFile.Writer.<a name="line.367"></a>
+<span class="sourceLineNo">368</span>       * @param family<a name="line.368"></a>
+<span class="sourceLineNo">369</span>       * @return A WriterLength, containing a new StoreFile.Writer.<a name="line.369"></a>
+<span class="sourceLineNo">370</span>       * @throws IOException<a name="line.370"></a>
+<span class="sourceLineNo">371</span>       */<a name="line.371"></a>
+<span class="sourceLineNo">372</span>      @edu.umd.cs.findbugs.annotations.SuppressWarnings(value="BX_UNBOXING_IMMEDIATELY_REBOXED",<a name="line.372"></a>
+<span class="sourceLineNo">373</span>          justification="Not important")<a name="line.373"></a>
+<span class="sourceLineNo">374</span>      private WriterLength getNewWriter(byte[] tableName, byte[] family, Configuration<a name="line.374"></a>
+<span class="sourceLineNo">375</span>              conf, InetSocketAddress[] favoredNodes) throws IOException {<a name="line.375"></a>
+<span class="sourceLineNo">376</span>        byte[] tableAndFamily = getTableNameSuffixedWithFamily(tableName, family);<a name="line.376"></a>
+<span class="sourceLineNo">377</span>        Path familydir = new Path(outputDir, Bytes.toString(family));<a name="line.377"></a>
+<span class="sourceLineNo">378</span>        if (writeMultipleTables) {<a name="line.378"></a>
+<span class="sourceLineNo">379</span>          familydir = new Path(outputDir,<a name="line.379"></a>
+<span class="sourceLineNo">380</span>                  new Path(Bytes.toString(tableName), Bytes.toString(family)));<a name="line.380"></a>
+<span class="sourceLineNo">381</span>        }<a name="line.381"></a>
+<span class="sourceLineNo">382</span>        WriterLength wl = new WriterLength();<a name="line.382"></a>
+<span class="sourceLineNo">383</span>        Algorithm compression = compressionMap.get(tableAndFamily);<a name="line.383"></a>
+<span class="sourceLineNo">384</span>        compression = compression == null ? defaultCompression : compression;<a name="line.384"></a>
+<span class="sourceLineNo">385</span>        BloomType bloomType = bloomTypeMap.get(tableAndFamily);<a name="line.385"></a>
+<span class="sourceLineNo">386</span>        bloomType = bloomType == null ? BloomType.NONE : bloomType;<a name="line.386"></a>
+<span class="sourceLineNo">387</span>        Integer blockSize = blockSizeMap.get(tableAndFamily);<a name="line.387"></a>
+<span class="sourceLineNo">388</span>        blockSize = blockSize == null ? HConstants.DEFAULT_BLOCKSIZE : blockSize;<a name="line.388"></a>
+<span class="sourceLineNo">389</span>        DataBlockEncoding encoding = overriddenEncoding;<a name="line.389"></a>
+<span class="sourceLineNo">390</span>        encoding = encoding == null ? datablockEncodingMap.get(tableAndFamily) : encoding;<a name="line.390"></a>
+<span class="sourceLineNo">391</span>        encoding = encoding == null ? DataBlockEncoding.NONE : encoding;<a name="line.391"></a>
+<span class="sourceLineNo">392</span>        Configuration tempConf = new Configuration(conf);<a name="line.392"></a>
+<span class="sourceLineNo">393</span>        tempConf.setFloat(HConstants.HFILE_BLOCK_CACHE_SIZE_KEY, 0.0f);<a name="line.393"></a>
+<span class="sourceLineNo">394</span>        HFileContextBuilder contextBuilder = new HFileContextBuilder()<a name="line.394"></a>
+<span class="sourceLineNo">395</span>                                    .withCompression(compression)<a name="line.395"></a>
+<span class="sourceLineNo">396</span>                                    .withChecksumType(HStore.getChecksumType(conf))<a name="line.396"></a>
+<span class="sourceLineNo">397</span>                                    .withBytesPerCheckSum(HStore.getBytesPerChecksum(conf))<a name="line.397"></a>
+<span class="sourceLineNo">398</span>                                    .withBlockSize(blockSize);<a name="line.398"></a>
+<span class="sourceLineNo">399</span><a name="line.399"></a>
+<span class="sourceLineNo">400</span>        if (HFile.getFormatVersion(conf) &gt;= HFile.MIN_FORMAT_VERSION_WITH_TAGS) {<a name="line.400"></a>
+<span class="sourceLineNo">401</span>          contextBuilder.withIncludesTags(true);<a name="line.401"></a>
+<span class="sourceLineNo">402</span>        }<a name="line.402"></a>
+<span class="sourceLineNo">403</span><a name="line.403"></a>
+<span class="sourceLineNo">404</span>        contextBuilder.withDataBlockEncoding(encoding);<a name="line.404"></a>
+<span class="sourceLineNo">405</span>        HFileContext hFileContext = contextBuilder.build();<a name="line.405"></a>
+<span class="sourceLineNo">406</span>        if (null == favoredNodes) {<a name="line.406"></a>
+<span class="sourceLineNo">407</span>          wl.writer =<a name="line.407"></a>
+<span class="sourceLineNo">408</span>              new StoreFileWriter.Builder(conf, new CacheConfig(tempConf), fs)<a name="line.408"></a>
+<span class="sourceLineNo">409</span>                  .withOutputDir(familydir).withBloomType(bloomType)<a name="line.409"></a>
+<span class="sourceLineNo">410</span>                  .withComparator(CellComparator.getInstance()).withFileContext(hFileContext).build();<a name="line.410"></a>
