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 org.junit.Assert.assertEquals;
22  import static org.junit.Assert.assertFalse;
23  import static org.junit.Assert.assertNotNull;
24  import static org.junit.Assert.assertNotSame;
25  import static org.junit.Assert.assertTrue;
26  import static org.junit.Assert.fail;
27  
28  import java.io.IOException;
29  import java.io.UnsupportedEncodingException;
30  import java.util.Arrays;
31  import java.util.HashMap;
32  import java.util.Map;
33  import java.util.Map.Entry;
34  import java.util.Random;
35  import java.util.Set;
36  import java.util.concurrent.Callable;
37  
38  import org.apache.commons.logging.Log;
39  import org.apache.commons.logging.LogFactory;
40  import org.apache.hadoop.conf.Configuration;
41  import org.apache.hadoop.fs.FileStatus;
42  import org.apache.hadoop.fs.FileSystem;
43  import org.apache.hadoop.fs.Path;
44  import org.apache.hadoop.hbase.CategoryBasedTimeout;
45  import org.apache.hadoop.hbase.Cell;
46  import org.apache.hadoop.hbase.CellUtil;
47  import org.apache.hadoop.hbase.CompatibilitySingletonFactory;
48  import org.apache.hadoop.hbase.HBaseConfiguration;
49  import org.apache.hadoop.hbase.HBaseTestingUtility;
50  import org.apache.hadoop.hbase.HColumnDescriptor;
51  import org.apache.hadoop.hbase.HConstants;
52  import org.apache.hadoop.hbase.HDFSBlocksDistribution;
53  import org.apache.hadoop.hbase.HTableDescriptor;
54  import org.apache.hadoop.hbase.HadoopShims;
55  import org.apache.hadoop.hbase.KeyValue;
56  import org.apache.hadoop.hbase.PerformanceEvaluation;
57  import org.apache.hadoop.hbase.TableName;
58  import org.apache.hadoop.hbase.client.Admin;
59  import org.apache.hadoop.hbase.client.Connection;
60  import org.apache.hadoop.hbase.client.ConnectionFactory;
61  import org.apache.hadoop.hbase.client.HTable;
62  import org.apache.hadoop.hbase.client.Put;
63  import org.apache.hadoop.hbase.client.RegionLocator;
64  import org.apache.hadoop.hbase.client.Result;
65  import org.apache.hadoop.hbase.client.ResultScanner;
66  import org.apache.hadoop.hbase.client.Scan;
67  import org.apache.hadoop.hbase.client.Table;
68  import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
69  import org.apache.hadoop.hbase.io.compress.Compression;
70  import org.apache.hadoop.hbase.io.compress.Compression.Algorithm;
71  import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding;
72  import org.apache.hadoop.hbase.io.hfile.CacheConfig;
73  import org.apache.hadoop.hbase.io.hfile.HFile;
74  import org.apache.hadoop.hbase.io.hfile.HFile.Reader;
75  import org.apache.hadoop.hbase.regionserver.BloomType;
76  import org.apache.hadoop.hbase.regionserver.HRegion;
77  import org.apache.hadoop.hbase.regionserver.StoreFile;
78  import org.apache.hadoop.hbase.regionserver.TimeRangeTracker;
79  import org.apache.hadoop.hbase.testclassification.LargeTests;
80  import org.apache.hadoop.hbase.testclassification.VerySlowMapReduceTests;
81  import org.apache.hadoop.hbase.util.Bytes;
82  import org.apache.hadoop.hbase.util.FSUtils;
83  import org.apache.hadoop.hbase.util.Threads;
84  import org.apache.hadoop.hbase.util.Writables;
85  import org.apache.hadoop.io.NullWritable;
86  import org.apache.hadoop.mapreduce.Job;
87  import org.apache.hadoop.mapreduce.Mapper;
88  import org.apache.hadoop.mapreduce.RecordWriter;
89  import org.apache.hadoop.mapreduce.TaskAttemptContext;
90  import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
91  import org.junit.Ignore;
92  import org.junit.Rule;
93  import org.junit.Test;
94  import org.junit.experimental.categories.Category;
95  import org.junit.rules.TestRule;
96  import org.mockito.Mockito;
97  
98  /**
99   * Simple test for {@link CellSortReducer} and {@link HFileOutputFormat2}.
100  * Sets up and runs a mapreduce job that writes hfile output.
101  * Creates a few inner classes to implement splits and an inputformat that
102  * emits keys and values like those of {@link PerformanceEvaluation}.
103  */
104 @Category({VerySlowMapReduceTests.class, LargeTests.class})
105 public class TestHFileOutputFormat2  {
106   @Rule public final TestRule timeout = CategoryBasedTimeout.builder().
107       withTimeout(this.getClass()).withLookingForStuckThread(true).build();
108   private final static int ROWSPERSPLIT = 1024;
109 
110   private static final byte[][] FAMILIES
111     = { Bytes.add(PerformanceEvaluation.FAMILY_NAME, Bytes.toBytes("-A"))
112       , Bytes.add(PerformanceEvaluation.FAMILY_NAME, Bytes.toBytes("-B"))};
113   private static final TableName TABLE_NAME =
114       TableName.valueOf("TestTable");
115 
116   private HBaseTestingUtility util = new HBaseTestingUtility();
117 
118   private static final Log LOG = LogFactory.getLog(TestHFileOutputFormat2.class);
119 
120   /**
121    * Simple mapper that makes KeyValue output.
