View Javadoc

1   /**
2    *
3    * Licensed to the Apache Software Foundation (ASF) under one
4    * or more contributor license agreements.  See the NOTICE file
5    * distributed with this work for additional information
6    * regarding copyright ownership.  The ASF licenses this file
7    * to you under the Apache License, Version 2.0 (the
8    * "License"); you may not use this file except in compliance
9    * with the License.  You may obtain a copy of the License at
10   *
11   *     http://www.apache.org/licenses/LICENSE-2.0
12   *
13   * Unless required by applicable law or agreed to in writing, software
14   * distributed under the License is distributed on an "AS IS" BASIS,
15   * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
16   * See the License for the specific language governing permissions and
17   * limitations under the License.
18   */
19  package org.apache.hadoop.hbase.mapreduce;
20  
21  import static java.lang.String.format;
22  
23  import java.io.FileNotFoundException;
24  import java.io.IOException;
25  import java.io.InterruptedIOException;
26  import java.nio.ByteBuffer;
27  import java.util.ArrayList;
28  import java.util.Arrays;
29  import java.util.Collection;
30  import java.util.Deque;
31  import java.util.HashMap;
32  import java.util.HashSet;
33  import java.util.LinkedList;
34  import java.util.List;
35  import java.util.Map;
36  import java.util.Map.Entry;
37  import java.util.Set;
38  import java.util.TreeMap;
39  import java.util.concurrent.Callable;
40  import java.util.concurrent.ExecutionException;
41  import java.util.concurrent.ExecutorService;
42  import java.util.concurrent.Future;
43  import java.util.concurrent.LinkedBlockingQueue;
44  import java.util.concurrent.ThreadPoolExecutor;
45  import java.util.concurrent.TimeUnit;
46  import java.util.concurrent.atomic.AtomicLong;
47  
48  import org.apache.commons.lang.mutable.MutableInt;
49  import org.apache.commons.logging.Log;
50  import org.apache.commons.logging.LogFactory;
51  import org.apache.hadoop.hbase.classification.InterfaceAudience;
52  import org.apache.hadoop.hbase.classification.InterfaceStability;
53  import org.apache.hadoop.conf.Configuration;
54  import org.apache.hadoop.conf.Configured;
55  import org.apache.hadoop.fs.permission.FsPermission;
56  import org.apache.hadoop.fs.FileStatus;
57  import org.apache.hadoop.fs.FileSystem;
58  import org.apache.hadoop.fs.FileUtil;
59  import org.apache.hadoop.fs.Path;
60  import org.apache.hadoop.hbase.HBaseConfiguration;
61  import org.apache.hadoop.hbase.HColumnDescriptor;
62  import org.apache.hadoop.hbase.HConstants;
63  import org.apache.hadoop.hbase.HTableDescriptor;
64  import org.apache.hadoop.hbase.KeyValue;
65  import org.apache.hadoop.hbase.TableName;
66  import org.apache.hadoop.hbase.TableNotFoundException;
67  import org.apache.hadoop.hbase.client.HBaseAdmin;
68  import org.apache.hadoop.hbase.client.HConnection;
69  import org.apache.hadoop.hbase.client.HTable;
70  import org.apache.hadoop.hbase.client.RegionServerCallable;
71  import org.apache.hadoop.hbase.client.RpcRetryingCallerFactory;
72  import org.apache.hadoop.hbase.client.coprocessor.SecureBulkLoadClient;
73  import org.apache.hadoop.hbase.coprocessor.CoprocessorHost;
74  import org.apache.hadoop.hbase.io.HFileLink;
75  import org.apache.hadoop.hbase.io.HalfStoreFileReader;
76  import org.apache.hadoop.hbase.io.Reference;
77  import org.apache.hadoop.hbase.io.compress.Compression.Algorithm;
78  import org.apache.hadoop.hbase.io.hfile.CacheConfig;
79  import org.apache.hadoop.hbase.io.hfile.HFile;
80  import org.apache.hadoop.hbase.io.hfile.HFileContext;
81  import org.apache.hadoop.hbase.io.hfile.HFileContextBuilder;
82  import org.apache.hadoop.hbase.io.hfile.HFileScanner;
83  import org.apache.hadoop.hbase.protobuf.ProtobufUtil;
84  import org.apache.hadoop.hbase.regionserver.BloomType;
85  import org.apache.hadoop.hbase.regionserver.HStore;
86  import org.apache.hadoop.hbase.regionserver.StoreFile;
87  import org.apache.hadoop.hbase.security.UserProvider;
88  import org.apache.hadoop.hbase.security.token.FsDelegationToken;
89  import org.apache.hadoop.hbase.util.Bytes;
90  import org.apache.hadoop.hbase.util.FSHDFSUtils;
91  import org.apache.hadoop.hbase.util.Pair;
92  import org.apache.hadoop.util.Tool;
93  import org.apache.hadoop.util.ToolRunner;
94  
95  import com.google.common.collect.HashMultimap;
96  import com.google.common.collect.Multimap;
97  import com.google.common.collect.Multimaps;
98  import com.google.common.util.concurrent.ThreadFactoryBuilder;
99  
100 import java.util.UUID;
101 
102 /**
103  * Tool to load the output of HFileOutputFormat into an existing table.
