001/* 002 * Licensed to the Apache Software Foundation (ASF) under one 003 * or more contributor license agreements. See the NOTICE file 004 * distributed with this work for additional information 005 * regarding copyright ownership. The ASF licenses this file 006 * to you under the Apache License, Version 2.0 (the 007 * "License"); you may not use this file except in compliance 008 * with the License. You may obtain a copy of the License at 009 * 010 * http://www.apache.org/licenses/LICENSE-2.0 011 * 012 * Unless required by applicable law or agreed to in writing, software 013 * distributed under the License is distributed on an "AS IS" BASIS, 014 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 015 * See the License for the specific language governing permissions and 016 * limitations under the License. 017 */ 018package org.apache.hadoop.hbase.io.hfile; 019 020import static org.junit.Assert.assertEquals; 021import static org.junit.Assert.assertFalse; 022import static org.junit.Assert.assertTrue; 023 024import java.io.ByteArrayOutputStream; 025import java.io.DataOutput; 026import java.io.DataOutputStream; 027import java.io.IOException; 028import java.nio.ByteBuffer; 029import java.util.ArrayList; 030import java.util.Arrays; 031import java.util.Collection; 032import java.util.HashSet; 033import java.util.List; 034import java.util.Random; 035import java.util.Set; 036import java.util.concurrent.atomic.AtomicInteger; 037import org.apache.hadoop.conf.Configuration; 038import org.apache.hadoop.fs.FSDataInputStream; 039import org.apache.hadoop.fs.FSDataOutputStream; 040import org.apache.hadoop.fs.FileSystem; 041import org.apache.hadoop.fs.Path; 042import org.apache.hadoop.hbase.CellBuilderType; 043import org.apache.hadoop.hbase.CellComparatorImpl; 044import org.apache.hadoop.hbase.CellUtil; 045import org.apache.hadoop.hbase.ExtendedCellBuilderFactory; 046import org.apache.hadoop.hbase.HBaseClassTestRule; 047import org.apache.hadoop.hbase.HBaseCommonTestingUtil; 048import org.apache.hadoop.hbase.HBaseTestingUtil; 049import org.apache.hadoop.hbase.HConstants; 050import org.apache.hadoop.hbase.KeyValue; 051import org.apache.hadoop.hbase.KeyValueUtil; 052import org.apache.hadoop.hbase.PrivateCellUtil; 053import org.apache.hadoop.hbase.fs.HFileSystem; 054import org.apache.hadoop.hbase.io.ByteBuffAllocator; 055import org.apache.hadoop.hbase.io.FSDataInputStreamWrapper; 056import org.apache.hadoop.hbase.io.compress.Compression; 057import org.apache.hadoop.hbase.io.compress.Compression.Algorithm; 058import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding; 059import org.apache.hadoop.hbase.io.encoding.IndexBlockEncoding; 060import org.apache.hadoop.hbase.io.hfile.HFile.Writer; 061import org.apache.hadoop.hbase.io.hfile.HFileBlockIndex.BlockIndexReader; 062import org.apache.hadoop.hbase.io.hfile.NoOpIndexBlockEncoder.NoOpEncodedSeeker; 063import org.apache.hadoop.hbase.nio.ByteBuff; 064import org.apache.hadoop.hbase.nio.MultiByteBuff; 065import org.apache.hadoop.hbase.nio.RefCnt; 066import org.apache.hadoop.hbase.testclassification.IOTests; 067import org.apache.hadoop.hbase.testclassification.MediumTests; 068import org.apache.hadoop.hbase.util.Bytes; 069import org.apache.hadoop.hbase.util.ClassSize; 070import org.apache.hadoop.hbase.util.EnvironmentEdgeManager; 071import org.junit.Before; 072import org.junit.ClassRule; 073import org.junit.Test; 074import org.junit.experimental.categories.Category; 075import org.junit.runner.RunWith; 076import org.junit.runners.Parameterized; 077import org.junit.runners.Parameterized.Parameters; 078import org.slf4j.Logger; 079import org.slf4j.LoggerFactory; 080 081import org.apache.hbase.thirdparty.io.netty.util.ResourceLeakDetector; 082 083@RunWith(Parameterized.class) 084@Category({ IOTests.class, MediumTests.class }) 085public class TestHFileBlockIndex { 086 087 @ClassRule 088 public static final HBaseClassTestRule CLASS_RULE = 089 HBaseClassTestRule.forClass(TestHFileBlockIndex.class); 090 091 @Parameters 092 public static Collection<Object[]> compressionAlgorithms() { 093 return HBaseCommonTestingUtil.COMPRESSION_ALGORITHMS_PARAMETERIZED; 094 } 095 096 public TestHFileBlockIndex(Compression.Algorithm compr) { 097 this.compr = compr; 098 } 099 100 private static final Logger LOG = LoggerFactory.getLogger(TestHFileBlockIndex.class); 101 private static final Random RNG = new Random(); // This test depends on Random#setSeed 102 private static final int NUM_DATA_BLOCKS = 1000; 103 private static final HBaseTestingUtil TEST_UTIL = new HBaseTestingUtil(); 104 105 