122    */
123   static class RandomKVGeneratingMapper
124       extends Mapper<NullWritable, NullWritable,
125                  ImmutableBytesWritable, Cell> {
126 
127     private int keyLength;
128     private static final int KEYLEN_DEFAULT=10;
129     private static final String KEYLEN_CONF="randomkv.key.length";
130 
131     private int valLength;
132     private static final int VALLEN_DEFAULT=10;
133     private static final String VALLEN_CONF="randomkv.val.length";
134     private static final byte [] QUALIFIER = Bytes.toBytes("data");
135 
136     @Override
137     protected void setup(Context context) throws IOException,
138         InterruptedException {
139       super.setup(context);
140 
141       Configuration conf = context.getConfiguration();
142       keyLength = conf.getInt(KEYLEN_CONF, KEYLEN_DEFAULT);
143       valLength = conf.getInt(VALLEN_CONF, VALLEN_DEFAULT);
144     }
145 
146     @Override
147     protected void map(
148         NullWritable n1, NullWritable n2,
149         Mapper<NullWritable, NullWritable,
150                ImmutableBytesWritable,Cell>.Context context)
151         throws java.io.IOException ,InterruptedException
152     {
153 
154       byte keyBytes[] = new byte[keyLength];
155       byte valBytes[] = new byte[valLength];
156 
157       int taskId = context.getTaskAttemptID().getTaskID().getId();
158       assert taskId < Byte.MAX_VALUE : "Unit tests dont support > 127 tasks!";
159 
160       Random random = new Random();
161       for (int i = 0; i < ROWSPERSPLIT; i++) {
162 
163         random.nextBytes(keyBytes);
164         // Ensure that unique tasks generate unique keys
165         keyBytes[keyLength - 1] = (byte)(taskId & 0xFF);
166         random.nextBytes(valBytes);
167         ImmutableBytesWritable key = new ImmutableBytesWritable(keyBytes);
168 
169         for (byte[] family : TestHFileOutputFormat2.FAMILIES) {
170           Cell kv = new KeyValue(keyBytes, family, QUALIFIER, valBytes);
171           context.write(key, kv);
172         }
173       }
174     }
175   }
176 
177   private void setupRandomGeneratorMapper(Job job) {
178     job.setInputFormatClass(NMapInputFormat.class);
179     job.setMapperClass(RandomKVGeneratingMapper.class);
180     job.setMapOutputKeyClass(ImmutableBytesWritable.class);
181     job.setMapOutputValueClass(KeyValue.class);
182   }
183 
184   /**
185    * Test that {@link HFileOutputFormat2} RecordWriter amends timestamps if
186    * passed a keyvalue whose timestamp is {@link HConstants#LATEST_TIMESTAMP}.
187    * @see <a href="https://issues.apache.org/jira/browse/HBASE-2615">HBASE-2615</a>
188    */
189   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
190   public void test_LATEST_TIMESTAMP_isReplaced()
191   throws Exception {
192     Configuration conf = new Configuration(this.util.getConfiguration());
193     RecordWriter<ImmutableBytesWritable, Cell> writer = null;
194     TaskAttemptContext context = null;
195     Path dir =
196       util.getDataTestDir("test_LATEST_TIMESTAMP_isReplaced");
197     try {
198       Job job = new Job(conf);
199       FileOutputFormat.setOutputPath(job, dir);
200       context = createTestTaskAttemptContext(job);
201       HFileOutputFormat2 hof = new HFileOutputFormat2();
202       writer = hof.getRecordWriter(context);
203       final byte [] b = Bytes.toBytes("b");
204 
205       // Test 1.  Pass a KV that has a ts of LATEST_TIMESTAMP.  It should be
206       // changed by call to write.  Check all in kv is same but ts.
207       KeyValue kv = new KeyValue(b, b, b);
208       KeyValue original = kv.clone();
209       writer.write(new ImmutableBytesWritable(), kv);
210       assertFalse(original.equals(kv));
211       assertTrue(Bytes.equals(CellUtil.cloneRow(original), CellUtil.cloneRow(kv)));
212       assertTrue(Bytes.equals(CellUtil.cloneFamily(original), CellUtil.cloneFamily(kv)));
213       assertTrue(Bytes.equals(CellUtil.cloneQualifier(original), CellUtil.cloneQualifier(kv)));
214       assertNotSame(original.getTimestamp(), kv.getTimestamp());
215       assertNotSame(HConstants.LATEST_TIMESTAMP, kv.getTimestamp());
216 
217       // Test 2. Now test passing a kv that has explicit ts.  It should not be
218       // changed by call to record write.
219       kv = new KeyValue(b, b, b, kv.getTimestamp() - 1, b);
220       original = kv.clone();
221       writer.write(new ImmutableBytesWritable(), kv);
222       assertTrue(original.equals(kv));
223     } finally {
224       if (writer != null && context != null) writer.close(context);
225       dir.getFileSystem(conf).delete(dir, true);
226     }
227   }
228 
229   private TaskAttemptContext createTestTaskAttemptContext(final Job job)
230   throws Exception {
231     HadoopShims hadoop = CompatibilitySingletonFactory.getInstance(HadoopShims.class);
232     TaskAttemptContext context = hadoop.createTestTaskAttemptContext(
233       job, "attempt_201402131733_0001_m_000000_0");
234     return context;
235   }
236 
237   /*
238    * Test that {@link HFileOutputFormat2} creates an HFile with TIMERANGE
239    * metadata used by time-restricted scans.
240    */
241   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
242   public void test_TIMERANGE() throws Exception {
243     Configuration conf = new Configuration(this.util.getConfiguration());
244     RecordWriter<ImmutableBytesWritable, Cell> writer = null;
245     TaskAttemptContext context = null;
246     Path dir =
247       util.getDataTestDir("test_TIMERANGE_present");
248     LOG.info("Timerange dir writing to dir: "+ dir);
249     try {
250       // build a record writer using HFileOutputFormat2
251       Job job = new Job(conf);
252       FileOutputFormat.setOutputPath(job, dir);
253       context = createTestTaskAttemptContext(job);
254       HFileOutputFormat2 hof = new HFileOutputFormat2();
255       writer = hof.getRecordWriter(context);
256 
257       // Pass two key values with explicit times stamps
258       final byte [] b = Bytes.toBytes("b");
259 
260       // value 1 with timestamp 2000
261       KeyValue kv = new KeyValue(b, b, b, 2000, b);
262       KeyValue original = kv.clone();
263       writer.write(new ImmutableBytesWritable(), kv);
264       assertEquals(original,kv);
265 
266       // value 2 with timestamp 1000
267       kv = new KeyValue(b, b, b, 1000, b);
268       original = kv.clone();
269       writer.write(new ImmutableBytesWritable(), kv);
270       assertEquals(original, kv);
271 
272       // verify that the file has the proper FileInfo.
273       writer.close(context);
274 
275       // the generated file lives 1 directory down from the attempt directory
276       // and is the only file, e.g.
277       // _attempt__0000_r_000000_0/b/1979617994050536795
278       FileSystem fs = FileSystem.get(conf);
279       Path attemptDirectory = hof.getDefaultWorkFile(context, "").getParent();
280       FileStatus[] sub1 = fs.listStatus(attemptDirectory);
281       FileStatus[] file = fs.listStatus(sub1[0].getPath());
282 
283       // open as HFile Reader and pull out TIMERANGE FileInfo.
284       HFile.Reader rd = HFile.createReader(fs, file[0].getPath(),
285           new CacheConfig(conf), conf);
286       Map<byte[],byte[]> finfo = rd.loadFileInfo();
287       byte[] range = finfo.get("TIMERANGE".getBytes());
288       assertNotNull(range);
289 
290       // unmarshall and check values.