104  * @see #usage()
105  */
106 @InterfaceAudience.Public
107 @InterfaceStability.Stable
108 public class LoadIncrementalHFiles extends Configured implements Tool {
109   private static final Log LOG = LogFactory.getLog(LoadIncrementalHFiles.class);
110   private HBaseAdmin hbAdmin;
111 
112   public static final String NAME = "completebulkload";
113   public static final String MAX_FILES_PER_REGION_PER_FAMILY
114     = "hbase.mapreduce.bulkload.max.hfiles.perRegion.perFamily";
115   private static final String ASSIGN_SEQ_IDS = "hbase.mapreduce.bulkload.assign.sequenceNumbers";
116   public final static String CREATE_TABLE_CONF_KEY = "create.table";
117 
118   private int maxFilesPerRegionPerFamily;
119   private boolean assignSeqIds;
120 
121   // Source filesystem
122   private FileSystem fs;
123   // Source delegation token
124   private FsDelegationToken fsDelegationToken;
125   private String bulkToken;
126   private UserProvider userProvider;
127 
128   private LoadIncrementalHFiles() {}
129 
130   public LoadIncrementalHFiles(Configuration conf) throws Exception {
131     super(conf);
132     initialize();
133   }
134 
135   private void initialize() throws Exception {
136     if (hbAdmin == null) {
137       // make a copy, just to be sure we're not overriding someone else's config
138       setConf(HBaseConfiguration.create(getConf()));
139       Configuration conf = getConf();
140       // disable blockcache for tool invocation, see HBASE-10500
141       conf.setFloat(HConstants.HFILE_BLOCK_CACHE_SIZE_KEY, 0);
142       this.hbAdmin = new HBaseAdmin(conf);
143       this.userProvider = UserProvider.instantiate(conf);
144       this.fsDelegationToken = new FsDelegationToken(userProvider, "renewer");
145       assignSeqIds = conf.getBoolean(ASSIGN_SEQ_IDS, true);
146       maxFilesPerRegionPerFamily = conf.getInt(MAX_FILES_PER_REGION_PER_FAMILY, 32);
147     }
148   }
149 
150   private void usage() {
151     System.err.println("usage: " + NAME + " /path/to/hfileoutputformat-output tablename" + "\n -D"
152         + CREATE_TABLE_CONF_KEY + "=no - can be used to avoid creation of table by this tool\n"
153         + "  Note: if you set this to 'no', then the target table must already exist in HBase\n"
154         + "\n");
155   }
156 
157   /**
158    * Represents an HFile waiting to be loaded. An queue is used
159    * in this class in order to support the case where a region has
160    * split during the process of the load. When this happens,
161    * the HFile is split into two physical parts across the new
162    * region boundary, and each part is added back into the queue.
163    * The import process finishes when the queue is empty.
164    */
165   static class LoadQueueItem {
166     final byte[] family;
167     final Path hfilePath;
168 
169     public LoadQueueItem(byte[] family, Path hfilePath) {
170       this.family = family;
171       this.hfilePath = hfilePath;
172     }
173 
174     public String toString() {
175       return "family:"+ Bytes.toString(family) + " path:" + hfilePath.toString();
176     }
177   }
178 
179   /**
180    * Walk the given directory for all HFiles, and return a Queue
181    * containing all such files.
182    */
183   private void discoverLoadQueue(Deque<LoadQueueItem> ret, Path hfofDir)
184   throws IOException {
185     fs = hfofDir.getFileSystem(getConf());
186 
187     if (!fs.exists(hfofDir)) {
188       throw new FileNotFoundException("HFileOutputFormat dir " +
189           hfofDir + " not found");
190     }
191 
192     FileStatus[] familyDirStatuses = fs.listStatus(hfofDir);
193     if (familyDirStatuses == null) {
194       throw new FileNotFoundException("No families found in " + hfofDir);
195     }
196 
197     for (FileStatus stat : familyDirStatuses) {
198       if (!stat.isDir()) {
199         LOG.warn("Skipping non-directory " + stat.getPath());
200         continue;
201       }
202       Path familyDir = stat.getPath();
203       if (familyDir.getName().equals("_logs")) {
204         // Family name of "_logs" is not supported.
205         // This is due to the presence of history directory under _logs directory
206         // when hadoop-1 is used
207         continue;
208       }
209       byte[] family = familyDir.getName().getBytes();
210       Path[] hfiles = FileUtil.stat2Paths(fs.listStatus(familyDir));
211       for (Path hfile : hfiles) {
212         if (hfile.getName().startsWith("_")) continue;
213         ret.add(new LoadQueueItem(family, hfile));
214       }
215     }
216   }
217 
218   /**
219    * Perform a bulk load of the given directory into the given
220    * pre-existing table.  This method is not threadsafe.