private static final int SMALL_BLOCK_SIZE = 4096; 106 private static final int NUM_KV = 10000; 107 108 private static FileSystem fs; 109 private Path path; 110 private long rootIndexOffset; 111 private int numRootEntries; 112 private int numLevels; 113 private static final List<byte[]> keys = new ArrayList<>(); 114 private final Compression.Algorithm compr; 115 private byte[] firstKeyInFile; 116 private Configuration conf; 117 118 private static final int[] INDEX_CHUNK_SIZES = { 4096, 512, 384 }; 119 private static final int[] EXPECTED_NUM_LEVELS = { 2, 3, 4 }; 120 private static final int[] UNCOMPRESSED_INDEX_SIZES = { 19187, 21813, 23086 }; 121 122 private static final boolean includesMemstoreTS = true; 123 124 static { 125 assert INDEX_CHUNK_SIZES.length == EXPECTED_NUM_LEVELS.length; 126 assert INDEX_CHUNK_SIZES.length == UNCOMPRESSED_INDEX_SIZES.length; 127 } 128 129 @Before 130 public void setUp() throws IOException { 131 keys.clear(); 132 firstKeyInFile = null; 133 conf = TEST_UTIL.getConfiguration(); 134 RNG.setSeed(2389757); 135 136 // This test requires at least HFile format version 2. 137 conf.setInt(HFile.FORMAT_VERSION_KEY, HFile.MAX_FORMAT_VERSION); 138 139 fs = HFileSystem.get(conf); 140 } 141 142 @Test 143 public void testBlockIndex() throws IOException { 144 testBlockIndexInternals(false); 145 clear(); 146 testBlockIndexInternals(true); 147 } 148 149 private void writeDataBlocksAndCreateIndex(HFileBlock.Writer hbw, FSDataOutputStream outputStream, 150 HFileBlockIndex.BlockIndexWriter biw) throws IOException { 151 for (int i = 0; i < NUM_DATA_BLOCKS; ++i) { 152 hbw.startWriting(BlockType.DATA).write(Bytes.toBytes(String.valueOf(RNG.nextInt(1000)))); 153 long blockOffset = outputStream.getPos(); 154 hbw.writeHeaderAndData(outputStream); 155 156 byte[] firstKey = null; 157 byte[] family = Bytes.toBytes("f"); 158 byte[] qualifier = Bytes.toBytes("q"); 159 for (int j = 0; j < 16; ++j) { 160 byte[] k = new KeyValue(RandomKeyValueUtil.randomOrderedKey(RNG, i * 16 + j), family, 161 qualifier, EnvironmentEdgeManager.currentTime(), KeyValue.Type.Put).getKey(); 162 keys.add(k); 163 if (j == 8) { 164 firstKey = k; 165 } 166 } 167 assertTrue(firstKey != null); 168 if (firstKeyInFile == null) { 169 firstKeyInFile = firstKey; 170 } 171 biw.addEntry(firstKey, blockOffset, hbw.getOnDiskSizeWithHeader()); 172 173 writeInlineBlocks(hbw, outputStream, biw, false); 174 } 175 writeInlineBlocks(hbw, outputStream, biw, true); 176 rootIndexOffset = biw.writeIndexBlocks(outputStream); 177 outputStream.close(); 178 } 179 180 @Test 181 public void testBlockIndexWithOffHeapBuffer() throws Exception { 182 ResourceLeakDetector.setLevel(ResourceLeakDetector.Level.PARANOID); 183 path = new Path(TEST_UTIL.getDataTestDir(), "block_index_testBlockIndexWithOffHeapBuffer"); 184 assertEquals(0, keys.size()); 185 HFileContext meta = new HFileContextBuilder().withHBaseCheckSum(true) 186 .withIncludesMvcc(includesMemstoreTS).withIncludesTags(true).withCompression(compr) 187 .withBytesPerCheckSum(HFile.DEFAULT_BYTES_PER_CHECKSUM).build(); 188 ByteBuffAllocator allocator = ByteBuffAllocator.create(TEST_UTIL.getConfiguration(), true); 189 HFileBlock.Writer hbw = new HFileBlock.Writer(TEST_UTIL.getConfiguration(), null, meta, 190 allocator, meta.getBlocksize()); 191 FSDataOutputStream outputStream = fs.create(path); 192 193 final AtomicInteger counter = new AtomicInteger(); 194 RefCnt.detector.setLeakListener(new ResourceLeakDetector.LeakListener() { 195 @Override 196 public void onLeak(String s, String s1) { 197 counter.incrementAndGet(); 198 } 199 }); 200 201 long maxSize = NUM_DATA_BLOCKS * 1000; 202 long blockSize = 1000; 203 LruBlockCache cache = new LruBlockCache(maxSize, blockSize); 204 CacheConfig cacheConfig = new CacheConfig(TEST_UTIL.getConfiguration(), null, cache, allocator); 205 206 HFileBlockIndex.BlockIndexWriter biw = 207 new HFileBlockIndex.BlockIndexWriter(hbw, cacheConfig, path.getName(), null); 208 209 writeDataBlocksAndCreateIndex(hbw, outputStream, biw); 210 211 System.gc(); 212 Thread.sleep(1000); 213 214 allocator.allocate(128 * 1024).release(); 215 216 assertEquals(0, counter.get()); 217 } 218 219 private void clear() throws IOException { 220 keys.clear(); 221 firstKeyInFile = null; 222 conf = TEST_UTIL.getConfiguration(); 223 RNG.setSeed(2389757); 224 225 // This test requires at least HFile format version 2. 