291       TimeRangeTracker timeRangeTracker = new TimeRangeTracker();
292       Writables.copyWritable(range, timeRangeTracker);
293       LOG.info(timeRangeTracker.getMinimumTimestamp() +
294           "...." + timeRangeTracker.getMaximumTimestamp());
295       assertEquals(1000, timeRangeTracker.getMinimumTimestamp());
296       assertEquals(2000, timeRangeTracker.getMaximumTimestamp());
297       rd.close();
298     } finally {
299       if (writer != null && context != null) writer.close(context);
300       dir.getFileSystem(conf).delete(dir, true);
301     }
302   }
303 
304   /**
305    * Run small MR job.
306    */
307   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
308   public void testWritingPEData() throws Exception {
309     Configuration conf = util.getConfiguration();
310     Path testDir = util.getDataTestDirOnTestFS("testWritingPEData");
311     FileSystem fs = testDir.getFileSystem(conf);
312 
313     // Set down this value or we OOME in eclipse.
314     conf.setInt("mapreduce.task.io.sort.mb", 20);
315     // Write a few files.
316     conf.setLong(HConstants.HREGION_MAX_FILESIZE, 64 * 1024);
317 
318     Job job = new Job(conf, "testWritingPEData");
319     setupRandomGeneratorMapper(job);
320     // This partitioner doesn't work well for number keys but using it anyways
321     // just to demonstrate how to configure it.
322     byte[] startKey = new byte[RandomKVGeneratingMapper.KEYLEN_DEFAULT];
323     byte[] endKey = new byte[RandomKVGeneratingMapper.KEYLEN_DEFAULT];
324 
325     Arrays.fill(startKey, (byte)0);
326     Arrays.fill(endKey, (byte)0xff);
327 
328     job.setPartitionerClass(SimpleTotalOrderPartitioner.class);
329     // Set start and end rows for partitioner.
330     SimpleTotalOrderPartitioner.setStartKey(job.getConfiguration(), startKey);
331     SimpleTotalOrderPartitioner.setEndKey(job.getConfiguration(), endKey);
332     job.setReducerClass(KeyValueSortReducer.class);
333     job.setOutputFormatClass(HFileOutputFormat2.class);
334     job.setNumReduceTasks(4);
335     job.getConfiguration().setStrings("io.serializations", conf.get("io.serializations"),
336         MutationSerialization.class.getName(), ResultSerialization.class.getName(),
337         KeyValueSerialization.class.getName());
338 
339     FileOutputFormat.setOutputPath(job, testDir);
340     assertTrue(job.waitForCompletion(false));
341     FileStatus [] files = fs.listStatus(testDir);
342     assertTrue(files.length > 0);
343   }
344 
345   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
346   public void testJobConfiguration() throws Exception {
347     Configuration conf = new Configuration(this.util.getConfiguration());
348     conf.set("hbase.fs.tmp.dir", util.getDataTestDir("testJobConfiguration").toString());
349     Job job = new Job(conf);
350     job.setWorkingDirectory(util.getDataTestDir("testJobConfiguration"));
351     Table table = Mockito.mock(Table.class);
352     RegionLocator regionLocator = Mockito.mock(RegionLocator.class);
353     setupMockStartKeys(regionLocator);
354     setupMockTableName(regionLocator);
355     HFileOutputFormat2.configureIncrementalLoad(job, table.getTableDescriptor(), regionLocator);
356     assertEquals(job.getNumReduceTasks(), 4);
357   }
358 
359   private byte [][] generateRandomStartKeys(int numKeys) {
360     Random random = new Random();
361     byte[][] ret = new byte[numKeys][];
362     // first region start key is always empty
363     ret[0] = HConstants.EMPTY_BYTE_ARRAY;
364     for (int i = 1; i < numKeys; i++) {
365       ret[i] =
366         PerformanceEvaluation.generateData(random, PerformanceEvaluation.DEFAULT_VALUE_LENGTH);
367     }
368     return ret;
369   }
370 
371   private byte[][] generateRandomSplitKeys(int numKeys) {
372     Random random = new Random();
373     byte[][] ret = new byte[numKeys][];
374     for (int i = 0; i < numKeys; i++) {
375       ret[i] =
376           PerformanceEvaluation.generateData(random, PerformanceEvaluation.DEFAULT_VALUE_LENGTH);
377     }
378     return ret;
379   }
380 
381   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
382   public void testMRIncrementalLoad() throws Exception {
383     LOG.info("\nStarting test testMRIncrementalLoad\n");
384     doIncrementalLoadTest(false, false);
385   }
386 
387   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
388   public void testMRIncrementalLoadWithSplit() throws Exception {
389     LOG.info("\nStarting test testMRIncrementalLoadWithSplit\n");
390     doIncrementalLoadTest(true, false);
391   }
392 
393   /**
394    * Test for HFileOutputFormat2.LOCALITY_SENSITIVE_CONF_KEY = true
395    * This test could only check the correctness of original logic if LOCALITY_SENSITIVE_CONF_KEY
396    * is set to true. Because MiniHBaseCluster always run with single hostname (and different ports),
397    * it's not possible to check the region locality by comparing region locations and DN hostnames.
398    * When MiniHBaseCluster supports explicit hostnames parameter (just like MiniDFSCluster does),
399    * we could test region locality features more easily.
400    */
401   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
402   public void testMRIncrementalLoadWithLocality() throws Exception {
403     LOG.info("\nStarting test testMRIncrementalLoadWithLocality\n");
404     doIncrementalLoadTest(false, true);
405     doIncrementalLoadTest(true, true);
406   }
407 
408   private void doIncrementalLoadTest(boolean shouldChangeRegions, boolean shouldKeepLocality)
409       throws Exception {
410     util = new HBaseTestingUtility();
411     Configuration conf = util.getConfiguration();
412     conf.setBoolean(HFileOutputFormat2.LOCALITY_SENSITIVE_CONF_KEY, shouldKeepLocality);
413     int hostCount = 1;
414     int regionNum = 5;
415     if(shouldKeepLocality) {
416       // We should change host count higher than hdfs replica count when MiniHBaseCluster supports
417       // explicit hostnames parameter just like MiniDFSCluster does.