221    *
222    * @param hfofDir the directory that was provided as the output path
223    * of a job using HFileOutputFormat
224    * @param table the table to load into
225    * @throws TableNotFoundException if table does not yet exist
226    */
227   @SuppressWarnings("deprecation")
228   public void doBulkLoad(Path hfofDir, final HTable table)
229     throws TableNotFoundException, IOException
230   {
231     final HConnection conn = table.getConnection();
232 
233     if (!conn.isTableAvailable(table.getName())) {
234       throw new TableNotFoundException("Table " +
235           Bytes.toStringBinary(table.getTableName()) +
236           "is not currently available.");
237     }
238 
239     // initialize thread pools
240     int nrThreads = getConf().getInt("hbase.loadincremental.threads.max",
241       Runtime.getRuntime().availableProcessors());
242     ThreadFactoryBuilder builder = new ThreadFactoryBuilder();
243     builder.setNameFormat("LoadIncrementalHFiles-%1$d");
244     ExecutorService pool = new ThreadPoolExecutor(nrThreads, nrThreads,
245         60, TimeUnit.SECONDS,
246         new LinkedBlockingQueue<Runnable>(),
247         builder.build());
248     ((ThreadPoolExecutor)pool).allowCoreThreadTimeOut(true);
249 
250     // LQI queue does not need to be threadsafe -- all operations on this queue
251     // happen in this thread
252     Deque<LoadQueueItem> queue = new LinkedList<LoadQueueItem>();
253     try {
254       discoverLoadQueue(queue, hfofDir);
255       // check whether there is invalid family name in HFiles to be bulkloaded
256       Collection<HColumnDescriptor> families = table.getTableDescriptor().getFamilies();
257       ArrayList<String> familyNames = new ArrayList<String>();
258       for (HColumnDescriptor family : families) {
259         familyNames.add(family.getNameAsString());
260       }
261       ArrayList<String> unmatchedFamilies = new ArrayList<String>();
262       for (LoadQueueItem lqi : queue) {
263         String familyNameInHFile = Bytes.toString(lqi.family);
264         if (!familyNames.contains(familyNameInHFile)) {
265           unmatchedFamilies.add(familyNameInHFile);
266         }
267       }
268       if (unmatchedFamilies.size() > 0) {
269         String msg =
270             "Unmatched family names found: unmatched family names in HFiles to be bulkloaded: "
271                 + unmatchedFamilies + "; valid family names of table "
272                 + Bytes.toString(table.getTableName()) + " are: " + familyNames;
273         LOG.error(msg);
274         throw new IOException(msg);
275       }
276       int count = 0;
277 
278       if (queue.isEmpty()) {
279         LOG.warn("Bulk load operation did not find any files to load in " +
280             "directory " + hfofDir.toUri() + ".  Does it contain files in " +
281             "subdirectories that correspond to column family names?");
282         return;
283       }
284 
285       //If using secure bulk load, get source delegation token, and
286       //prepare staging directory and token
287       // fs is the source filesystem
288       fsDelegationToken.acquireDelegationToken(fs);
289       if(isSecureBulkLoadEndpointAvailable()) {
290         bulkToken = new SecureBulkLoadClient(table).prepareBulkLoad(table.getName());
291       }
292 
293       // Assumes that region splits can happen while this occurs.
294       while (!queue.isEmpty()) {
295         // need to reload split keys each iteration.
296         final Pair<byte[][], byte[][]> startEndKeys = table.getStartEndKeys();
297         if (count != 0) {
298           LOG.info("Split occured while grouping HFiles, retry attempt " +
299               + count + " with " + queue.size() + " files remaining to group or split");
300         }
301 
302         int maxRetries = getConf().getInt("hbase.bulkload.retries.number", 0);
303         if (maxRetries != 0 && count >= maxRetries) {
304           throw new IOException("Retry attempted " + count +
305             " times without completing, bailing out");
306         }
307         count++;
308 
309         // Using ByteBuffer for byte[] equality semantics
310         Multimap<ByteBuffer, LoadQueueItem> regionGroups = groupOrSplitPhase(table,
311             pool, queue, startEndKeys);
312 
313         if (!checkHFilesCountPerRegionPerFamily(regionGroups)) {
314           // Error is logged inside checkHFilesCountPerRegionPerFamily.
315           throw new IOException("Trying to load more than " + maxFilesPerRegionPerFamily
316             + " hfiles to one family of one region");
317         }
318 
319         bulkLoadPhase(table, conn, pool, queue, regionGroups);
320 
321         // NOTE: The next iteration's split / group could happen in parallel to
322         // atomic bulkloads assuming that there are splits and no merges, and
323         // that we can atomically pull out the groups we want to retry.
324       }
325 
326     } finally {
327       fsDelegationToken.releaseDelegationToken();
328       if(bulkToken != null) {
329         new SecureBulkLoadClient(table).cleanupBulkLoad(bulkToken);
330       }
331       pool.shutdown();
332       if (queue != null && !queue.isEmpty()) {
333         StringBuilder err = new StringBuilder();
334         err.append("-------------------------------------------------\n");
335         err.append("Bulk load aborted with some files not yet loaded:\n");
336         err.append("-------------------------------------------------\n");
337         for (LoadQueueItem q : queue) {
338           err.append("  ").append(q.hfilePath).append('\n');
339         }
340         LOG.error(err);
341       }
342     }
343 
344     if (queue != null && !queue.isEmpty()) {
345         throw new RuntimeException("Bulk load aborted with some files not yet loaded."