+<span class="sourceLineNo">411</span>        } else {<a name="line.411"></a>
+<span class="sourceLineNo">412</span>          wl.writer =<a name="line.412"></a>
+<span class="sourceLineNo">413</span>              new StoreFileWriter.Builder(conf, new CacheConfig(tempConf), new HFileSystem(fs))<a name="line.413"></a>
+<span class="sourceLineNo">414</span>                  .withOutputDir(familydir).withBloomType(bloomType)<a name="line.414"></a>
+<span class="sourceLineNo">415</span>                  .withComparator(CellComparator.getInstance()).withFileContext(hFileContext)<a name="line.415"></a>
+<span class="sourceLineNo">416</span>                  .withFavoredNodes(favoredNodes).build();<a name="line.416"></a>
+<span class="sourceLineNo">417</span>        }<a name="line.417"></a>
+<span class="sourceLineNo">418</span><a name="line.418"></a>
+<span class="sourceLineNo">419</span>        this.writers.put(tableAndFamily, wl);<a name="line.419"></a>
+<span class="sourceLineNo">420</span>        return wl;<a name="line.420"></a>
+<span class="sourceLineNo">421</span>      }<a name="line.421"></a>
+<span class="sourceLineNo">422</span><a name="line.422"></a>
+<span class="sourceLineNo">423</span>      private void close(final StoreFileWriter w) throws IOException {<a name="line.423"></a>
+<span class="sourceLineNo">424</span>        if (w != null) {<a name="line.424"></a>
+<span class="sourceLineNo">425</span>          w.appendFileInfo(BULKLOAD_TIME_KEY,<a name="line.425"></a>
+<span class="sourceLineNo">426</span>              Bytes.toBytes(System.currentTimeMillis()));<a name="line.426"></a>
+<span class="sourceLineNo">427</span>          w.appendFileInfo(BULKLOAD_TASK_KEY,<a name="line.427"></a>
+<span class="sourceLineNo">428</span>              Bytes.toBytes(context.getTaskAttemptID().toString()));<a name="line.428"></a>
+<span class="sourceLineNo">429</span>          w.appendFileInfo(MAJOR_COMPACTION_KEY,<a name="line.429"></a>
+<span class="sourceLineNo">430</span>              Bytes.toBytes(true));<a name="line.430"></a>
+<span class="sourceLineNo">431</span>          w.appendFileInfo(EXCLUDE_FROM_MINOR_COMPACTION_KEY,<a name="line.431"></a>
+<span class="sourceLineNo">432</span>              Bytes.toBytes(compactionExclude));<a name="line.432"></a>
+<span class="sourceLineNo">433</span>          w.appendTrackedTimestampsToMetadata();<a name="line.433"></a>
+<span class="sourceLineNo">434</span>          w.close();<a name="line.434"></a>
+<span class="sourceLineNo">435</span>        }<a name="line.435"></a>
+<span class="sourceLineNo">436</span>      }<a name="line.436"></a>
+<span class="sourceLineNo">437</span><a name="line.437"></a>
+<span class="sourceLineNo">438</span>      @Override<a name="line.438"></a>
+<span class="sourceLineNo">439</span>      public void close(TaskAttemptContext c)<a name="line.439"></a>
+<span class="sourceLineNo">440</span>      throws IOException, InterruptedException {<a name="line.440"></a>
+<span class="sourceLineNo">441</span>        for (WriterLength wl: this.writers.values()) {<a name="line.441"></a>
+<span class="sourceLineNo">442</span>          close(wl.writer);<a name="line.442"></a>
+<span class="sourceLineNo">443</span>        }<a name="line.443"></a>
+<span class="sourceLineNo">444</span>      }<a name="line.444"></a>
+<span class="sourceLineNo">445</span>    };<a name="line.445"></a>
+<span class="sourceLineNo">446</span>  }<a name="line.446"></a>
+<span class="sourceLineNo">447</span><a name="line.447"></a>
+<span class="sourceLineNo">448</span>  /**<a name="line.448"></a>
+<span class="sourceLineNo">449</span>   * Configure block storage policy for CF after the directory is created.<a name="line.449"></a>
+<span class="sourceLineNo">450</span>   */<a name="line.450"></a>
+<span class="sourceLineNo">451</span>  static void configureStoragePolicy(final Configuration conf, final FileSystem fs,<a name="line.451"></a>
+<span class="sourceLineNo">452</span>      byte[] tableAndFamily, Path cfPath) {<a name="line.452"></a>
+<span class="sourceLineNo">453</span>    if (null == conf || null == fs || null == tableAndFamily || null == cfPath) {<a name="line.453"></a>
+<span class="sourceLineNo">454</span>      return;<a name="line.454"></a>
+<span class="sourceLineNo">455</span>    }<a name="line.455"></a>
+<span class="sourceLineNo">456</span><a name="line.456"></a>
+<span class="sourceLineNo">457</span>    String policy =<a name="line.457"></a>
+<span class="sourceLineNo">458</span>        conf.get(STORAGE_POLICY_PROPERTY_CF_PREFIX + Bytes.toString(tableAndFamily),<a name="line.458"></a>
+<span class="sourceLineNo">459</span>          conf.get(STORAGE_POLICY_PROPERTY));<a name="line.459"></a>
+<span class="sourceLineNo">460</span>    FSUtils.setStoragePolicy(fs, cfPath, policy);<a name="line.460"></a>
+<span class="sourceLineNo">461</span>  }<a name="line.461"></a>
+<span class="sourceLineNo">462</span><a name="line.462"></a>
+<span class="sourceLineNo">463</span>  /*<a name="line.463"></a>