418       hostCount = 3;
419       regionNum = 20;
420     }
421 
422     byte[][] splitKeys = generateRandomSplitKeys(regionNum - 1);
423     String[] hostnames = new String[hostCount];
424     for(int i = 0; i < hostCount; ++i) {
425       hostnames[i] = "datanode_" + i;
426     }
427     util.startMiniCluster(1, hostCount, hostnames);
428 
429     Table table = util.createTable(TABLE_NAME, FAMILIES, splitKeys);
430     Path testDir = util.getDataTestDirOnTestFS("testLocalMRIncrementalLoad");
431     try (RegionLocator r = util.getConnection().getRegionLocator(TABLE_NAME); Admin admin =
432         util.getConnection().getAdmin();) {
433       assertEquals("Should start with empty table", 0, util.countRows(table));
434       int numRegions = r.getStartKeys().length;
435       assertEquals("Should make " + regionNum + " regions", numRegions, regionNum);
436 
437       // Generate the bulk load files
438       runIncrementalPELoad(conf, table.getTableDescriptor(), r, testDir);
439       // This doesn't write into the table, just makes files
440       assertEquals("HFOF should not touch actual table", 0, util.countRows(table));
441 
442       // Make sure that a directory was created for every CF
443       int dir = 0;
444       for (FileStatus f : testDir.getFileSystem(conf).listStatus(testDir)) {
445         for (byte[] family : FAMILIES) {
446           if (Bytes.toString(family).equals(f.getPath().getName())) {
447             ++dir;
448           }
449         }
450       }
451       assertEquals("Column family not found in FS.", FAMILIES.length, dir);
452 
453       // handle the split case
454       if (shouldChangeRegions) {
455         LOG.info("Changing regions in table");
456         admin.disableTable(table.getName());
457         while(util.getMiniHBaseCluster().getMaster().getAssignmentManager().
458             getRegionStates().isRegionsInTransition()) {
459           Threads.sleep(200);
460           LOG.info("Waiting on table to finish disabling");
461         }
462         util.deleteTable(table.getName());
463         byte[][] newSplitKeys = generateRandomSplitKeys(14);
464         table = util.createTable(TABLE_NAME, FAMILIES, newSplitKeys);
465 
466         while (util.getConnection().getRegionLocator(TABLE_NAME)
467             .getAllRegionLocations().size() != 15 ||
468             !admin.isTableAvailable(table.getName())) {
469           Thread.sleep(200);
470           LOG.info("Waiting for new region assignment to happen");
471         }
472       }
473 
474       // Perform the actual load
475       new LoadIncrementalHFiles(conf).doBulkLoad(testDir, admin, table, r);
476 
477       // Ensure data shows up
478       int expectedRows = NMapInputFormat.getNumMapTasks(conf) * ROWSPERSPLIT;
479       assertEquals("LoadIncrementalHFiles should put expected data in table",
480           expectedRows, util.countRows(table));
481       Scan scan = new Scan();
482       ResultScanner results = table.getScanner(scan);
483       for (Result res : results) {
484         assertEquals(FAMILIES.length, res.rawCells().length);
485         Cell first = res.rawCells()[0];
486         for (Cell kv : res.rawCells()) {
487           assertTrue(CellUtil.matchingRow(first, kv));
488           assertTrue(Bytes.equals(CellUtil.cloneValue(first), CellUtil.cloneValue(kv)));
489         }
490       }
491       results.close();
492       String tableDigestBefore = util.checksumRows(table);
493 
494       // Check region locality
495       HDFSBlocksDistribution hbd = new HDFSBlocksDistribution();
496       for (HRegion region : util.getHBaseCluster().getRegions(TABLE_NAME)) {
497         hbd.add(region.getHDFSBlocksDistribution());
498       }
499       for (String hostname : hostnames) {
500         float locality =  hbd.getBlockLocalityIndex(hostname);
501         LOG.info("locality of [" + hostname + "]: " + locality);
502         assertEquals(100, (int) (locality * 100));
503       }
504 
505       // Cause regions to reopen
506       admin.disableTable(TABLE_NAME);
507       while (!admin.isTableDisabled(TABLE_NAME)) {
508         Thread.sleep(200);
509         LOG.info("Waiting for table to disable");
510       }
511       admin.enableTable(TABLE_NAME);
512       util.waitTableAvailable(TABLE_NAME);
513       assertEquals("Data should remain after reopening of regions",
514           tableDigestBefore, util.checksumRows(table));
515     } finally {
516       testDir.getFileSystem(conf).delete(testDir, true);
517       util.deleteTable(TABLE_NAME);
518       util.shutdownMiniCluster();
519     }
520   }
521 
522   private void runIncrementalPELoad(Configuration conf, HTableDescriptor tableDescriptor,
523       RegionLocator regionLocator, Path outDir) throws IOException, UnsupportedEncodingException,
524       InterruptedException, ClassNotFoundException {
525     Job job = new Job(conf, "testLocalMRIncrementalLoad");
526     job.setWorkingDirectory(util.getDataTestDirOnTestFS("runIncrementalPELoad"));
527     job.getConfiguration().setStrings("io.serializations", conf.get("io.serializations"),
528         MutationSerialization.class.getName(), ResultSerialization.class.getName(),
529         KeyValueSerialization.class.getName());
530     setupRandomGeneratorMapper(job);
531     HFileOutputFormat2.configureIncrementalLoad(job, tableDescriptor, regionLocator);
532     FileOutputFormat.setOutputPath(job, outDir);
533 
534     assertFalse(util.getTestFileSystem().exists(outDir)) ;
535 
536     assertEquals(regionLocator.getAllRegionLocations().size(), job.getNumReduceTasks());
537 
538     assertTrue(job.waitForCompletion(true));
539   }
540 
541   /**
542    * Test for {@link HFileOutputFormat2#configureCompression(org.apache.hadoop.hbase.client.Table,
543    * Configuration)} and {@link HFileOutputFormat2#createFamilyCompressionMap
544    * (Configuration)}.