346           + "Please check log for more details.");
347     }
348   }
349 
350   /**
351    * This takes the LQI's grouped by likely regions and attempts to bulk load
352    * them.  Any failures are re-queued for another pass with the
353    * groupOrSplitPhase.
354    */
355   protected void bulkLoadPhase(final HTable table, final HConnection conn,
356       ExecutorService pool, Deque<LoadQueueItem> queue,
357       final Multimap<ByteBuffer, LoadQueueItem> regionGroups) throws IOException {
358     // atomically bulk load the groups.
359     Set<Future<List<LoadQueueItem>>> loadingFutures = new HashSet<Future<List<LoadQueueItem>>>();
360     for (Entry<ByteBuffer, ? extends Collection<LoadQueueItem>> e: regionGroups.asMap().entrySet()) {
361       final byte[] first = e.getKey().array();
362       final Collection<LoadQueueItem> lqis =  e.getValue();
363 
364       final Callable<List<LoadQueueItem>> call = new Callable<List<LoadQueueItem>>() {
365         public List<LoadQueueItem> call() throws Exception {
366           List<LoadQueueItem> toRetry =
367               tryAtomicRegionLoad(conn, table.getName(), first, lqis);
368           return toRetry;
369         }
370       };
371       loadingFutures.add(pool.submit(call));
372     }
373 
374     // get all the results.
375     for (Future<List<LoadQueueItem>> future : loadingFutures) {
376       try {
377         List<LoadQueueItem> toRetry = future.get();
378 
379         // LQIs that are requeued to be regrouped.
380         queue.addAll(toRetry);
381 
382       } catch (ExecutionException e1) {
383         Throwable t = e1.getCause();
384         if (t instanceof IOException) {
385           // At this point something unrecoverable has happened.
386           // TODO Implement bulk load recovery
387           throw new IOException("BulkLoad encountered an unrecoverable problem", t);
388         }
389         LOG.error("Unexpected execution exception during bulk load", e1);
390         throw new IllegalStateException(t);
391       } catch (InterruptedException e1) {
392         LOG.error("Unexpected interrupted exception during bulk load", e1);
393         throw (InterruptedIOException)new InterruptedIOException().initCause(e1);
394       }
395     }
396   }
397 
398   private boolean checkHFilesCountPerRegionPerFamily(
399       final Multimap<ByteBuffer, LoadQueueItem> regionGroups) {
400     for (Entry<ByteBuffer,
401         ? extends Collection<LoadQueueItem>> e: regionGroups.asMap().entrySet()) {
402       final Collection<LoadQueueItem> lqis =  e.getValue();
403       HashMap<byte[], MutableInt> filesMap = new HashMap<byte[], MutableInt>();
404       for (LoadQueueItem lqi: lqis) {
405         MutableInt count = filesMap.get(lqi.family);
406         if (count == null) {
407           count = new MutableInt();
408           filesMap.put(lqi.family, count);
409         }
410         count.increment();
411         if (count.intValue() > maxFilesPerRegionPerFamily) {
412           LOG.error("Trying to load more than " + maxFilesPerRegionPerFamily
413             + " hfiles to family " + Bytes.toStringBinary(lqi.family)
414             + " of region with start key "
415             + Bytes.toStringBinary(e.getKey()));
416           return false;
417         }
418       }
419     }
420     return true;
421   }
422 
423   /**
424    * @return A Multimap<startkey, LoadQueueItem> that groups LQI by likely
425    * bulk load region targets.
426    */
427   private Multimap<ByteBuffer, LoadQueueItem> groupOrSplitPhase(final HTable table,
428       ExecutorService pool, Deque<LoadQueueItem> queue,
429       final Pair<byte[][], byte[][]> startEndKeys) throws IOException {
430     // <region start key, LQI> need synchronized only within this scope of this
431     // phase because of the puts that happen in futures.
432     Multimap<ByteBuffer, LoadQueueItem> rgs = HashMultimap.create();
433     final Multimap<ByteBuffer, LoadQueueItem> regionGroups = Multimaps.synchronizedMultimap(rgs);
434 
435     // drain LQIs and figure out bulk load groups
436     Set<Future<List<LoadQueueItem>>> splittingFutures = new HashSet<Future<List<LoadQueueItem>>>();
437     while (!queue.isEmpty()) {
438       final LoadQueueItem item = queue.remove();
439 
440       final Callable<List<LoadQueueItem>> call = new Callable<List<LoadQueueItem>>() {
441         public List<LoadQueueItem> call() throws Exception {
442           List<LoadQueueItem> splits = groupOrSplit(regionGroups, item, table, startEndKeys);
443           return splits;
444         }
445       };
446       splittingFutures.add(pool.submit(call));
447     }
448     // get all the results.  All grouping and splitting must finish before
449     // we can attempt the atomic loads.