226 conf.setInt(HFile.FORMAT_VERSION_KEY, 3); 227 228 fs = HFileSystem.get(conf); 229 } 230 231 private void testBlockIndexInternals(boolean useTags) throws IOException { 232 path = new Path(TEST_UTIL.getDataTestDir(), "block_index_" + compr + useTags); 233 writeWholeIndex(useTags); 234 readIndex(useTags); 235 } 236 237 /** 238 * A wrapper around a block reader which only caches the results of the last operation. Not 239 * thread-safe. 240 */ 241 private static class BlockReaderWrapper implements HFile.CachingBlockReader { 242 243 private HFileBlock.FSReader realReader; 244 private long prevOffset; 245 private long prevOnDiskSize; 246 private boolean prevPread; 247 private HFileBlock prevBlock; 248 249 public int hitCount = 0; 250 public int missCount = 0; 251 252 public BlockReaderWrapper(HFileBlock.FSReader realReader) { 253 this.realReader = realReader; 254 } 255 256 @Override 257 public HFileBlock readBlock(long offset, long onDiskSize, boolean cacheBlock, boolean pread, 258 boolean isCompaction, boolean updateCacheMetrics, BlockType expectedBlockType, 259 DataBlockEncoding expectedDataBlockEncoding) throws IOException { 260 return readBlock(offset, onDiskSize, cacheBlock, pread, isCompaction, updateCacheMetrics, 261 expectedBlockType, expectedDataBlockEncoding, false); 262 } 263 264 @Override 265 public HFileBlock readBlock(long offset, long onDiskSize, boolean cacheBlock, boolean pread, 266 boolean isCompaction, boolean updateCacheMetrics, BlockType expectedBlockType, 267 DataBlockEncoding expectedDataBlockEncoding, boolean cacheOnly) throws IOException { 268 if (offset == prevOffset && onDiskSize == prevOnDiskSize && pread == prevPread) { 269 hitCount += 1; 270 return prevBlock; 271 } 272 273 missCount += 1; 274 prevBlock = realReader.readBlockData(offset, onDiskSize, pread, false, true); 275 prevOffset = offset; 276 prevOnDiskSize = onDiskSize; 277 prevPread = pread; 278 279 return prevBlock; 280 } 281 } 282 283 private void readIndex(boolean useTags) throws IOException { 284 long fileSize = fs.getFileStatus(path).getLen(); 285 LOG.info("Size of {}: {} compression={}", path, fileSize, compr.toString()); 286 287 FSDataInputStream istream = fs.open(path); 288 HFileContext meta = 289 new HFileContextBuilder().withHBaseCheckSum(true).withIncludesMvcc(includesMemstoreTS) 290 .withIncludesTags(useTags).withCompression(compr).build(); 291 ReaderContext context = new ReaderContextBuilder().withFileSystemAndPath(fs, path).build(); 292 HFileBlock.FSReader blockReader = 293 new HFileBlock.FSReaderImpl(context, meta, ByteBuffAllocator.HEAP, conf); 294 295 BlockReaderWrapper brw = new BlockReaderWrapper(blockReader); 296 HFileBlockIndex.BlockIndexReader indexReader = 297 new HFileBlockIndex.CellBasedKeyBlockIndexReader(CellComparatorImpl.COMPARATOR, numLevels); 298 299 indexReader.readRootIndex(blockReader.blockRange(rootIndexOffset, fileSize) 300 .nextBlockWithBlockType(BlockType.ROOT_INDEX), numRootEntries); 301 302 long prevOffset = -1; 303 int i = 0; 304 int expectedHitCount = 0; 305 int expectedMissCount = 0; 306 LOG.info("Total number of keys: " + keys.size()); 307 for (byte[] key : keys) { 308 assertTrue(key != null); 309 assertTrue(indexReader != null); 310 KeyValue.KeyOnlyKeyValue keyOnlyKey = new KeyValue.KeyOnlyKeyValue(key, 0, key.length); 311 HFileBlock b = indexReader.seekToDataBlock(keyOnlyKey, null, true, true, false, null, brw); 312 if ( 313 PrivateCellUtil.compare(CellComparatorImpl.COMPARATOR, keyOnlyKey, firstKeyInFile, 0, 314 firstKeyInFile.length) < 0 315 ) { 316 assertTrue(b == null); 317 ++i; 318 continue; 319 } 320 321 String keyStr = "key #" + i + ", " + Bytes.toStringBinary(key); 322 323 assertTrue("seekToDataBlock failed for " + keyStr, b != null); 324 325 if (prevOffset == b.getOffset()) { 326 assertEquals(++expectedHitCount, brw.hitCount); 327 } else { 328 LOG.info("First key in a new block: " + keyStr + ", block offset: " + b.getOffset() + ")"); 329 assertTrue(b.getOffset() > prevOffset); 330 assertEquals(++expectedMissCount, brw.missCount); 331 prevOffset = b.getOffset(); 332 } 333 ++i; 334 } 335 336 istream.close(); 337 } 338 339 private void writeWholeIndex(boolean useTags) throws IOException { 340 assertEquals(0, keys.size()); 341 HFileContext meta = new HFileContextBuilder().withHBaseCheckSum(true) 342 .withIncludesMvcc(includesMemstoreTS).withIncludesTags(useTags).withCompression(compr) 343 .withBytesPerCheckSum(HFile.DEFAULT_BYTES_PER_CHECKSUM).build(); 344 HFileBlock.Writer hbw = new HFileBlock.Writer(TEST_UTIL.getConfiguration(), null, meta); 345 FSDataOutputStream outputStream = fs.create(path); 346 HFileBlockIndex.BlockIndexWriter biw = 347 new HFileBlockIndex.BlockIndexWriter(hbw, null, null, null); 348 writeDataBlocksAndCreateIndex(hbw, outputStream, biw); 