545    * Tests that the compression map is correctly serialized into
546    * and deserialized from configuration
547    *
548    * @throws IOException
549    */
550   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
551   public void testSerializeDeserializeFamilyCompressionMap() throws IOException {
552     for (int numCfs = 0; numCfs <= 3; numCfs++) {
553       Configuration conf = new Configuration(this.util.getConfiguration());
554       Map<String, Compression.Algorithm> familyToCompression =
555           getMockColumnFamiliesForCompression(numCfs);
556       Table table = Mockito.mock(HTable.class);
557       setupMockColumnFamiliesForCompression(table, familyToCompression);
558       HFileOutputFormat2.configureCompression(conf, table.getTableDescriptor());
559 
560       // read back family specific compression setting from the configuration
561       Map<byte[], Algorithm> retrievedFamilyToCompressionMap = HFileOutputFormat2
562           .createFamilyCompressionMap(conf);
563 
564       // test that we have a value for all column families that matches with the
565       // used mock values
566       for (Entry<String, Algorithm> entry : familyToCompression.entrySet()) {
567         assertEquals("Compression configuration incorrect for column family:"
568             + entry.getKey(), entry.getValue(),
569             retrievedFamilyToCompressionMap.get(entry.getKey().getBytes()));
570       }
571     }
572   }
573 
574   private void setupMockColumnFamiliesForCompression(Table table,
575       Map<String, Compression.Algorithm> familyToCompression) throws IOException {
576     HTableDescriptor mockTableDescriptor = new HTableDescriptor(TABLE_NAME);
577     for (Entry<String, Compression.Algorithm> entry : familyToCompression.entrySet()) {
578       mockTableDescriptor.addFamily(new HColumnDescriptor(entry.getKey())
579           .setMaxVersions(1)
580           .setCompressionType(entry.getValue())
581           .setBlockCacheEnabled(false)
582           .setTimeToLive(0));
583     }
584     Mockito.doReturn(mockTableDescriptor).when(table).getTableDescriptor();
585   }
586 
587   /**
588    * @return a map from column family names to compression algorithms for
589    *         testing column family compression. Column family names have special characters
590    */
591   private Map<String, Compression.Algorithm>
592       getMockColumnFamiliesForCompression (int numCfs) {
593     Map<String, Compression.Algorithm> familyToCompression
594       = new HashMap<String, Compression.Algorithm>();
595     // use column family names having special characters
596     if (numCfs-- > 0) {
597       familyToCompression.put("Family1!@#!@#&", Compression.Algorithm.LZO);
598     }
599     if (numCfs-- > 0) {
600       familyToCompression.put("Family2=asdads&!AASD", Compression.Algorithm.SNAPPY);
601     }
602     if (numCfs-- > 0) {
603       familyToCompression.put("Family2=asdads&!AASD", Compression.Algorithm.GZ);
604     }
605     if (numCfs-- > 0) {
606       familyToCompression.put("Family3", Compression.Algorithm.NONE);
607     }
608     return familyToCompression;
609   }
610 
611 
612   /**
613    * Test for {@link HFileOutputFormat2#configureBloomType(org.apache.hadoop.hbase.client.Table,
614    * Configuration)} and {@link HFileOutputFormat2#createFamilyBloomTypeMap
615    * (Configuration)}.
616    * Tests that the compression map is correctly serialized into
617    * and deserialized from configuration
618    *
619    * @throws IOException
620    */
621   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
622   public void testSerializeDeserializeFamilyBloomTypeMap() throws IOException {
623     for (int numCfs = 0; numCfs <= 2; numCfs++) {
624       Configuration conf = new Configuration(this.util.getConfiguration());
625       Map<String, BloomType> familyToBloomType =
626           getMockColumnFamiliesForBloomType(numCfs);
627       Table table = Mockito.mock(HTable.class);
628       setupMockColumnFamiliesForBloomType(table,
629           familyToBloomType);
630       HFileOutputFormat2.configureBloomType(table.getTableDescriptor(), conf);
631 
632       // read back family specific data block encoding settings from the
633       // configuration
634       Map<byte[], BloomType> retrievedFamilyToBloomTypeMap =
635           HFileOutputFormat2
636               .createFamilyBloomTypeMap(conf);
637 
638       // test that we have a value for all column families that matches with the
639       // used mock values
640       for (Entry<String, BloomType> entry : familyToBloomType.entrySet()) {
641         assertEquals("BloomType configuration incorrect for column family:"
642             + entry.getKey(), entry.getValue(),
643             retrievedFamilyToBloomTypeMap.get(entry.getKey().getBytes()));
644       }
645     }
646   }
647 
648   private void setupMockColumnFamiliesForBloomType(Table table,
649       Map<String, BloomType> familyToDataBlockEncoding) throws IOException {
650     HTableDescriptor mockTableDescriptor = new HTableDescriptor(TABLE_NAME);
651     for (Entry<String, BloomType> entry : familyToDataBlockEncoding.entrySet()) {
652       mockTableDescriptor.addFamily(new HColumnDescriptor(entry.getKey())
653           .setMaxVersions(1)
654           .setBloomFilterType(entry.getValue())
655           .setBlockCacheEnabled(false)
656           .setTimeToLive(0));
657     }
658     Mockito.doReturn(mockTableDescriptor).when(table).getTableDescriptor();
659   }
660 
661   /**
662    * @return a map from column family names to compression algorithms for
663    *         testing column family compression. Column family names have special characters
664    */
665   private Map<String, BloomType>
666   getMockColumnFamiliesForBloomType (int numCfs) {
667     Map<String, BloomType> familyToBloomType =
668         new HashMap<String, BloomType>();
669     // use column family names having special characters
670     if (numCfs-- > 0) {
671       familyToBloomType.put("Family1!@#!@#&", BloomType.ROW);
672     }
673     if (numCfs-- > 0) {
674       familyToBloomType.put("Family2=asdads&!AASD",
675           BloomType.ROWCOL);
676     }
677     if (numCfs-- > 0) {
678       familyToBloomType.put("Family3", BloomType.NONE);
679     }
680     return familyToBloomType;
681   }
682 
683   /**
684    * Test for {@link HFileOutputFormat2#configureBlockSize(org.apache.hadoop.hbase.client.Table,
685    * Configuration)} and {@link HFileOutputFormat2#createFamilyBlockSizeMap
686    * (Configuration)}.