450     for (Future<List<LoadQueueItem>> lqis : splittingFutures) {
451       try {
452         List<LoadQueueItem> splits = lqis.get();
453         if (splits != null) {
454           queue.addAll(splits);
455         }
456       } catch (ExecutionException e1) {
457         Throwable t = e1.getCause();
458         if (t instanceof IOException) {
459           LOG.error("IOException during splitting", e1);
460           throw (IOException)t; // would have been thrown if not parallelized,
461         }
462         LOG.error("Unexpected execution exception during splitting", e1);
463         throw new IllegalStateException(t);
464       } catch (InterruptedException e1) {
465         LOG.error("Unexpected interrupted exception during splitting", e1);
466         throw (InterruptedIOException)new InterruptedIOException().initCause(e1);
467       }
468     }
469     return regionGroups;
470   }
471 
472   // unique file name for the table
473   private String getUniqueName() {
474     return UUID.randomUUID().toString().replaceAll("-", "");
475   }
476 
477   protected List<LoadQueueItem> splitStoreFile(final LoadQueueItem item,
478       final HTable table, byte[] startKey,
479       byte[] splitKey) throws IOException {
480     final Path hfilePath = item.hfilePath;
481 
482     // We use a '_' prefix which is ignored when walking directory trees
483     // above.
484     final Path tmpDir = new Path(item.hfilePath.getParent(), "_tmp");
485 
486     LOG.info("HFile at " + hfilePath + " no longer fits inside a single " +
487     "region. Splitting...");
488 
489     String uniqueName = getUniqueName();
490     HColumnDescriptor familyDesc = table.getTableDescriptor().getFamily(item.family);
491     Path botOut = new Path(tmpDir, uniqueName + ".bottom");
492     Path topOut = new Path(tmpDir, uniqueName + ".top");
493     splitStoreFile(getConf(), hfilePath, familyDesc, splitKey,
494         botOut, topOut);
495 
496     FileSystem fs = tmpDir.getFileSystem(getConf());
497     fs.setPermission(tmpDir, FsPermission.valueOf("-rwxrwxrwx"));
498     fs.setPermission(botOut, FsPermission.valueOf("-rwxrwxrwx"));
499 
500     // Add these back at the *front* of the queue, so there's a lower
501     // chance that the region will just split again before we get there.
502     List<LoadQueueItem> lqis = new ArrayList<LoadQueueItem>(2);
503     lqis.add(new LoadQueueItem(item.family, botOut));
504     lqis.add(new LoadQueueItem(item.family, topOut));
505 
506     LOG.info("Successfully split into new HFiles " + botOut + " and " + topOut);
507     return lqis;
508   }
509 
510   /**
511    * Attempt to assign the given load queue item into its target region group.
512    * If the hfile boundary no longer fits into a region, physically splits
513    * the hfile such that the new bottom half will fit and returns the list of
514    * LQI's corresponding to the resultant hfiles.
515    *
516    * protected for testing
517    * @throws IOException
518    */
519   protected List<LoadQueueItem> groupOrSplit(Multimap<ByteBuffer, LoadQueueItem> regionGroups,
520       final LoadQueueItem item, final HTable table,
521       final Pair<byte[][], byte[][]> startEndKeys)
522       throws IOException {
523     final Path hfilePath = item.hfilePath;
524     HFile.Reader hfr = HFile.createReader(fs, hfilePath,
525         new CacheConfig(getConf()), getConf());
526     final byte[] first, last;
527     try {
528       hfr.loadFileInfo();
529       first = hfr.getFirstRowKey();
530       last = hfr.getLastRowKey();
531     }  finally {
532       hfr.close();
533     }
534 
535     LOG.info("Trying to load hfile=" + hfilePath +
536         " first=" + Bytes.toStringBinary(first) +
537         " last="  + Bytes.toStringBinary(last));
538     if (first == null || last == null) {
539       assert first == null && last == null;
540       // TODO what if this is due to a bad HFile?
541       LOG.info("hfile " + hfilePath + " has no entries, skipping");
542       return null;
543     }
544     if (Bytes.compareTo(first, last) > 0) {
545       throw new IllegalArgumentException(
546       "Invalid range: " + Bytes.toStringBinary(first) +
547       " > " + Bytes.toStringBinary(last));
548     }
549     int idx = Arrays.binarySearch(startEndKeys.getFirst(), first,
550         Bytes.BYTES_COMPARATOR);
551     if (idx < 0) {
552       // not on boundary, returns -(insertion index).  Calculate region it
553       // would be in.
554       idx = -(idx + 1) - 1;
555     }
556     final int indexForCallable = idx;
557 
558     /**
559      * we can consider there is a region hole in following conditions. 1) if idx < 0,then first
560      * region info is lost. 2) if the endkey of a region is not equal to the startkey of the next
561      * region. 3) if the endkey of the last region is not empty.