+<span class="sourceLineNo">464</span>   * Data structure to hold a Writer and amount of data written on it.<a name="line.464"></a>
+<span class="sourceLineNo">465</span>   */<a name="line.465"></a>
+<span class="sourceLineNo">466</span>  static class WriterLength {<a name="line.466"></a>
+<span class="sourceLineNo">467</span>    long written = 0;<a name="line.467"></a>
+<span class="sourceLineNo">468</span>    StoreFileWriter writer = null;<a name="line.468"></a>
+<span class="sourceLineNo">469</span>  }<a name="line.469"></a>
+<span class="sourceLineNo">470</span><a name="line.470"></a>
+<span class="sourceLineNo">471</span>  /**<a name="line.471"></a>
+<span class="sourceLineNo">472</span>   * Return the start keys of all of the regions in this table,<a name="line.472"></a>
+<span class="sourceLineNo">473</span>   * as a list of ImmutableBytesWritable.<a name="line.473"></a>
+<span class="sourceLineNo">474</span>   */<a name="line.474"></a>
+<span class="sourceLineNo">475</span>  private static List&lt;ImmutableBytesWritable&gt; getRegionStartKeys(List&lt;RegionLocator&gt; regionLocators,<a name="line.475"></a>
+<span class="sourceLineNo">476</span>                                                                 boolean writeMultipleTables)<a name="line.476"></a>
+<span class="sourceLineNo">477</span>          throws IOException {<a name="line.477"></a>
+<span class="sourceLineNo">478</span><a name="line.478"></a>
+<span class="sourceLineNo">479</span>    ArrayList&lt;ImmutableBytesWritable&gt; ret = new ArrayList&lt;&gt;();<a name="line.479"></a>
+<span class="sourceLineNo">480</span>    for(RegionLocator regionLocator : regionLocators)<a name="line.480"></a>
+<span class="sourceLineNo">481</span>    {<a name="line.481"></a>
+<span class="sourceLineNo">482</span>      TableName tableName = regionLocator.getName();<a name="line.482"></a>
+<span class="sourceLineNo">483</span>      LOG.info("Looking up current regions for table " + tableName);<a name="line.483"></a>
+<span class="sourceLineNo">484</span>      byte[][] byteKeys = regionLocator.getStartKeys();<a name="line.484"></a>
+<span class="sourceLineNo">485</span>      for (byte[] byteKey : byteKeys) {<a name="line.485"></a>
+<span class="sourceLineNo">486</span>        byte[] fullKey = byteKey; //HFileOutputFormat2 use case<a name="line.486"></a>
+<span class="sourceLineNo">487</span>        if (writeMultipleTables)<a name="line.487"></a>
+<span class="sourceLineNo">488</span>        {<a name="line.488"></a>
+<span class="sourceLineNo">489</span>          //MultiTableHFileOutputFormat use case<a name="line.489"></a>
+<span class="sourceLineNo">490</span>          fullKey = combineTableNameSuffix(tableName.getName(), byteKey);<a name="line.490"></a>
+<span class="sourceLineNo">491</span>        }<a name="line.491"></a>
+<span class="sourceLineNo">492</span>        if (LOG.isDebugEnabled()) {<a name="line.492"></a>
+<span class="sourceLineNo">493</span>          LOG.debug("SplitPoint startkey for table [" + tableName + "]: [" + Bytes.toStringBinary<a name="line.493"></a>
+<span class="sourceLineNo">494</span>                  (fullKey) + "]");<a name="line.494"></a>
+<span class="sourceLineNo">495</span>        }<a name="line.495"></a>
+<span class="sourceLineNo">496</span>        ret.add(new ImmutableBytesWritable(fullKey));<a name="line.496"></a>
+<span class="sourceLineNo">497</span>      }<a name="line.497"></a>
+<span class="sourceLineNo">498</span>    }<a name="line.498"></a>
+<span class="sourceLineNo">499</span>    return ret;<a name="line.499"></a>
+<span class="sourceLineNo">500</span>  }<a name="line.500"></a>
+<span class="sourceLineNo">501</span><a name="line.501"></a>
+<span class="sourceLineNo">502</span>  /**<a name="line.502"></a>
+<span class="sourceLineNo">503</span>   * Write out a {@link SequenceFile} that can be read by<a name="line.503"></a>
+<span class="sourceLineNo">504</span>   * {@link TotalOrderPartitioner} that contains the split points in startKeys.<a name="line.504"></a>
+<span class="sourceLineNo">505</span>   */<a name="line.505"></a>
+<span class="sourceLineNo">506</span>  @SuppressWarnings("deprecation")<a name="line.506"></a>
+<span class="sourceLineNo">507</span>  private static void writePartitions(Configuration conf, Path partitionsPath,<a name="line.507"></a>
+<span class="sourceLineNo">508</span>      List&lt;ImmutableBytesWritable&gt; startKeys, boolean writeMultipleTables) throws IOException {<a name="line.508"></a>
+<span class="sourceLineNo">509</span>    LOG.info("Writing partition information to " + partitionsPath);<a name="line.509"></a>
+<span class="sourceLineNo">510</span>    if (startKeys.isEmpty()) {<a name="line.510"></a>
+<span class="sourceLineNo">511</span>      throw new IllegalArgumentException("No regions passed");<a name="line.511"></a>
+<span class="sourceLineNo">512</span>    }<a name="line.512"></a>
+<span class="sourceLineNo">513</span><a name="line.513"></a>
+<span class="sourceLineNo">514</span>    // We're generating a list of split points, and we don't ever<a name="line.514"></a>