687    * Tests that the compression map is correctly serialized into
688    * and deserialized from configuration
689    *
690    * @throws IOException
691    */
692   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
693   public void testSerializeDeserializeFamilyBlockSizeMap() throws IOException {
694     for (int numCfs = 0; numCfs <= 3; numCfs++) {
695       Configuration conf = new Configuration(this.util.getConfiguration());
696       Map<String, Integer> familyToBlockSize =
697           getMockColumnFamiliesForBlockSize(numCfs);
698       Table table = Mockito.mock(HTable.class);
699       setupMockColumnFamiliesForBlockSize(table,
700           familyToBlockSize);
701       HFileOutputFormat2.configureBlockSize(table.getTableDescriptor(), conf);
702 
703       // read back family specific data block encoding settings from the
704       // configuration
705       Map<byte[], Integer> retrievedFamilyToBlockSizeMap =
706           HFileOutputFormat2
707               .createFamilyBlockSizeMap(conf);
708 
709       // test that we have a value for all column families that matches with the
710       // used mock values
711       for (Entry<String, Integer> entry : familyToBlockSize.entrySet()
712           ) {
713         assertEquals("BlockSize configuration incorrect for column family:"
714             + entry.getKey(), entry.getValue(),
715             retrievedFamilyToBlockSizeMap.get(entry.getKey().getBytes()));
716       }
717     }
718   }
719 
720   private void setupMockColumnFamiliesForBlockSize(Table table,
721       Map<String, Integer> familyToDataBlockEncoding) throws IOException {
722     HTableDescriptor mockTableDescriptor = new HTableDescriptor(TABLE_NAME);
723     for (Entry<String, Integer> entry : familyToDataBlockEncoding.entrySet()) {
724       mockTableDescriptor.addFamily(new HColumnDescriptor(entry.getKey())
725           .setMaxVersions(1)
726           .setBlocksize(entry.getValue())
727           .setBlockCacheEnabled(false)
728           .setTimeToLive(0));
729     }
730     Mockito.doReturn(mockTableDescriptor).when(table).getTableDescriptor();
731   }
732 
733   /**
734    * @return a map from column family names to compression algorithms for
735    *         testing column family compression. Column family names have special characters
736    */
737   private Map<String, Integer>
738   getMockColumnFamiliesForBlockSize (int numCfs) {
739     Map<String, Integer> familyToBlockSize =
740         new HashMap<String, Integer>();
741     // use column family names having special characters
742     if (numCfs-- > 0) {
743       familyToBlockSize.put("Family1!@#!@#&", 1234);
744     }
745     if (numCfs-- > 0) {
746       familyToBlockSize.put("Family2=asdads&!AASD",
747           Integer.MAX_VALUE);
748     }
749     if (numCfs-- > 0) {
750       familyToBlockSize.put("Family2=asdads&!AASD",
751           Integer.MAX_VALUE);
752     }
753     if (numCfs-- > 0) {
754       familyToBlockSize.put("Family3", 0);
755     }
756     return familyToBlockSize;
757   }
758 
759   /**
760    * Test for {@link HFileOutputFormat2#configureDataBlockEncoding(HTableDescriptor, Configuration)}
761    * and {@link HFileOutputFormat2#createFamilyDataBlockEncodingMap(Configuration)}.
762    * Tests that the compression map is correctly serialized into
763    * and deserialized from configuration
764    *
765    * @throws IOException
766    */
767   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
768   public void testSerializeDeserializeFamilyDataBlockEncodingMap() throws IOException {
769     for (int numCfs = 0; numCfs <= 3; numCfs++) {
770       Configuration conf = new Configuration(this.util.getConfiguration());
771       Map<String, DataBlockEncoding> familyToDataBlockEncoding =
772           getMockColumnFamiliesForDataBlockEncoding(numCfs);
773       Table table = Mockito.mock(HTable.class);
774       setupMockColumnFamiliesForDataBlockEncoding(table,
775           familyToDataBlockEncoding);
776       HTableDescriptor tableDescriptor = table.getTableDescriptor();
777       HFileOutputFormat2.configureDataBlockEncoding(tableDescriptor, conf);
778 
779       // read back family specific data block encoding settings from the
780       // configuration
781       Map<byte[], DataBlockEncoding> retrievedFamilyToDataBlockEncodingMap =
782           HFileOutputFormat2
783           .createFamilyDataBlockEncodingMap(conf);
784 
785       // test that we have a value for all column families that matches with the
786       // used mock values
787       for (Entry<String, DataBlockEncoding> entry : familyToDataBlockEncoding.entrySet()) {
788         assertEquals("DataBlockEncoding configuration incorrect for column family:"
789             + entry.getKey(), entry.getValue(),
790             retrievedFamilyToDataBlockEncodingMap.get(entry.getKey().getBytes()));
791       }
792     }
793   }
794 
795   private void setupMockColumnFamiliesForDataBlockEncoding(Table table,
796       Map<String, DataBlockEncoding> familyToDataBlockEncoding) throws IOException {
797     HTableDescriptor mockTableDescriptor = new HTableDescriptor(TABLE_NAME);
798     for (Entry<String, DataBlockEncoding> entry : familyToDataBlockEncoding.entrySet()) {
799       mockTableDescriptor.addFamily(new HColumnDescriptor(entry.getKey())
800           .setMaxVersions(1)
801           .setDataBlockEncoding(entry.getValue())
802           .setBlockCacheEnabled(false)
803           .setTimeToLive(0));
804     }
805     Mockito.doReturn(mockTableDescriptor).when(table).getTableDescriptor();
806   }
807 
808   /**
809    * @return a map from column family names to compression algorithms for
810    *         testing column family compression. Column family names have special characters
811    */
812   private Map<String, DataBlockEncoding>
813       getMockColumnFamiliesForDataBlockEncoding (int numCfs) {
814     Map<String, DataBlockEncoding> familyToDataBlockEncoding =
815         new HashMap<String, DataBlockEncoding>();
816     // use column family names having special characters
817     if (numCfs-- > 0) {
818       familyToDataBlockEncoding.put("Family1!@#!@#&", DataBlockEncoding.DIFF);
819     }
820     if (numCfs-- > 0) {
821       familyToDataBlockEncoding.put("Family2=asdads&!AASD",
822           DataBlockEncoding.FAST_DIFF);
823     }
824     if (numCfs-- > 0) {
825       familyToDataBlockEncoding.put("Family2=asdads&!AASD",
826           DataBlockEncoding.PREFIX);
827     }
828     if (numCfs-- > 0) {
829       familyToDataBlockEncoding.put("Family3", DataBlockEncoding.NONE);
830     }
831     return familyToDataBlockEncoding;
832   }
833 
834   private void setupMockStartKeys(RegionLocator table) throws IOException {
835     byte[][] mockKeys = new byte[][] {
836         HConstants.EMPTY_BYTE_ARRAY,
837         Bytes.toBytes("aaa"),
838         Bytes.toBytes("ggg"),
839         Bytes.toBytes("zzz")
840     };
841     Mockito.doReturn(mockKeys).when(table).getStartKeys();
842   }
843 
844   private void setupMockTableName(RegionLocator table) throws IOException {
845     TableName mockTableName = TableName.valueOf("mock_table");
846     Mockito.doReturn(mockTableName).when(table).getName();
847   }
848 
849   /**
850    * Test that {@link HFileOutputFormat2} RecordWriter uses compression and
851    * bloom filter settings from the column family descriptor
852    */
853   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
854   public void testColumnFamilySettings() throws Exception {
855     Configuration conf = new Configuration(this.util.getConfiguration());
856     RecordWriter<ImmutableBytesWritable, Cell> writer = null;
857     TaskAttemptContext context = null;
858     Path dir = util.getDataTestDir("testColumnFamilySettings");
859 
860     // Setup table descriptor
861     Table table = Mockito.mock(Table.class);
862     RegionLocator regionLocator = Mockito.mock(RegionLocator.class);
863     HTableDescriptor htd = new HTableDescriptor(TABLE_NAME);
864     Mockito.doReturn(htd).when(table).getTableDescriptor();
865     for (HColumnDescriptor hcd: HBaseTestingUtility.generateColumnDescriptors()) {
866       htd.addFamily(hcd);
867     }
868 
869     // set up the table to return some mock keys
870     setupMockStartKeys(regionLocator);
871 
872     try {
873       // partial map red setup to get an operational writer for testing
874       // We turn off the sequence file compression, because DefaultCodec
875       // pollutes the GZip codec pool with an incompatible compressor.