562      */
563     if (indexForCallable < 0) {
564       throw new IOException("The first region info for table "
565           + Bytes.toString(table.getTableName())
566           + " cann't be found in hbase:meta.Please use hbck tool to fix it first.");
567     } else if ((indexForCallable == startEndKeys.getFirst().length - 1)
568         && !Bytes.equals(startEndKeys.getSecond()[indexForCallable], HConstants.EMPTY_BYTE_ARRAY)) {
569       throw new IOException("The last region info for table "
570           + Bytes.toString(table.getTableName())
571           + " cann't be found in hbase:meta.Please use hbck tool to fix it first.");
572     } else if (indexForCallable + 1 < startEndKeys.getFirst().length
573         && !(Bytes.compareTo(startEndKeys.getSecond()[indexForCallable],
574           startEndKeys.getFirst()[indexForCallable + 1]) == 0)) {
575       throw new IOException("The endkey of one region for table "
576           + Bytes.toString(table.getTableName())
577           + " is not equal to the startkey of the next region in hbase:meta."
578           + "Please use hbck tool to fix it first.");
579     }
580 
581     boolean lastKeyInRange =
582       Bytes.compareTo(last, startEndKeys.getSecond()[idx]) < 0 ||
583       Bytes.equals(startEndKeys.getSecond()[idx], HConstants.EMPTY_BYTE_ARRAY);
584     if (!lastKeyInRange) {
585       List<LoadQueueItem> lqis = splitStoreFile(item, table,
586           startEndKeys.getFirst()[indexForCallable],
587           startEndKeys.getSecond()[indexForCallable]);
588       return lqis;
589     }
590 
591     // group regions.
592     regionGroups.put(ByteBuffer.wrap(startEndKeys.getFirst()[idx]), item);
593     return null;
594   }
595 
596   /**
597    * @deprecated Use {@link #tryAtomicRegionLoad(HConnection, TableName, byte[], Collection)}
598    */
599   @Deprecated
600   protected List<LoadQueueItem> tryAtomicRegionLoad(final HConnection conn,
601       final byte [] tableName, final byte[] first, Collection<LoadQueueItem> lqis)
602   throws IOException {
603     return tryAtomicRegionLoad(conn, TableName.valueOf(tableName), first, lqis);
604   }
605 
606   /**
607    * Attempts to do an atomic load of many hfiles into a region.  If it fails,
608    * it returns a list of hfiles that need to be retried.  If it is successful
609    * it will return an empty list.
610    *
611    * NOTE: To maintain row atomicity guarantees, region server callable should
612    * succeed atomically and fails atomically.
613    *
614    * Protected for testing.
615    *
616    * @return empty list if success, list of items to retry on recoverable
617    * failure
618    */
619   protected List<LoadQueueItem> tryAtomicRegionLoad(final HConnection conn,
620       final TableName tableName, final byte[] first, Collection<LoadQueueItem> lqis)
621   throws IOException {
622     final List<Pair<byte[], String>> famPaths =
623       new ArrayList<Pair<byte[], String>>(lqis.size());
624     for (LoadQueueItem lqi : lqis) {
625       famPaths.add(Pair.newPair(lqi.family, lqi.hfilePath.toString()));
626     }
627 
628     final RegionServerCallable<Boolean> svrCallable =
629         new RegionServerCallable<Boolean>(conn, tableName, first) {
630       @Override
631       public Boolean call() throws Exception {
632         SecureBulkLoadClient secureClient = null;
633         boolean success = false;
634 
635         try {
636           LOG.debug("Going to connect to server " + getLocation() + " for row "
637               + Bytes.toStringBinary(getRow()) + " with hfile group " + famPaths);
638           byte[] regionName = getLocation().getRegionInfo().getRegionName();
639           if (!isSecureBulkLoadEndpointAvailable()) {
640             success = ProtobufUtil.bulkLoadHFile(getStub(), famPaths, regionName, assignSeqIds);
641           } else {
642             HTable table = new HTable(conn.getConfiguration(), getTableName());
643             secureClient = new SecureBulkLoadClient(table);
644             success = secureClient.bulkLoadHFiles(famPaths, fsDelegationToken.getUserToken(),
645               bulkToken, getLocation().getRegionInfo().getStartKey());
646           }
647           return success;
648         } finally {
649           //Best effort copying of files that might not have been imported
650           //from the staging directory back to original location
651           //in user directory
652           if(secureClient != null && !success) {
653             FileSystem targetFs = FileSystem.get(getConf());
654             // Check to see if the source and target filesystems are the same
655             // If they are the same filesystem, we will try move the files back
656             // because previously we moved them to the staging directory.