349 350 numLevels = biw.getNumLevels(); 351 numRootEntries = biw.getNumRootEntries(); 352 353 LOG.info("Index written: numLevels=" + numLevels + ", numRootEntries=" + numRootEntries 354 + ", rootIndexOffset=" + rootIndexOffset); 355 } 356 357 private void writeInlineBlocks(HFileBlock.Writer hbw, FSDataOutputStream outputStream, 358 HFileBlockIndex.BlockIndexWriter biw, boolean isClosing) throws IOException { 359 while (biw.shouldWriteBlock(isClosing)) { 360 long offset = outputStream.getPos(); 361 biw.writeInlineBlock(hbw.startWriting(biw.getInlineBlockType())); 362 hbw.writeHeaderAndData(outputStream); 363 biw.blockWritten(offset, hbw.getOnDiskSizeWithHeader(), 364 hbw.getUncompressedSizeWithoutHeader()); 365 LOG.info( 366 "Wrote an inline index block at " + offset + ", size " + hbw.getOnDiskSizeWithHeader()); 367 } 368 } 369 370 private static final long getDummyFileOffset(int i) { 371 return i * 185 + 379; 372 } 373 374 private static final int getDummyOnDiskSize(int i) { 375 return i * i * 37 + i * 19 + 13; 376 } 377 378 @Test 379 public void testSecondaryIndexBinarySearch() throws IOException { 380 int numTotalKeys = 99; 381 assertTrue(numTotalKeys % 2 == 1); // Ensure no one made this even. 382 383 // We only add odd-index keys into the array that we will binary-search. 384 int numSearchedKeys = (numTotalKeys - 1) / 2; 385 386 ByteArrayOutputStream baos = new ByteArrayOutputStream(); 387 DataOutputStream dos = new DataOutputStream(baos); 388 389 dos.writeInt(numSearchedKeys); 390 int curAllEntriesSize = 0; 391 int numEntriesAdded = 0; 392 393 // Only odd-index elements of this array are used to keep the secondary 394 // index entries of the corresponding keys. 395 int secondaryIndexEntries[] = new int[numTotalKeys]; 396 397 for (int i = 0; i < numTotalKeys; ++i) { 398 byte[] k = RandomKeyValueUtil.randomOrderedKey(RNG, i * 2); 399 KeyValue cell = new KeyValue(k, Bytes.toBytes("f"), Bytes.toBytes("q"), Bytes.toBytes("val")); 400 // KeyValue cell = new KeyValue.KeyOnlyKeyValue(k, 0, k.length); 401 keys.add(cell.getKey()); 402 String msgPrefix = "Key #" + i + " (" + Bytes.toStringBinary(k) + "): "; 403 StringBuilder padding = new StringBuilder(); 404 while (msgPrefix.length() + padding.length() < 70) 405 padding.append(' '); 406 msgPrefix += padding; 407 if (i % 2 == 1) { 408 dos.writeInt(curAllEntriesSize); 409 secondaryIndexEntries[i] = curAllEntriesSize; 410 LOG.info( 411 msgPrefix + "secondary index entry #" + ((i - 1) / 2) + ", offset " + curAllEntriesSize); 412 curAllEntriesSize += cell.getKey().length + HFileBlockIndex.SECONDARY_INDEX_ENTRY_OVERHEAD; 413 ++numEntriesAdded; 414 } else { 415 secondaryIndexEntries[i] = -1; 416 LOG.info(msgPrefix + "not in the searched array"); 417 } 418 } 419 420 // Make sure the keys are increasing. 421 for (int i = 0; i < keys.size() - 1; ++i) 422 assertTrue(CellComparatorImpl.COMPARATOR.compare( 423 new KeyValue.KeyOnlyKeyValue(keys.get(i), 0, keys.get(i).length), 424 new KeyValue.KeyOnlyKeyValue(keys.get(i + 1), 0, keys.get(i + 1).length)) < 0); 425 426 dos.writeInt(curAllEntriesSize); 427 assertEquals(numSearchedKeys, numEntriesAdded); 428 int secondaryIndexOffset = dos.size(); 429 assertEquals(Bytes.SIZEOF_INT * (numSearchedKeys + 2), secondaryIndexOffset); 430 431 for (int i = 1; i <= numTotalKeys - 1; i += 2) { 432 assertEquals(dos.size(), secondaryIndexOffset + secondaryIndexEntries[i]); 433 long dummyFileOffset = getDummyFileOffset(i); 434 int dummyOnDiskSize = getDummyOnDiskSize(i); 435 LOG.debug("Storing file offset=" + dummyFileOffset + " and onDiskSize=" + dummyOnDiskSize 436 + " at offset " + dos.size()); 437 dos.writeLong(dummyFileOffset); 438 dos.writeInt(dummyOnDiskSize); 439 LOG.debug("Stored key " + ((i - 1) / 2) + " at offset " + dos.size()); 440 dos.write(keys.get(i)); 441 } 442 443 dos.writeInt(curAllEntriesSize); 444 445 ByteBuffer nonRootIndex = ByteBuffer.wrap(baos.toByteArray()); 446 for (int i = 0; i < numTotalKeys; ++i) { 447 byte[] searchKey = keys.get(i); 448 byte[] arrayHoldingKey = new byte[searchKey.length + searchKey.length / 2]; 449 450 // To make things a bit more interesting, store the key we are looking 451 // for at a non-zero offset in a new array. 