876       conf.set("io.seqfile.compression.type", "NONE");
877       conf.set("hbase.fs.tmp.dir", dir.toString());
878       // turn locality off to eliminate getRegionLocation fail-and-retry time when writing kvs
879       conf.setBoolean(HFileOutputFormat2.LOCALITY_SENSITIVE_CONF_KEY, false);
880 
881       Job job = new Job(conf, "testLocalMRIncrementalLoad");
882       job.setWorkingDirectory(util.getDataTestDirOnTestFS("testColumnFamilySettings"));
883       setupRandomGeneratorMapper(job);
884       HFileOutputFormat2.configureIncrementalLoad(job, table.getTableDescriptor(), regionLocator);
885       FileOutputFormat.setOutputPath(job, dir);
886       context = createTestTaskAttemptContext(job);
887       HFileOutputFormat2 hof = new HFileOutputFormat2();
888       writer = hof.getRecordWriter(context);
889 
890       // write out random rows
891       writeRandomKeyValues(writer, context, htd.getFamiliesKeys(), ROWSPERSPLIT);
892       writer.close(context);
893 
894       // Make sure that a directory was created for every CF
895       FileSystem fs = dir.getFileSystem(conf);
896 
897       // commit so that the filesystem has one directory per column family
898       hof.getOutputCommitter(context).commitTask(context);
899       hof.getOutputCommitter(context).commitJob(context);
900       FileStatus[] families = FSUtils.listStatus(fs, dir, new FSUtils.FamilyDirFilter(fs));
901       assertEquals(htd.getFamilies().size(), families.length);
902       for (FileStatus f : families) {
903         String familyStr = f.getPath().getName();
904         HColumnDescriptor hcd = htd.getFamily(Bytes.toBytes(familyStr));
905         // verify that the compression on this file matches the configured
906         // compression
907         Path dataFilePath = fs.listStatus(f.getPath())[0].getPath();
908         Reader reader = HFile.createReader(fs, dataFilePath, new CacheConfig(conf), conf);
909         Map<byte[], byte[]> fileInfo = reader.loadFileInfo();
910 
911         byte[] bloomFilter = fileInfo.get(StoreFile.BLOOM_FILTER_TYPE_KEY);
912         if (bloomFilter == null) bloomFilter = Bytes.toBytes("NONE");
913         assertEquals("Incorrect bloom filter used for column family " + familyStr +
914           "(reader: " + reader + ")",
915           hcd.getBloomFilterType(), BloomType.valueOf(Bytes.toString(bloomFilter)));
916         assertEquals("Incorrect compression used for column family " + familyStr +
917           "(reader: " + reader + ")", hcd.getCompressionType(), reader.getFileContext().getCompression());
918       }
919     } finally {
920       dir.getFileSystem(conf).delete(dir, true);
921     }
922   }
923 
924   /**
925    * Write random values to the writer assuming a table created using
926    * {@link #FAMILIES} as column family descriptors
927    */
928   private void writeRandomKeyValues(RecordWriter<ImmutableBytesWritable, Cell> writer,
929       TaskAttemptContext context, Set<byte[]> families, int numRows)
930       throws IOException, InterruptedException {
931     byte keyBytes[] = new byte[Bytes.SIZEOF_INT];
932     int valLength = 10;
933     byte valBytes[] = new byte[valLength];
934 
935     int taskId = context.getTaskAttemptID().getTaskID().getId();
936     assert taskId < Byte.MAX_VALUE : "Unit tests dont support > 127 tasks!";
937     final byte [] qualifier = Bytes.toBytes("data");
938     Random random = new Random();
939     for (int i = 0; i < numRows; i++) {
940 
941       Bytes.putInt(keyBytes, 0, i);
942       random.nextBytes(valBytes);
943       ImmutableBytesWritable key = new ImmutableBytesWritable(keyBytes);
944 
945       for (byte[] family : families) {
946         Cell kv = new KeyValue(keyBytes, family, qualifier, valBytes);
947         writer.write(key, kv);
948       }
949     }
950   }
951 
952   /**
953    * This test is to test the scenario happened in HBASE-6901.
954    * All files are bulk loaded and excluded from minor compaction.
955    * Without the fix of HBASE-6901, an ArrayIndexOutOfBoundsException
956    * will be thrown.