657             if (FSHDFSUtils.isSameHdfs(getConf(), fs, targetFs)) {
658               for(Pair<byte[], String> el : famPaths) {
659                 Path hfileStagingPath = null;
660                 Path hfileOrigPath = new Path(el.getSecond());
661                 try {
662                   hfileStagingPath= new Path(secureClient.getStagingPath(bulkToken, el.getFirst()),
663                     hfileOrigPath.getName());
664                   if(targetFs.rename(hfileStagingPath, hfileOrigPath)) {
665                     LOG.debug("Moved back file " + hfileOrigPath + " from " +
666                         hfileStagingPath);
667                   } else if(targetFs.exists(hfileStagingPath)){
668                     LOG.debug("Unable to move back file " + hfileOrigPath + " from " +
669                         hfileStagingPath);
670                   }
671                 } catch(Exception ex) {
672                   LOG.debug("Unable to move back file " + hfileOrigPath + " from " +
673                       hfileStagingPath, ex);
674                 }
675               }
676             }
677           }
678         }
679       }
680     };
681 
682     try {
683       List<LoadQueueItem> toRetry = new ArrayList<LoadQueueItem>();
684       Configuration conf = getConf();
685       boolean success = RpcRetryingCallerFactory.instantiate(conf).<Boolean> newCaller()
686           .callWithRetries(svrCallable);
687       if (!success) {
688         LOG.warn("Attempt to bulk load region containing "
689             + Bytes.toStringBinary(first) + " into table "
690             + tableName  + " with files " + lqis
691             + " failed.  This is recoverable and they will be retried.");
692         toRetry.addAll(lqis); // return lqi's to retry
693       }
694       // success
695       return toRetry;
696     } catch (IOException e) {
697       LOG.error("Encountered unrecoverable error from region server, additional details: "
698           + svrCallable.getExceptionMessageAdditionalDetail(), e);
699       throw e;
700     }
701   }
702   
703   private boolean isSecureBulkLoadEndpointAvailable() {
704     String classes = getConf().get(CoprocessorHost.REGION_COPROCESSOR_CONF_KEY, "");
705     return classes.contains("org.apache.hadoop.hbase.security.access.SecureBulkLoadEndpoint");
706   }
707 
708   /**
709    * Split a storefile into a top and bottom half, maintaining
710    * the metadata, recreating bloom filters, etc.
711    */
712   static void splitStoreFile(
713       Configuration conf, Path inFile,
714       HColumnDescriptor familyDesc, byte[] splitKey,
715       Path bottomOut, Path topOut) throws IOException
716   {
717     // Open reader with no block cache, and not in-memory
718     Reference topReference = Reference.createTopReference(splitKey);
719     Reference bottomReference = Reference.createBottomReference(splitKey);
720 
721     copyHFileHalf(conf, inFile, topOut, topReference, familyDesc);
722     copyHFileHalf(conf, inFile, bottomOut, bottomReference, familyDesc);
723   }
724 
725   /**
726    * Copy half of an HFile into a new HFile.
727    */
728   private static void copyHFileHalf(
729       Configuration conf, Path inFile, Path outFile, Reference reference,
730       HColumnDescriptor familyDescriptor)
731   throws IOException {
732     FileSystem fs = inFile.getFileSystem(conf);
733     CacheConfig cacheConf = new CacheConfig(conf);
734     HalfStoreFileReader halfReader = null;
735     StoreFile.Writer halfWriter = null;
736     try {
737       halfReader = new HalfStoreFileReader(fs, inFile, cacheConf, reference, conf);
738       Map<byte[], byte[]> fileInfo = halfReader.loadFileInfo();
739 
740       int blocksize = familyDescriptor.getBlocksize();
741       Algorithm compression = familyDescriptor.getCompression();
742       BloomType bloomFilterType = familyDescriptor.getBloomFilterType();
743       HFileContext hFileContext = new HFileContextBuilder()
744                                   .withCompression(compression)
745                                   .withChecksumType(HStore.getChecksumType(conf))
746                                   .withBytesPerCheckSum(HStore.getBytesPerChecksum(conf))
747                                   .withBlockSize(blocksize)
748                                   .withDataBlockEncoding(familyDescriptor.getDataBlockEncoding())
749                                   .build();
750       halfWriter = new StoreFile.WriterBuilder(conf, cacheConf,
751           fs)
752               .withFilePath(outFile)
753               .withBloomType(bloomFilterType)
754               .withFileContext(hFileContext)
755               .build();
756       HFileScanner scanner = halfReader.getScanner(false, false, false);
757       scanner.seekTo();
758       do {
759         KeyValue kv = scanner.getKeyValue();
760         halfWriter.append(kv);
761       } while (scanner.next());
762 
763       for (Map.Entry<byte[],byte[]> entry : fileInfo.entrySet()) {
764         if (shouldCopyHFileMetaKey(entry.getKey())) {
765           halfWriter.appendFileInfo(entry.getKey(), entry.getValue());
766         }
767       }
768     } finally {
769       if (halfWriter != null) halfWriter.close();
770       if (halfReader != null) halfReader.close(cacheConf.shouldEvictOnClose());
771     }
772   }
773 
774   private static boolean shouldCopyHFileMetaKey(byte[] key) {
775     return !HFile.isReservedFileInfoKey(key);
776   }
777 
778   private boolean doesTableExist(TableName tableName) throws Exception {
779     return hbAdmin.tableExists(tableName);
780   }
781 
782   /*
783    * Infers region boundaries for a new table.
784    * Parameter:
785    *   bdryMap is a map between keys to an integer belonging to {+1, -1}
786    *     If a key is a start key of a file, then it maps to +1
787    *     If a key is an end key of a file, then it maps to -1
788    * Algo:
789    * 1) Poll on the keys in order:
790    *    a) Keep adding the mapped values to these keys (runningSum)
791    *    b) Each time runningSum reaches 0, add the start Key from when the runningSum had started to a boundary list.