452 System.arraycopy(searchKey, 0, arrayHoldingKey, searchKey.length / 2, searchKey.length); 453 454 KeyValue.KeyOnlyKeyValue cell = 455 new KeyValue.KeyOnlyKeyValue(arrayHoldingKey, searchKey.length / 2, searchKey.length); 456 int searchResult = BlockIndexReader.binarySearchNonRootIndex(cell, 457 new MultiByteBuff(nonRootIndex), CellComparatorImpl.COMPARATOR); 458 String lookupFailureMsg = 459 "Failed to look up key #" + i + " (" + Bytes.toStringBinary(searchKey) + ")"; 460 461 int expectedResult; 462 int referenceItem; 463 464 if (i % 2 == 1) { 465 // This key is in the array we search as the element (i - 1) / 2. Make 466 // sure we find it. 467 expectedResult = (i - 1) / 2; 468 referenceItem = i; 469 } else { 470 // This key is not in the array but between two elements on the array, 471 // in the beginning, or in the end. The result should be the previous 472 // key in the searched array, or -1 for i = 0. 473 expectedResult = i / 2 - 1; 474 referenceItem = i - 1; 475 } 476 477 assertEquals(lookupFailureMsg, expectedResult, searchResult); 478 479 // Now test we can get the offset and the on-disk-size using a 480 // higher-level API function.s 481 boolean locateBlockResult = 482 (BlockIndexReader.locateNonRootIndexEntry(new MultiByteBuff(nonRootIndex), cell, 483 CellComparatorImpl.COMPARATOR) != -1); 484 485 if (i == 0) { 486 assertFalse(locateBlockResult); 487 } else { 488 assertTrue(locateBlockResult); 489 String errorMsg = "i=" + i + ", position=" + nonRootIndex.position(); 490 assertEquals(errorMsg, getDummyFileOffset(referenceItem), nonRootIndex.getLong()); 491 assertEquals(errorMsg, getDummyOnDiskSize(referenceItem), nonRootIndex.getInt()); 492 } 493 } 494 495 } 496 497 @Test 498 public void testBlockIndexChunk() throws IOException { 499 BlockIndexChunk c = new HFileBlockIndex.BlockIndexChunkImpl(); 500 HFileIndexBlockEncoder indexBlockEncoder = NoOpIndexBlockEncoder.INSTANCE; 501 ByteArrayOutputStream baos = new ByteArrayOutputStream(); 502 int N = 1000; 503 int[] numSubEntriesAt = new int[N]; 504 int numSubEntries = 0; 505 for (int i = 0; i < N; ++i) { 506 baos.reset(); 507 DataOutputStream dos = new DataOutputStream(baos); 508 indexBlockEncoder.encode(c, false, dos); 509 assertEquals(c.getNonRootSize(), dos.size()); 510 511 baos.reset(); 512 dos = new DataOutputStream(baos); 513 indexBlockEncoder.encode(c, true, dos); 514 assertEquals(c.getRootSize(), dos.size()); 515 516 byte[] k = RandomKeyValueUtil.randomOrderedKey(RNG, i); 517 numSubEntries += RNG.nextInt(5) + 1; 518 keys.add(k); 519 c.add(k, getDummyFileOffset(i), getDummyOnDiskSize(i), numSubEntries); 520 } 521 522 // Test the ability to look up the entry that contains a particular 523 // deeper-level index block's entry ("sub-entry"), assuming a global 524 // 0-based ordering of sub-entries. This is needed for mid-key calculation. 525 for (int i = 0; i < N; ++i) { 526 for (int j = i == 0 ? 0 : numSubEntriesAt[i - 1]; j < numSubEntriesAt[i]; ++j) { 527 assertEquals(i, c.getEntryBySubEntry(j)); 528 } 529 } 530 } 531 532 /** Checks if the HeapSize calculator is within reason */ 533 @Test 534 public void testHeapSizeForBlockIndex() throws IOException { 535 Class<HFileBlockIndex.BlockIndexReader> cl = HFileBlockIndex.BlockIndexReader.class; 536 long expected = ClassSize.estimateBase(cl, false); 537 538 HFileBlockIndex.BlockIndexReader bi = new HFileBlockIndex.ByteArrayKeyBlockIndexReader(1); 539 long actual = bi.heapSize(); 540 541 // Since the arrays in BlockIndex(byte [][] blockKeys, long [] blockOffsets, 542 // int [] blockDataSizes) are all null they are not going to show up in the 543 // HeapSize calculation, so need to remove those array costs from expected. 544 // Already the block keys are not there in this case 545 expected -= ClassSize.align(2 * ClassSize.ARRAY); 546 547 if (expected != actual) { 548 expected = ClassSize.estimateBase(cl, true); 549 assertEquals(expected, actual); 550 } 551 } 552 553 /** 554 * to check if looks good when midKey on a leaf index block boundary 555 */ 556 @Test 557 public void testMidKeyOnLeafIndexBlockBoundary() throws IOException { 558 Path hfilePath = new Path(TEST_UTIL.getDataTestDir(), "hfile_for_midkey"); 559 int maxChunkSize = 512; 560 conf.setInt(HFileBlockIndex.MAX_CHUNK_SIZE_KEY, maxChunkSize); 561 // should open hfile.block.index.cacheonwrite 562 conf.setBoolean(CacheConfig.CACHE_INDEX_BLOCKS_ON_WRITE_KEY, true); 563 CacheConfig cacheConf = new CacheConfig(conf, BlockCacheFactory.createBlockCache(conf)); 564 BlockCache blockCache = cacheConf.getBlockCache().get(); 565 // Evict all blocks that were cached-on-write by the previous invocation. 