957    */
958   @Ignore ("Flakey: See HBASE-9051") @Test
959   public void testExcludeAllFromMinorCompaction() throws Exception {
960     Configuration conf = util.getConfiguration();
961     conf.setInt("hbase.hstore.compaction.min", 2);
962     generateRandomStartKeys(5);
963 
964     util.startMiniCluster();
965     try (Connection conn = ConnectionFactory.createConnection();
966         Admin admin = conn.getAdmin();
967         Table table = util.createTable(TABLE_NAME, FAMILIES);
968         RegionLocator locator = conn.getRegionLocator(TABLE_NAME)) {
969       final FileSystem fs = util.getDFSCluster().getFileSystem();
970       assertEquals("Should start with empty table", 0, util.countRows(table));
971 
972       // deep inspection: get the StoreFile dir
973       final Path storePath = new Path(
974         FSUtils.getTableDir(FSUtils.getRootDir(conf), TABLE_NAME),
975           new Path(admin.getTableRegions(TABLE_NAME).get(0).getEncodedName(),
976             Bytes.toString(FAMILIES[0])));
977       assertEquals(0, fs.listStatus(storePath).length);
978 
979       // Generate two bulk load files
980       conf.setBoolean("hbase.mapreduce.hfileoutputformat.compaction.exclude",
981           true);
982 
983       for (int i = 0; i < 2; i++) {
984         Path testDir = util.getDataTestDirOnTestFS("testExcludeAllFromMinorCompaction_" + i);
985         runIncrementalPELoad(conf, table.getTableDescriptor(), conn.getRegionLocator(TABLE_NAME),
986             testDir);
987         // Perform the actual load
988         new LoadIncrementalHFiles(conf).doBulkLoad(testDir, admin, table, locator);
989       }
990 
991       // Ensure data shows up
992       int expectedRows = 2 * NMapInputFormat.getNumMapTasks(conf) * ROWSPERSPLIT;
993       assertEquals("LoadIncrementalHFiles should put expected data in table",
994           expectedRows, util.countRows(table));
995 
996       // should have a second StoreFile now
997       assertEquals(2, fs.listStatus(storePath).length);
998 
999       // minor compactions shouldn't get rid of the file
1000       admin.compact(TABLE_NAME);
1001       try {
1002         quickPoll(new Callable<Boolean>() {
1003           @Override
1004           public Boolean call() throws Exception {
1005             return fs.listStatus(storePath).length == 1;
1006           }
1007         }, 5000);
1008         throw new IOException("SF# = " + fs.listStatus(storePath).length);
1009       } catch (AssertionError ae) {
1010         // this is expected behavior
1011       }
1012 
1013       // a major compaction should work though
1014       admin.majorCompact(TABLE_NAME);
1015       quickPoll(new Callable<Boolean>() {
1016         @Override
1017         public Boolean call() throws Exception {
1018           return fs.listStatus(storePath).length == 1;
1019         }
1020       }, 5000);
1021 
1022     } finally {
1023       util.shutdownMiniCluster();
1024     }
1025   }
1026 
1027   @Ignore("Goes zombie too frequently; needs work. See HBASE-14563") @Test
1028   public void testExcludeMinorCompaction() throws Exception {
1029     Configuration conf = util.getConfiguration();
1030     conf.setInt("hbase.hstore.compaction.min", 2);
1031     generateRandomStartKeys(5);
1032 
1033     util.startMiniCluster();
1034     try (Connection conn = ConnectionFactory.createConnection(conf);
1035         Admin admin = conn.getAdmin()){
1036       Path testDir = util.getDataTestDirOnTestFS("testExcludeMinorCompaction");
1037       final FileSystem fs = util.getDFSCluster().getFileSystem();
1038       Table table = util.createTable(TABLE_NAME, FAMILIES);
1039       assertEquals("Should start with empty table", 0, util.countRows(table));
1040 
1041       // deep inspection: get the StoreFile dir
1042       final Path storePath = new Path(
1043         FSUtils.getTableDir(FSUtils.getRootDir(conf), TABLE_NAME),
1044           new Path(admin.getTableRegions(TABLE_NAME).get(0).getEncodedName(),
1045             Bytes.toString(FAMILIES[0])));
1046       assertEquals(0, fs.listStatus(storePath).length);
1047 
1048       // put some data in it and flush to create a storefile
1049       Put p = new Put(Bytes.toBytes("test"));
1050       p.addColumn(FAMILIES[0], Bytes.toBytes("1"), Bytes.toBytes("1"));
1051       table.put(p);
1052       admin.flush(TABLE_NAME);
1053       assertEquals(1, util.countRows(table));
1054       quickPoll(new Callable<Boolean>() {
1055         @Override
1056         public Boolean call() throws Exception {
1057           return fs.listStatus(storePath).length == 1;
1058         }
1059       }, 5000);
1060 
1061       // Generate a bulk load file with more rows
1062       conf.setBoolean("hbase.mapreduce.hfileoutputformat.compaction.exclude",
1063           true);
1064 
1065       RegionLocator regionLocator = conn.getRegionLocator(TABLE_NAME);
1066       runIncrementalPELoad(conf, table.getTableDescriptor(), regionLocator, testDir);
1067 
1068       // Perform the actual load
1069       new LoadIncrementalHFiles(conf).doBulkLoad(testDir, admin, table, regionLocator);
1070 
1071       // Ensure data shows up
1072       int expectedRows = NMapInputFormat.getNumMapTasks(conf) * ROWSPERSPLIT;
1073       assertEquals("LoadIncrementalHFiles should put expected data in table",
1074           expectedRows + 1, util.countRows(table));
1075 
1076       // should have a second StoreFile now
1077       assertEquals(2, fs.listStatus(storePath).length);
1078 
1079       // minor compactions shouldn't get rid of the file
1080       admin.compact(TABLE_NAME);
1081       try {
1082         quickPoll(new Callable<Boolean>() {
1083           @Override
1084           public Boolean call() throws Exception {
1085             return fs.listStatus(storePath).length == 1;
1086           }
1087         }, 5000);
1088         throw new IOException("SF# = " + fs.listStatus(storePath).length);
1089       } catch (AssertionError ae) {
1090         // this is expected behavior
1091       }
1092 
1093       // a major compaction should work though
1094       admin.majorCompact(TABLE_NAME);
1095       quickPoll(new Callable<Boolean>() {
1096         @Override
1097         public Boolean call() throws Exception {
1098           return fs.listStatus(storePath).length == 1;
1099         }
1100       }, 5000);
1101 
1102     } finally {
1103       util.shutdownMiniCluster();
1104     }
1105   }
1106 
1107   private void quickPoll(Callable<Boolean> c, int waitMs) throws Exception {
1108     int sleepMs = 10;
1109     int retries = (int) Math.ceil(((double) waitMs) / sleepMs);
1110     while (retries-- > 0) {
1111       if (c.call().booleanValue()) {
1112         return;
1113       }
1114       Thread.sleep(sleepMs);
1115     }
1116     fail();
1117   }
1118 
1119   public static void main(String args[]) throws Exception {
1120     new TestHFileOutputFormat2().manualTest(args);
1121   }
1122 
1123   public void manualTest(String args[]) throws Exception {
1124     Configuration conf = HBaseConfiguration.create();
1125     util = new HBaseTestingUtility(conf);
1126     if ("newtable".equals(args[0])) {
1127       TableName tname = TableName.valueOf(args[1]);
1128       byte[][] splitKeys = generateRandomSplitKeys(4);
1129       try (Table table = util.createTable(tname, FAMILIES, splitKeys)) {
1130       }
1131     } else if ("incremental".equals(args[0])) {
1132       TableName tname = TableName.valueOf(args[1]);
1133       try(Connection c = ConnectionFactory.createConnection(conf);
1134           Admin admin = c.getAdmin();
1135           RegionLocator regionLocator = c.getRegionLocator(tname)) {
1136         Path outDir = new Path("incremental-out");
1137         runIncrementalPELoad(conf, admin.getTableDescriptor(tname), regionLocator, outDir);
1138       }
1139     } else {
1140       throw new RuntimeException(
1141           "usage: TestHFileOutputFormat2 newtable | incremental");
1142     }
1143   }
1144 
1145 }
1146