792    * 2) Return the boundary list.
793    */
794   public static byte[][] inferBoundaries(TreeMap<byte[], Integer> bdryMap) {
795     ArrayList<byte[]> keysArray = new ArrayList<byte[]>();
796     int runningValue = 0;
797     byte[] currStartKey = null;
798     boolean firstBoundary = true;
799 
800     for (Map.Entry<byte[], Integer> item: bdryMap.entrySet()) {
801       if (runningValue == 0) currStartKey = item.getKey();
802       runningValue += item.getValue();
803       if (runningValue == 0) {
804         if (!firstBoundary) keysArray.add(currStartKey);
805         firstBoundary = false;
806       }
807     }
808 
809     return keysArray.toArray(new byte[0][0]);
810   }
811 
812   /*
813    * If the table is created for the first time, then "completebulkload" reads the files twice.
814    * More modifications necessary if we want to avoid doing it.
815    */
816   private void createTable(TableName tableName, String dirPath) throws Exception {
817     Path hfofDir = new Path(dirPath);
818     FileSystem fs = hfofDir.getFileSystem(getConf());
819 
820     if (!fs.exists(hfofDir)) {
821       throw new FileNotFoundException("HFileOutputFormat dir " +
822           hfofDir + " not found");
823     }
824 
825     FileStatus[] familyDirStatuses = fs.listStatus(hfofDir);
826     if (familyDirStatuses == null) {
827       throw new FileNotFoundException("No families found in " + hfofDir);
828     }
829 
830     HTableDescriptor htd = new HTableDescriptor(tableName);
831     HColumnDescriptor hcd;
832 
833     // Add column families
834     // Build a set of keys
835     byte[][] keys;
836     TreeMap<byte[], Integer> map = new TreeMap<byte[], Integer>(Bytes.BYTES_COMPARATOR);
837 
838     for (FileStatus stat : familyDirStatuses) {
839       if (!stat.isDir()) {
840         LOG.warn("Skipping non-directory " + stat.getPath());
841         continue;
842       }
843       Path familyDir = stat.getPath();
844       if (familyDir.getName().equals("_logs")) {
845         // Family name of "_logs" is not supported.
846         // This is due to the presence of history directory under _logs directory
847         // when hadoop-1 is used
848         continue;
849       }
850       byte[] family = familyDir.getName().getBytes();
851 
852       hcd = new HColumnDescriptor(family);
853       htd.addFamily(hcd);
854 
855       Path[] hfiles = FileUtil.stat2Paths(fs.listStatus(familyDir));
856       for (Path hfile : hfiles) {
857         if (hfile.getName().startsWith("_")) continue;
858         HFile.Reader reader = HFile.createReader(fs, hfile,
859             new CacheConfig(getConf()), getConf());
860         final byte[] first, last;
861         try {
862           if (hcd.getCompressionType() != reader.getFileContext().getCompression()) {
863             hcd.setCompressionType(reader.getFileContext().getCompression());
864             LOG.info("Setting compression " + hcd.getCompressionType().name() +
865                      " for family " + hcd.toString());
866           }
867           reader.loadFileInfo();
868           first = reader.getFirstRowKey();
869           last =  reader.getLastRowKey();
870 
871           LOG.info("Trying to figure out region boundaries hfile=" + hfile +
872             " first=" + Bytes.toStringBinary(first) +
873             " last="  + Bytes.toStringBinary(last));
874 
875           // To eventually infer start key-end key boundaries
876           Integer value = map.containsKey(first)? map.get(first):0;
877           map.put(first, value+1);
878 
879           value = map.containsKey(last)? map.get(last):0;
880           map.put(last, value-1);
881         }  finally {
882           reader.close();
883         }
884       }
885     }
886 
887     keys = LoadIncrementalHFiles.inferBoundaries(map);
888     this.hbAdmin.createTable(htd,keys);
889 
890     LOG.info("Table "+ tableName +" is available!!");
891   }
892 
893   @Override
894   public int run(String[] args) throws Exception {
895     if (args.length != 2) {
896       usage();
897       return -1;
898     }
899 
900     initialize();
901 
902     String dirPath = args[0];
903     TableName tableName = TableName.valueOf(args[1]);
904 
905     boolean tableExists = this.doesTableExist(tableName);
906     if (!tableExists) {
907       if ("yes".equalsIgnoreCase(getConf().get(CREATE_TABLE_CONF_KEY, "yes"))) {
908         this.createTable(tableName, dirPath);
909       } else {
910         String errorMsg = format("Table '%s' does not exist.", tableName);
911         LOG.error(errorMsg);
912         throw new TableNotFoundException(errorMsg);
913       }
914     }
915 
916     Path hfofDir = new Path(dirPath);
917     HTable table = new HTable(getConf(), tableName);
918 
919     doBulkLoad(hfofDir, table);
920     return 0;
921   }
922 
923   public static void main(String[] args) throws Exception {
924     Configuration conf = HBaseConfiguration.create();
925     int ret = ToolRunner.run(conf, new LoadIncrementalHFiles(), args);
926     System.exit(ret);
927   }
928 
929 }