+<span class="sourceLineNo">515</span>    // have keys &lt; the first region (which has an empty start key)<a name="line.515"></a>
+<span class="sourceLineNo">516</span>    // so we need to remove it. Otherwise we would end up with an<a name="line.516"></a>
+<span class="sourceLineNo">517</span>    // empty reducer with index 0<a name="line.517"></a>
+<span class="sourceLineNo">518</span>    TreeSet&lt;ImmutableBytesWritable&gt; sorted = new TreeSet&lt;&gt;(startKeys);<a name="line.518"></a>
+<span class="sourceLineNo">519</span>    ImmutableBytesWritable first = sorted.first();<a name="line.519"></a>
+<span class="sourceLineNo">520</span>    if (writeMultipleTables) {<a name="line.520"></a>
+<span class="sourceLineNo">521</span>      first = new ImmutableBytesWritable(MultiTableHFileOutputFormat.getSuffix(sorted.first<a name="line.521"></a>
+<span class="sourceLineNo">522</span>              ().get()));<a name="line.522"></a>
+<span class="sourceLineNo">523</span>    }<a name="line.523"></a>
+<span class="sourceLineNo">524</span>    if (!first.equals(HConstants.EMPTY_BYTE_ARRAY)) {<a name="line.524"></a>
+<span class="sourceLineNo">525</span>      throw new IllegalArgumentException(<a name="line.525"></a>
+<span class="sourceLineNo">526</span>          "First region of table should have empty start key. Instead has: "<a name="line.526"></a>
+<span class="sourceLineNo">527</span>          + Bytes.toStringBinary(first.get()));<a name="line.527"></a>
+<span class="sourceLineNo">528</span>    }<a name="line.528"></a>
+<span class="sourceLineNo">529</span>    sorted.remove(sorted.first());<a name="line.529"></a>
+<span class="sourceLineNo">530</span><a name="line.530"></a>
+<span class="sourceLineNo">531</span>    // Write the actual file<a name="line.531"></a>
+<span class="sourceLineNo">532</span>    FileSystem fs = partitionsPath.getFileSystem(conf);<a name="line.532"></a>
+<span class="sourceLineNo">533</span>    SequenceFile.Writer writer = SequenceFile.createWriter(<a name="line.533"></a>
+<span class="sourceLineNo">534</span>      fs, conf, partitionsPath, ImmutableBytesWritable.class,<a name="line.534"></a>
+<span class="sourceLineNo">535</span>      NullWritable.class);<a name="line.535"></a>
+<span class="sourceLineNo">536</span><a name="line.536"></a>
+<span class="sourceLineNo">537</span>    try {<a name="line.537"></a>
+<span class="sourceLineNo">538</span>      for (ImmutableBytesWritable startKey : sorted) {<a name="line.538"></a>
+<span class="sourceLineNo">539</span>        writer.append(startKey, NullWritable.get());<a name="line.539"></a>
+<span class="sourceLineNo">540</span>      }<a name="line.540"></a>
+<span class="sourceLineNo">541</span>    } finally {<a name="line.541"></a>
+<span class="sourceLineNo">542</span>      writer.close();<a name="line.542"></a>
+<span class="sourceLineNo">543</span>    }<a name="line.543"></a>
+<span class="sourceLineNo">544</span>  }<a name="line.544"></a>
+<span class="sourceLineNo">545</span><a name="line.545"></a>
+<span class="sourceLineNo">546</span>  /**<a name="line.546"></a>
+<span class="sourceLineNo">547</span>   * Configure a MapReduce Job to perform an incremental load into the given<a name="line.547"></a>
+<span class="sourceLineNo">548</span>   * table. This<a name="line.548"></a>
+<span class="sourceLineNo">549</span>   * &lt;ul&gt;<a name="line.549"></a>
+<span class="sourceLineNo">550</span>   *   &lt;li&gt;Inspects the table to configure a total order partitioner&lt;/li&gt;<a name="line.550"></a>
+<span class="sourceLineNo">551</span>   *   &lt;li&gt;Uploads the partitions file to the cluster and adds it to the DistributedCache&lt;/li&gt;<a name="line.551"></a>
+<span class="sourceLineNo">552</span>   *   &lt;li&gt;Sets the number of reduce tasks to match the current number of regions&lt;/li&gt;<a name="line.552"></a>
+<span class="sourceLineNo">553</span>   *   &lt;li&gt;Sets the output key/value class to match HFileOutputFormat2's requirements&lt;/li&gt;<a name="line.553"></a>
+<span class="sourceLineNo">554</span>   *   &lt;li&gt;Sets the reducer up to perform the appropriate sorting (either KeyValueSortReducer or<a name="line.554"></a>
+<span class="sourceLineNo">555</span>   *     PutSortReducer)&lt;/li&gt;<a name="line.555"></a>
+<span class="sourceLineNo">556</span>   * &lt;/ul&gt;<a name="line.556"></a>
+<span class="sourceLineNo">557</span>   * The user should be sure to set the map output value class to either KeyValue or Put before<a name="line.557"></a>
+<span class="sourceLineNo">558</span>   * running this function.<a name="line.558"></a>
+<span class="sourceLineNo">559</span>   */<a name="line.559"></a>
+<span class="sourceLineNo">560</span>  public static void configureIncrementalLoad(Job job, Table table, RegionLocator regionLocator)<a name="line.560"></a>
+<span class="sourceLineNo">561</span>      throws IOException {<a name="line.561"></a>
+<span class="sourceLineNo">562</span>    configureIncrementalLoad(job, table.getDescriptor(), regionLocator);<a name="line.562"></a>
+<span class="sourceLineNo">563</span>  }<a name="line.563"></a>