566 blockCache.evictBlocksByHfileName(hfilePath.getName()); 567 // Write the HFile 568 HFileContext meta = new HFileContextBuilder().withBlockSize(SMALL_BLOCK_SIZE) 569 .withCompression(Algorithm.NONE).withDataBlockEncoding(DataBlockEncoding.NONE).build(); 570 HFile.Writer writer = HFile.getWriterFactory(conf, cacheConf).withPath(fs, hfilePath) 571 .withFileContext(meta).create(); 572 Random rand = new Random(19231737); 573 byte[] family = Bytes.toBytes("f"); 574 byte[] qualifier = Bytes.toBytes("q"); 575 int kvNumberToBeWritten = 16; 576 // the new generated hfile will contain 2 leaf-index blocks and 16 data blocks, 577 // midkey is just on the boundary of the first leaf-index block 578 for (int i = 0; i < kvNumberToBeWritten; ++i) { 579 byte[] row = RandomKeyValueUtil.randomOrderedFixedLengthKey(rand, i, 30); 580 581 // Key will be interpreted by KeyValue.KEY_COMPARATOR 582 KeyValue kv = new KeyValue(row, family, qualifier, EnvironmentEdgeManager.currentTime(), 583 RandomKeyValueUtil.randomFixedLengthValue(rand, SMALL_BLOCK_SIZE)); 584 writer.append(kv); 585 } 586 writer.close(); 587 588 // close hfile.block.index.cacheonwrite 589 conf.setBoolean(CacheConfig.CACHE_INDEX_BLOCKS_ON_WRITE_KEY, false); 590 591 // Read the HFile 592 HFile.Reader reader = HFile.createReader(fs, hfilePath, cacheConf, true, conf); 593 594 boolean hasArrayIndexOutOfBoundsException = false; 595 try { 596 // get the mid-key. 597 reader.midKey(); 598 } catch (ArrayIndexOutOfBoundsException e) { 599 hasArrayIndexOutOfBoundsException = true; 600 } finally { 601 reader.close(); 602 } 603 604 // to check if ArrayIndexOutOfBoundsException occurred 605 assertFalse(hasArrayIndexOutOfBoundsException); 606 } 607 608 /** 609 * Testing block index through the HFile writer/reader APIs. Allows to test setting index block 610 * size through configuration, intermediate-level index blocks, and caching index blocks on write. 611 */ 612 @Test 613 public void testHFileWriterAndReader() throws IOException { 614 Path hfilePath = new Path(TEST_UTIL.getDataTestDir(), "hfile_for_block_index"); 615 CacheConfig cacheConf = new CacheConfig(conf, BlockCacheFactory.createBlockCache(conf)); 616 BlockCache blockCache = cacheConf.getBlockCache().get(); 617 618 for (int testI = 0; testI < INDEX_CHUNK_SIZES.length; ++testI) { 619 int indexBlockSize = INDEX_CHUNK_SIZES[testI]; 620 int expectedNumLevels = EXPECTED_NUM_LEVELS[testI]; 621 LOG.info("Index block size: " + indexBlockSize + ", compression: " + compr); 622 // Evict all blocks that were cached-on-write by the previous invocation. 623 blockCache.evictBlocksByHfileName(hfilePath.getName()); 624 625 conf.setInt(HFileBlockIndex.MAX_CHUNK_SIZE_KEY, indexBlockSize); 626 Set<String> keyStrSet = new HashSet<>(); 627 byte[][] keys = new byte[NUM_KV][]; 628 byte[][] values = new byte[NUM_KV][]; 629 630 // Write the HFile 631 { 632 HFileContext meta = 633 new HFileContextBuilder().withBlockSize(SMALL_BLOCK_SIZE).withCompression(compr).build(); 634 HFile.Writer writer = HFile.getWriterFactory(conf, cacheConf).withPath(fs, hfilePath) 635 .withFileContext(meta).create(); 636 Random rand = new Random(19231737); 637 byte[] family = Bytes.toBytes("f"); 638 byte[] qualifier = Bytes.toBytes("q"); 639 for (int i = 0; i < NUM_KV; ++i) { 640 byte[] row = RandomKeyValueUtil.randomOrderedKey(rand, i); 641 642 // Key will be interpreted by KeyValue.KEY_COMPARATOR 643 KeyValue kv = new KeyValue(row, family, qualifier, EnvironmentEdgeManager.currentTime(), 644 RandomKeyValueUtil.randomValue(rand)); 645 byte[] k = kv.getKey(); 646 writer.append(kv); 647 keys[i] = k; 648 values[i] = CellUtil.cloneValue(kv); 649 keyStrSet.add(Bytes.toStringBinary(k)); 650 if (i > 0) { 651 assertTrue((PrivateCellUtil.compare(CellComparatorImpl.COMPARATOR, kv, keys[i - 1], 0, 652 keys[i - 1].length)) > 0); 653 } 654 } 655 656 writer.close(); 657 } 658 659 // Read the HFile 660 HFile.Reader reader = HFile.createReader(fs, hfilePath, cacheConf, true, conf); 661 assertEquals(expectedNumLevels, reader.getTrailer().getNumDataIndexLevels()); 662 663 assertTrue(Bytes.equals(keys[0], ((KeyValue) reader.getFirstKey().get()).getKey())); 664 assertTrue(Bytes.equals(keys[NUM_KV - 1], ((KeyValue) reader.getLastKey().get()).getKey())); 665 LOG.info("Last key: " + Bytes.toStringBinary(keys[NUM_KV - 1])); 666 667 for (boolean pread : new boolean[] { false, true }) { 668 HFileScanner scanner = reader.getScanner(conf, true, pread); 669 for (int i = 0; i < NUM_KV; ++i) { 670 checkSeekTo(keys, scanner, i); 671 checkKeyValue("i=" + i, keys[i], values[i], 672 ByteBuffer.wrap(((KeyValue) scanner.getKey()).getKey()), scanner.getValue()); 673 } 674 assertTrue(scanner.seekTo()); 675 for (int i = NUM_KV - 1; i >= 0; --i) { 676 checkSeekTo(keys, scanner, i); 677 checkKeyValue("i=" + i, keys[i], values[i], 678 ByteBuffer.wrap(((KeyValue) scanner.getKey()).getKey()), scanner.getValue()); 679 } 680 } 681 682 // Manually compute the mid-key and validate it. 683 HFile.Reader reader2 = reader; 684 HFileBlock.FSReader fsReader = reader2.getUncachedBlockReader(); 685 686 HFileBlock.BlockIterator iter = 687 fsReader.blockRange(0, reader.getTrailer().getLoadOnOpenDataOffset()); 688 HFileBlock block; 689 List<byte[]> blockKeys = new ArrayList<>(); 690 while ((block = iter.nextBlock()) != null) { 691 if (block.getBlockType() != BlockType.LEAF_INDEX) return; 692 ByteBuff b = block.getBufferReadOnly(); 693 int n = b.getIntAfterPosition(0); 694 // One int for the number of items, and n + 1 for the secondary index. 