+<span class="sourceLineNo">564</span><a name="line.564"></a>
+<span class="sourceLineNo">565</span>  /**<a name="line.565"></a>
+<span class="sourceLineNo">566</span>   * Configure a MapReduce Job to perform an incremental load into the given<a name="line.566"></a>
+<span class="sourceLineNo">567</span>   * table. This<a name="line.567"></a>
+<span class="sourceLineNo">568</span>   * &lt;ul&gt;<a name="line.568"></a>
+<span class="sourceLineNo">569</span>   *   &lt;li&gt;Inspects the table to configure a total order partitioner&lt;/li&gt;<a name="line.569"></a>
+<span class="sourceLineNo">570</span>   *   &lt;li&gt;Uploads the partitions file to the cluster and adds it to the DistributedCache&lt;/li&gt;<a name="line.570"></a>
+<span class="sourceLineNo">571</span>   *   &lt;li&gt;Sets the number of reduce tasks to match the current number of regions&lt;/li&gt;<a name="line.571"></a>
+<span class="sourceLineNo">572</span>   *   &lt;li&gt;Sets the output key/value class to match HFileOutputFormat2's requirements&lt;/li&gt;<a name="line.572"></a>
+<span class="sourceLineNo">573</span>   *   &lt;li&gt;Sets the reducer up to perform the appropriate sorting (either KeyValueSortReducer or<a name="line.573"></a>
+<span class="sourceLineNo">574</span>   *     PutSortReducer)&lt;/li&gt;<a name="line.574"></a>
+<span class="sourceLineNo">575</span>   * &lt;/ul&gt;<a name="line.575"></a>
+<span class="sourceLineNo">576</span>   * The user should be sure to set the map output value class to either KeyValue or Put before<a name="line.576"></a>
+<span class="sourceLineNo">577</span>   * running this function.<a name="line.577"></a>
+<span class="sourceLineNo">578</span>   */<a name="line.578"></a>
+<span class="sourceLineNo">579</span>  public static void configureIncrementalLoad(Job job, TableDescriptor tableDescriptor,<a name="line.579"></a>
+<span class="sourceLineNo">580</span>      RegionLocator regionLocator) throws IOException {<a name="line.580"></a>
+<span class="sourceLineNo">581</span>    ArrayList&lt;TableInfo&gt; singleTableInfo = new ArrayList&lt;&gt;();<a name="line.581"></a>
+<span class="sourceLineNo">582</span>    singleTableInfo.add(new TableInfo(tableDescriptor, regionLocator));<a name="line.582"></a>
+<span class="sourceLineNo">583</span>    configureIncrementalLoad(job, singleTableInfo, HFileOutputFormat2.class);<a name="line.583"></a>
+<span class="sourceLineNo">584</span>  }<a name="line.584"></a>
+<span class="sourceLineNo">585</span><a name="line.585"></a>
+<span class="sourceLineNo">586</span>  static void configureIncrementalLoad(Job job, List&lt;TableInfo&gt; multiTableInfo,<a name="line.586"></a>
+<span class="sourceLineNo">587</span>      Class&lt;? extends OutputFormat&lt;?, ?&gt;&gt; cls) throws IOException {<a name="line.587"></a>
+<span class="sourceLineNo">588</span>    Configuration conf = job.getConfiguration();<a name="line.588"></a>
+<span class="sourceLineNo">589</span>    job.setOutputKeyClass(ImmutableBytesWritable.class);<a name="line.589"></a>
+<span class="sourceLineNo">590</span>    job.setOutputValueClass(MapReduceCell.class);<a name="line.590"></a>
+<span class="sourceLineNo">591</span>    job.setOutputFormatClass(cls);<a name="line.591"></a>
+<span class="sourceLineNo">592</span><a name="line.592"></a>
+<span class="sourceLineNo">593</span>    if (multiTableInfo.stream().distinct().count() != multiTableInfo.size()) {<a name="line.593"></a>
+<span class="sourceLineNo">594</span>      throw new IllegalArgumentException("Duplicate entries found in TableInfo argument");<a name="line.594"></a>
+<span class="sourceLineNo">595</span>    }<a name="line.595"></a>
+<span class="sourceLineNo">596</span>    boolean writeMultipleTables = false;<a name="line.596"></a>
+<span class="sourceLineNo">597</span>    if (MultiTableHFileOutputFormat.class.equals(cls)) {<a name="line.597"></a>
+<span class="sourceLineNo">598</span>      writeMultipleTables = true;<a name="line.598"></a>
+<span class="sourceLineNo">599</span>      conf.setBoolean(MULTI_TABLE_HFILEOUTPUTFORMAT_CONF_KEY, true);<a name="line.599"></a>
+<span class="sourceLineNo">600</span>    }<a name="line.600"></a>
+<span class="sourceLineNo">601</span>    // Based on the configured map output class, set the correct reducer to properly<a name="line.601"></a>
+<span class="sourceLineNo">602</span>    // sort the incoming values.<a name="line.602"></a>
+<span class="sourceLineNo">603</span>    // TODO it would be nice to pick one or the other of these formats.<a name="line.603"></a>
+<span class="sourceLineNo">604</span>    if (KeyValue.class.equals(job.getMapOutputValueClass())<a name="line.604"></a>
+<span class="sourceLineNo">605</span>        || MapReduceCell.class.equals(job.getMapOutputValueClass())) {<a name="line.605"></a>
+<span class="sourceLineNo">606</span>      job.setReducerClass(CellSortReducer.class);<a name="line.606"></a>
+<span class="sourceLineNo">607</span>    } else if (Put.class.equals(job.getMapOutputValueClass())) {<a name="line.607"></a>
+<span class="sourceLineNo">608</span>      job.setReducerClass(PutSortReducer.class);<a name="line.608"></a>