695 int entriesOffset = Bytes.SIZEOF_INT * (n + 2); 696 697 // Get all the keys from the leaf index block. S 698 for (int i = 0; i < n; ++i) { 699 int keyRelOffset = b.getIntAfterPosition(Bytes.SIZEOF_INT * (i + 1)); 700 int nextKeyRelOffset = b.getIntAfterPosition(Bytes.SIZEOF_INT * (i + 2)); 701 int keyLen = nextKeyRelOffset - keyRelOffset; 702 int keyOffset = b.arrayOffset() + entriesOffset + keyRelOffset 703 + HFileBlockIndex.SECONDARY_INDEX_ENTRY_OVERHEAD; 704 byte[] blockKey = Arrays.copyOfRange(b.array(), keyOffset, keyOffset + keyLen); 705 String blockKeyStr = Bytes.toString(blockKey); 706 blockKeys.add(blockKey); 707 708 // If the first key of the block is not among the keys written, we 709 // are not parsing the non-root index block format correctly. 710 assertTrue("Invalid block key from leaf-level block: " + blockKeyStr, 711 keyStrSet.contains(blockKeyStr)); 712 } 713 } 714 715 // Validate the mid-key. 716 assertEquals(Bytes.toStringBinary(blockKeys.get((blockKeys.size() - 1) / 2)), 717 reader.midKey()); 718 719 assertEquals(UNCOMPRESSED_INDEX_SIZES[testI], 720 reader.getTrailer().getUncompressedDataIndexSize()); 721 722 reader.close(); 723 reader2.close(); 724 } 725 } 726 727 private void checkSeekTo(byte[][] keys, HFileScanner scanner, int i) throws IOException { 728 assertEquals("Failed to seek to key #" + i + " (" + Bytes.toStringBinary(keys[i]) + ")", 0, 729 scanner.seekTo(KeyValueUtil.createKeyValueFromKey(keys[i]))); 730 } 731 732 private void assertArrayEqualsBuffer(String msgPrefix, byte[] arr, ByteBuffer buf) { 733 assertEquals( 734 msgPrefix + ": expected " + Bytes.toStringBinary(arr) + ", actual " 735 + Bytes.toStringBinary(buf), 736 0, Bytes.compareTo(arr, 0, arr.length, buf.array(), buf.arrayOffset(), buf.limit())); 737 } 738 739 /** Check a key/value pair after it was read by the reader */ 740 private void checkKeyValue(String msgPrefix, byte[] expectedKey, byte[] expectedValue, 741 ByteBuffer keyRead, ByteBuffer valueRead) { 742 if (!msgPrefix.isEmpty()) msgPrefix += ". "; 743 744 assertArrayEqualsBuffer(msgPrefix + "Invalid key", expectedKey, keyRead); 745 assertArrayEqualsBuffer(msgPrefix + "Invalid value", expectedValue, valueRead); 746 } 747 748 @Test 749 public void testIntermediateLevelIndicesWithLargeKeys() throws IOException { 750 testIntermediateLevelIndicesWithLargeKeys(16); 751 } 752 753 @Test 754 public void testIntermediateLevelIndicesWithLargeKeysWithMinNumEntries() throws IOException { 755 // because of the large rowKeys, we will end up with a 50-level block index without sanity check 756 testIntermediateLevelIndicesWithLargeKeys(2); 757 } 758 759 public void testIntermediateLevelIndicesWithLargeKeys(int minNumEntries) throws IOException { 760 Path hfPath = 761 new Path(TEST_UTIL.getDataTestDir(), "testIntermediateLevelIndicesWithLargeKeys.hfile"); 762 int maxChunkSize = 1024; 763 FileSystem fs = FileSystem.get(conf); 764 CacheConfig cacheConf = new CacheConfig(conf); 765 conf.setInt(HFileBlockIndex.MAX_CHUNK_SIZE_KEY, maxChunkSize); 766 conf.setInt(HFileBlockIndex.MIN_INDEX_NUM_ENTRIES_KEY, minNumEntries); 767 HFileContext context = new HFileContextBuilder().withBlockSize(16).build(); 768 HFile.Writer hfw = new HFile.WriterFactory(conf, cacheConf).withFileContext(context) 769 .withPath(fs, hfPath).create(); 770 List<byte[]> keys = new ArrayList<>(); 771 772 // This should result in leaf-level indices and a root level index 773 for (int i = 0; i < 100; i++) { 774 byte[] rowkey = new byte[maxChunkSize + 1]; 775 byte[] b = Bytes.toBytes(i); 776 System.arraycopy(b, 0, rowkey, rowkey.length - b.length, b.length); 777 keys.add(rowkey); 778 hfw.append(ExtendedCellBuilderFactory.create(CellBuilderType.DEEP_COPY).setRow(rowkey) 779 .setFamily(HConstants.EMPTY_BYTE_ARRAY).setQualifier(HConstants.EMPTY_BYTE_ARRAY) 780 .setTimestamp(HConstants.LATEST_TIMESTAMP).setType(KeyValue.Type.Maximum.getCode()) 781 .setValue(HConstants.EMPTY_BYTE_ARRAY).build()); 