+<span class="sourceLineNo">609</span>    } else if (Text.class.equals(job.getMapOutputValueClass())) {<a name="line.609"></a>
+<span class="sourceLineNo">610</span>      job.setReducerClass(TextSortReducer.class);<a name="line.610"></a>
+<span class="sourceLineNo">611</span>    } else {<a name="line.611"></a>
+<span class="sourceLineNo">612</span>      LOG.warn("Unknown map output value type:" + job.getMapOutputValueClass());<a name="line.612"></a>
+<span class="sourceLineNo">613</span>    }<a name="line.613"></a>
+<span class="sourceLineNo">614</span><a name="line.614"></a>
+<span class="sourceLineNo">615</span>    conf.setStrings("io.serializations", conf.get("io.serializations"),<a name="line.615"></a>
+<span class="sourceLineNo">616</span>        MutationSerialization.class.getName(), ResultSerialization.class.getName(),<a name="line.616"></a>
+<span class="sourceLineNo">617</span>        CellSerialization.class.getName());<a name="line.617"></a>
+<span class="sourceLineNo">618</span><a name="line.618"></a>
+<span class="sourceLineNo">619</span>    if (conf.getBoolean(LOCALITY_SENSITIVE_CONF_KEY, DEFAULT_LOCALITY_SENSITIVE)) {<a name="line.619"></a>
+<span class="sourceLineNo">620</span>      LOG.info("bulkload locality sensitive enabled");<a name="line.620"></a>
+<span class="sourceLineNo">621</span>    }<a name="line.621"></a>
+<span class="sourceLineNo">622</span><a name="line.622"></a>
+<span class="sourceLineNo">623</span>    /* Now get the region start keys for every table required */<a name="line.623"></a>
+<span class="sourceLineNo">624</span>    List&lt;String&gt; allTableNames = new ArrayList&lt;&gt;(multiTableInfo.size());<a name="line.624"></a>
+<span class="sourceLineNo">625</span>    List&lt;RegionLocator&gt; regionLocators = new ArrayList&lt;&gt;( multiTableInfo.size());<a name="line.625"></a>
+<span class="sourceLineNo">626</span>    List&lt;TableDescriptor&gt; tableDescriptors = new ArrayList&lt;&gt;( multiTableInfo.size());<a name="line.626"></a>
+<span class="sourceLineNo">627</span><a name="line.627"></a>
+<span class="sourceLineNo">628</span>    for( TableInfo tableInfo : multiTableInfo )<a name="line.628"></a>
+<span class="sourceLineNo">629</span>    {<a name="line.629"></a>
+<span class="sourceLineNo">630</span>      regionLocators.add(tableInfo.getRegionLocator());<a name="line.630"></a>
+<span class="sourceLineNo">631</span>      allTableNames.add(tableInfo.getRegionLocator().getName().getNameAsString());<a name="line.631"></a>
+<span class="sourceLineNo">632</span>      tableDescriptors.add(tableInfo.getTableDescriptor());<a name="line.632"></a>
+<span class="sourceLineNo">633</span>    }<a name="line.633"></a>
+<span class="sourceLineNo">634</span>    // Record tablenames for creating writer by favored nodes, and decoding compression, block size and other attributes of columnfamily per table<a name="line.634"></a>
+<span class="sourceLineNo">635</span>    conf.set(OUTPUT_TABLE_NAME_CONF_KEY, StringUtils.join(allTableNames, Bytes<a name="line.635"></a>
+<span class="sourceLineNo">636</span>            .toString(tableSeparator)));<a name="line.636"></a>
+<span class="sourceLineNo">637</span>    List&lt;ImmutableBytesWritable&gt; startKeys = getRegionStartKeys(regionLocators, writeMultipleTables);<a name="line.637"></a>
+<span class="sourceLineNo">638</span>    // Use table's region boundaries for TOP split points.<a name="line.638"></a>
+<span class="sourceLineNo">639</span>    LOG.info("Configuring " + startKeys.size() + " reduce partitions " +<a name="line.639"></a>
+<span class="sourceLineNo">640</span>        "to match current region count for all tables");<a name="line.640"></a>
+<span class="sourceLineNo">641</span>    job.setNumReduceTasks(startKeys.size());<a name="line.641"></a>
+<span class="sourceLineNo">642</span><a name="line.642"></a>
+<span class="sourceLineNo">643</span>    configurePartitioner(job, startKeys, writeMultipleTables);<a name="line.643"></a>
+<span class="sourceLineNo">644</span>    // Set compression algorithms based on column families<a name="line.644"></a>
+<span class="sourceLineNo">645</span><a name="line.645"></a>
+<span class="sourceLineNo">646</span>    conf.set(COMPRESSION_FAMILIES_CONF_KEY, serializeColumnFamilyAttribute(compressionDetails,<a name="line.646"></a>
+<span class="sourceLineNo">647</span>            tableDescriptors));<a name="line.647"></a>
+<span class="sourceLineNo">648</span>    conf.set(BLOCK_SIZE_FAMILIES_CONF_KEY, serializeColumnFamilyAttribute(blockSizeDetails,<a name="line.648"></a>
+<span class="sourceLineNo">649</span>            tableDescriptors));<a name="line.649"></a>
+<span class="sourceLineNo">650</span>    conf.set(BLOOM_TYPE_FAMILIES_CONF_KEY, serializeColumnFamilyAttribute(bloomTypeDetails,<a name="line.650"></a>
+<span class="sourceLineNo">651</span>            tableDescriptors));<a name="line.651"></a>
+<span class="sourceLineNo">652</span>    conf.set(DATABLOCK_ENCODING_FAMILIES_CONF_KEY,<a name="line.652"></a>
+<span class="sourceLineNo">653</span>            serializeColumnFamilyAttribute(dataBlockEncodingDetails, tableDescriptors));<a name="line.653"></a>
+<span class="sourceLineNo">654</span><a name="line.654"></a>
+<span class="sourceLineNo">655</span>    TableMapReduceUtil.addDependencyJars(job);<a name="line.655"></a>
+<span class="sourceLineNo">656</span>    TableMapReduceUtil.initCredentials(job);<a name="line.656"></a>
+<span class="sourceLineNo">657</span>    LOG.info("Incremental output configured for tables: " + StringUtils.join(allTableNames, ","));<a name="line.657"></a>
+<span class="sourceLineNo">658</span>  }<a name="line.658"></a>
+<span class="sourceLineNo">659</span><a name="line.659"></a>
+<span class="sourceLineNo">660</span>  public static void configureIncrementalLoadMap(Job job, TableDescriptor tableDescriptor) throws<a name="line.660"></a>
+<span class="sourceLineNo">661</span>      IOException {<a name="line.661"></a>
+<span class="sourceLineNo">662</span>    Configuration conf = job.getConfiguration();<a name="line.662"></a>
+<span class="sourceLineNo">663</span><a name="line.663"></a>
+<span class="sourceLineNo">664</span>    job.setOutputKeyClass(ImmutableBytesWritable.class);<a name="line.664"></a>
+<span class="sourceLineNo">665</span>    job.setOutputValueClass(MapReduceCell.class);<a name="line.665"></a>
+<span class="sourceLineNo">666</span>    job.setOutputFormatClass(HFileOutputFormat2.class);<a name="line.666"></a>
+<span class="sourceLineNo">667</span><a name="line.667"></a>
+<span class="sourceLineNo">668</span>    ArrayList&lt;TableDescriptor&gt; singleTableDescriptor = new ArrayList&lt;&gt;(1);<a name="line.668"></a>
+<span class="sourceLineNo">669</span>    singleTableDescriptor.add(tableDescriptor);<a name="line.669"></a>
+<span class="sourceLineNo">670</span><a name="line.670"></a>
+<span class="sourceLineNo">671</span>    conf.set(OUTPUT_TABLE_NAME_CONF_KEY, tableDescriptor.getTableName().getNameAsString());<a name="line.671"></a>
+<span class="sourceLineNo">672</span>    // Set compression algorithms based on column families<a name="line.672"></a>
+<span class="sourceLineNo">673</span>    conf.set(COMPRESSION_FAMILIES_CONF_KEY,<a name="line.673"></a>
+<span class="sourceLineNo">674</span>        serializeColumnFamilyAttribute(compressionDetails, singleTableDescriptor));<a name="line.674"></a>
+<span class="sourceLineNo">675</span>    conf.set(BLOCK_SIZE_FAMILIES_CONF_KEY,<a name="line.675"></a>
+<span class="sourceLineNo">676</span>        serializeColumnFamilyAttribute(blockSizeDetails, singleTableDescriptor));<a name="line.676"></a>
+<span class="sourceLineNo">677</span>    conf.set(BLOOM_TYPE_FAMILIES_CONF_KEY,<a name="line.677"></a>
+<span class="sourceLineNo">678</span>        serializeColumnFamilyAttribute(bloomTypeDetails, singleTableDescriptor));<a name="line.678"></a>
+<span class="sourceLineNo">679</span>    conf.set(DATABLOCK_ENCODING_FAMILIES_CONF_KEY,<a name="line.679"></a>
+<span class="sourceLineNo">680</span>        serializeColumnFamilyAttribute(dataBlockEncodingDetails, singleTableDescriptor));<a name="line.680"></a>
+<span class="sourceLineNo">681</span><a name="line.681"></a>
+<span class="sourceLineNo">682</span>    TableMapReduceUtil.addDependencyJars(job);<a name="line.682"></a>
+<span class="sourceLineNo">683</span>    TableMapReduceUtil.initCredentials(job);<a name="line.683"></a>
+<span class="sourceLineNo">684</span>    LOG.info("Incremental table " + tableDescriptor.getTableName() + " output configured.");<a name="line.684"></a>
+<span class="sourceLineNo">685</span>  }<a name="line.685"></a>
+<span class="sourceLineNo">686</span><a name="line.686"></a>
+<span class="sourceLineNo">687</span>  /**<a name="line.687"></a>
+<span class="sourceLineNo">688</span>   * Runs inside the task to deserialize column family to compression algorithm<a name="line.688"></a>
+<span class="sourceLineNo">689</span>   * map from the configuration.<a name="line.689"></a>
+<span class="sourceLineNo">690</span>   *<a name="line.690"></a>
+<span class="sourceLineNo">691</span>   * @param conf to read the serialized values from<a name="line.691"></a>
+<span class="sourceLineNo">692</span>   * @return a map from column family to the configured compression algorithm<a name="line.692"></a>
+<span class="sourceLineNo">693</span>   */<a name="line.693"></a>
+<span class="sourceLineNo">694</span>  @VisibleForTesting<a name="line.694"></a>
+<span class="sourceLineNo">695</span>  static Map&lt;byte[], Algorithm&gt; createFamilyCompressionMap(Configuration<a name="line.695"></a>
+<span class="sourceLineNo">696</span>      conf) {<a name="line.696"></a>
+<span class="sourceLineNo">697</span>    Map&lt;byte[], String&gt; stringMap = createFamilyConfValueMap(conf,<a name="line.697"></a>
+<span class="sourceLineNo">698</span>        COMPRESSION_FAMILIES_CONF_KEY);<a name="line.698"></a>
+<span class="sourceLineNo">699</span>    Map&lt;byte[], Algorithm&gt; compressionMap = new TreeMap&lt;&gt;(Bytes.BYTES_COMPARATOR);<a name="line.699"></a>
+<span class="sourceLineNo">700</span>    for (Map.Entry&lt;byte[], String&gt; e : stringMap.entrySet()) {<a name="line.700"></a>