782 } 783 hfw.close(); 784 785 HFile.Reader reader = HFile.createReader(fs, hfPath, cacheConf, true, conf); 786 // Scanner doesn't do Cells yet. Fix. 787 HFileScanner scanner = reader.getScanner(conf, true, true); 788 for (int i = 0; i < keys.size(); ++i) { 789 scanner.seekTo(ExtendedCellBuilderFactory.create(CellBuilderType.DEEP_COPY) 790 .setRow(keys.get(i)).setFamily(HConstants.EMPTY_BYTE_ARRAY) 791 .setQualifier(HConstants.EMPTY_BYTE_ARRAY).setTimestamp(HConstants.LATEST_TIMESTAMP) 792 .setType(KeyValue.Type.Maximum.getCode()).setValue(HConstants.EMPTY_BYTE_ARRAY).build()); 793 } 794 reader.close(); 795 } 796 797 /** 798 * This test is for HBASE-27940, which midkey metadata in root index block would always be ignored 799 * by {@link BlockIndexReader#readMultiLevelIndexRoot}. 800 */ 801 @Test 802 public void testMidKeyReadSuccessfullyFromRootIndexBlock() throws IOException { 803 conf.setInt(HFileBlockIndex.MAX_CHUNK_SIZE_KEY, 128); 804 Path hfilePath = 805 new Path(TEST_UTIL.getDataTestDir(), "testMidKeyReadSuccessfullyFromRootIndexBlock"); 806 Compression.Algorithm compressAlgo = Compression.Algorithm.NONE; 807 int entryCount = 50000; 808 HFileContext context = new HFileContextBuilder().withBlockSize(4096).withIncludesTags(false) 809 .withDataBlockEncoding(DataBlockEncoding.NONE).withCompression(compressAlgo).build(); 810 811 try (HFile.Writer writer = new HFile.WriterFactory(conf, new CacheConfig(conf)) 812 .withPath(fs, hfilePath).withFileContext(context).create()) { 813 814 List<KeyValue> keyValues = new ArrayList<>(entryCount); 815 for (int i = 0; i < entryCount; ++i) { 816 byte[] keyBytes = RandomKeyValueUtil.randomOrderedKey(RNG, i); 817 // A random-length random value. 818 byte[] valueBytes = RandomKeyValueUtil.randomValue(RNG); 819 KeyValue keyValue = 820 new KeyValue(keyBytes, null, null, HConstants.LATEST_TIMESTAMP, valueBytes); 821 writer.append(keyValue); 822 keyValues.add(keyValue); 823 } 824 } 825 826 try (FSDataInputStream fsdis = fs.open(hfilePath)) { 827 long fileSize = fs.getFileStatus(hfilePath).getLen(); 828 FixedFileTrailer trailer = FixedFileTrailer.readFromStream(fsdis, fileSize); 829 830 assertEquals(3, trailer.getMajorVersion()); 831 assertEquals(entryCount, trailer.getEntryCount()); 832 HFileContext meta = new HFileContextBuilder().withCompression(compressAlgo) 833 .withIncludesMvcc(false).withIncludesTags(false) 834 .withDataBlockEncoding(DataBlockEncoding.NONE).withHBaseCheckSum(true).build(); 835 ReaderContext readerContext = 836 new ReaderContextBuilder().withInputStreamWrapper(new FSDataInputStreamWrapper(fsdis)) 837 .withFilePath(hfilePath).withFileSystem(fs).withFileSize(fileSize).build(); 838 HFileBlock.FSReader blockReader = 839 new HFileBlock.FSReaderImpl(readerContext, meta, ByteBuffAllocator.HEAP, conf); 840 841 MyEncoder encoder = new MyEncoder(); 842 HFileBlockIndex.CellBasedKeyBlockIndexReaderV2 dataBlockIndexReader = 843 new HFileBlockIndex.CellBasedKeyBlockIndexReaderV2(trailer.createComparator(), 844 trailer.getNumDataIndexLevels(), encoder); 845 846 HFileBlock.BlockIterator blockIter = blockReader.blockRange(trailer.getLoadOnOpenDataOffset(), 847 fileSize - trailer.getTrailerSize()); 848 // Data index. We also read statistics about the block index written after 849 // the root level. 850 dataBlockIndexReader.readMultiLevelIndexRoot( 851 blockIter.nextBlockWithBlockType(BlockType.ROOT_INDEX), trailer.getDataIndexCount()); 852 NoOpEncodedSeeker noOpEncodedSeeker = (NoOpEncodedSeeker) encoder.encoderSeeker; 853 // Assert we have read midkey metadata successfully. 854 assertTrue(noOpEncodedSeeker.midLeafBlockOffset >= 0); 855 assertTrue(noOpEncodedSeeker.midLeafBlockOnDiskSize > 0); 856 assertTrue(noOpEncodedSeeker.midKeyEntry >= 0); 857 } 858 } 859 860 static class MyEncoder implements HFileIndexBlockEncoder { 861 862 EncodedSeeker encoderSeeker; 863 864 @Override 865 public void saveMetadata(Writer writer) throws IOException { 866 NoOpIndexBlockEncoder.INSTANCE.saveMetadata(writer); 867 868 } 869 870 @Override 871 public void encode(BlockIndexChunk blockIndexChunk, boolean rootIndexBlock, DataOutput out) 872 throws IOException { 873 NoOpIndexBlockEncoder.INSTANCE.encode(blockIndexChunk, rootIndexBlock, out); 874 } 875 876 @Override 877 public IndexBlockEncoding getIndexBlockEncoding() { 878 return NoOpIndexBlockEncoder.INSTANCE.getIndexBlockEncoding(); 879 } 880 881 @Override 882 public EncodedSeeker createSeeker() { 883 encoderSeeker = NoOpIndexBlockEncoder.INSTANCE.createSeeker(); 884 return encoderSeeker; 885 } 886 887 } 888}