+<span class="sourceLineNo">701</span>      Algorithm algorithm = HFileWriterImpl.compressionByName(e.getValue());<a name="line.701"></a>
+<span class="sourceLineNo">702</span>      compressionMap.put(e.getKey(), algorithm);<a name="line.702"></a>
+<span class="sourceLineNo">703</span>    }<a name="line.703"></a>
+<span class="sourceLineNo">704</span>    return compressionMap;<a name="line.704"></a>
+<span class="sourceLineNo">705</span>  }<a name="line.705"></a>
+<span class="sourceLineNo">706</span><a name="line.706"></a>
+<span class="sourceLineNo">707</span>  /**<a name="line.707"></a>
+<span class="sourceLineNo">708</span>   * Runs inside the task to deserialize column family to bloom filter type<a name="line.708"></a>
+<span class="sourceLineNo">709</span>   * map from the configuration.<a name="line.709"></a>
+<span class="sourceLineNo">710</span>   *<a name="line.710"></a>
+<span class="sourceLineNo">711</span>   * @param conf to read the serialized values from<a name="line.711"></a>
+<span class="sourceLineNo">712</span>   * @return a map from column family to the the configured bloom filter type<a name="line.712"></a>
+<span class="sourceLineNo">713</span>   */<a name="line.713"></a>
+<span class="sourceLineNo">714</span>  @VisibleForTesting<a name="line.714"></a>
+<span class="sourceLineNo">715</span>  static Map&lt;byte[], BloomType&gt; createFamilyBloomTypeMap(Configuration conf) {<a name="line.715"></a>
+<span class="sourceLineNo">716</span>    Map&lt;byte[], String&gt; stringMap = createFamilyConfValueMap(conf,<a name="line.716"></a>
+<span class="sourceLineNo">717</span>        BLOOM_TYPE_FAMILIES_CONF_KEY);<a name="line.717"></a>
+<span class="sourceLineNo">718</span>    Map&lt;byte[], BloomType&gt; bloomTypeMap = new TreeMap&lt;&gt;(Bytes.BYTES_COMPARATOR);<a name="line.718"></a>
+<span class="sourceLineNo">719</span>    for (Map.Entry&lt;byte[], String&gt; e : stringMap.entrySet()) {<a name="line.719"></a>
+<span class="sourceLineNo">720</span>      BloomType bloomType = BloomType.valueOf(e.getValue());<a name="line.720"></a>
+<span class="sourceLineNo">721</span>      bloomTypeMap.put(e.getKey(), bloomType);<a name="line.721"></a>
+<span class="sourceLineNo">722</span>    }<a name="line.722"></a>
+<span class="sourceLineNo">723</span>    return bloomTypeMap;<a name="line.723"></a>
+<span class="sourceLineNo">724</span>  }<a name="line.724"></a>
+<span class="sourceLineNo">725</span><a name="line.725"></a>
+<span class="sourceLineNo">726</span>  /**<a name="line.726"></a>
+<span class="sourceLineNo">727</span>   * Runs inside the task to deserialize column family to block size<a name="line.727"></a>
+<span class="sourceLineNo">728</span>   * map from the configuration.<a name="line.728"></a>
+<span class="sourceLineNo">729</span>   *<a name="line.729"></a>
+<span class="sourceLineNo">730</span>   * @param conf to read the serialized values from<a name="line.730"></a>
+<span class="sourceLineNo">731</span>   * @return a map from column family to the configured block size<a name="line.731"></a>
+<span class="sourceLineNo">732</span>   */<a name="line.732"></a>
+<span class="sourceLineNo">733</span>  @VisibleForTesting<a name="line.733"></a>
+<span class="sourceLineNo">734</span>  static Map&lt;byte[], Integer&gt; createFamilyBlockSizeMap(Configuration conf) {<a name="line.734"></a>
+<span class="sourceLineNo">735</span>    Map&lt;byte[], String&gt; stringMap = createFamilyConfValueMap(conf,<a name="line.735"></a>
+<span class="sourceLineNo">736</span>        BLOCK_SIZE_FAMILIES_CONF_KEY);<a name="line.736"></a>
+<span class="sourceLineNo">737</span>    Map&lt;byte[], Integer&gt; blockSizeMap = new TreeMap&lt;&gt;(Bytes.BYTES_COMPARATOR);<a name="line.737"></a>
+<span class="sourceLineNo">738</span>    for (Map.Entry&lt;byte[], String&gt; e : stringMap.entrySet()) {<a name="line.738"></a>
+<span class="sourceLineNo">739</span>      Integer blockSize = Integer.parseInt(e.getValue());<a name="line.739"></a>
+<span class="sourceLineNo">740</span>      blockSizeMap.put(e.getKey(), blockSize);<a name="line.740"></a>
+<span class="sourceLineNo">741</span>    }<a name="line.741"></a>
+<span class="sourceLineNo">742</span>    return blockSizeMap;<a name="line.742"></a>
+<span class="sourceLineNo">743</span>  }<a name="line.743"></a>
+<span class="sourceLineNo">744</span><a name="line.744"></a>
+<span class="sourceLineNo">745</span>  /**<a name="line.745"></a>
+<span class="sourceLineNo">746</span>   * Runs inside the task to deserialize column family to data block encoding<a name="line.746"></a>
+<span class="sourceLineNo">747</span>   * type map from the configuration.<a name="line.747"></a>
+<span class="sourceLineNo">748</span>   *<a name="line.748"></a>
+<span class="sourceLineNo">749</span>   * @param conf to read the serialized values from<a name="line.749"></a>
+<span class="sourceLineNo">750</span>   * @return a map from column family to HFileDataBlockEncoder for the<a name="line.750"></a>
+<span class="sourceLineNo">751</span>   * 

<TRUNCATED>

Mime
View raw message