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.regionserver; 019 020import java.io.IOException; 021import java.io.InterruptedIOException; 022import java.util.ArrayList; 023import java.util.List; 024import java.util.NavigableSet; 025import java.util.Optional; 026import java.util.concurrent.CountDownLatch; 027import java.util.concurrent.locks.ReentrantLock; 028import org.apache.hadoop.hbase.Cell; 029import org.apache.hadoop.hbase.CellComparator; 030import org.apache.hadoop.hbase.CellUtil; 031import org.apache.hadoop.hbase.DoNotRetryIOException; 032import org.apache.hadoop.hbase.HConstants; 033import org.apache.hadoop.hbase.KeyValue; 034import org.apache.hadoop.hbase.KeyValueUtil; 035import org.apache.hadoop.hbase.PrivateCellUtil; 036import org.apache.hadoop.hbase.client.IsolationLevel; 037import org.apache.hadoop.hbase.client.Scan; 038import org.apache.hadoop.hbase.executor.ExecutorService; 039import org.apache.hadoop.hbase.filter.Filter; 040import org.apache.hadoop.hbase.ipc.RpcCall; 041import org.apache.hadoop.hbase.ipc.RpcServer; 042import org.apache.hadoop.hbase.regionserver.ScannerContext.LimitScope; 043import org.apache.hadoop.hbase.regionserver.ScannerContext.NextState; 044import org.apache.hadoop.hbase.regionserver.handler.ParallelSeekHandler; 045import org.apache.hadoop.hbase.regionserver.querymatcher.CompactionScanQueryMatcher; 046import org.apache.hadoop.hbase.regionserver.querymatcher.ScanQueryMatcher; 047import org.apache.hadoop.hbase.regionserver.querymatcher.UserScanQueryMatcher; 048import org.apache.hadoop.hbase.util.EnvironmentEdgeManager; 049import org.apache.yetus.audience.InterfaceAudience; 050import org.slf4j.Logger; 051import org.slf4j.LoggerFactory; 052 053import org.apache.hbase.thirdparty.com.google.common.base.Preconditions; 054import org.apache.hbase.thirdparty.org.apache.commons.collections4.CollectionUtils; 055 056/** 057 * Scanner scans both the memstore and the Store. Coalesce KeyValue stream into List<KeyValue> 058 * for a single row. 059 * <p> 060 * The implementation is not thread safe. So there will be no race between next and close. The only 061 * exception is updateReaders, it will be called in the memstore flush thread to indicate that there 062 * is a flush. 063 */ 064@InterfaceAudience.Private 065public class StoreScanner extends NonReversedNonLazyKeyValueScanner 066 implements KeyValueScanner, InternalScanner, ChangedReadersObserver { 067 private static final Logger LOG = LoggerFactory.getLogger(StoreScanner.class); 068 // In unit tests, the store could be null 069 protected final HStore store; 070 private final CellComparator comparator; 071 private ScanQueryMatcher matcher; 072 protected KeyValueHeap heap; 073 private boolean cacheBlocks; 074 075 private long countPerRow = 0; 076 private int storeLimit = -1; 077 private int storeOffset = 0; 078 079 // Used to indicate that the scanner has closed (see HBASE-1107) 080 private volatile boolean closing = false; 081 private final boolean get; 082 private final boolean explicitColumnQuery; 083 private final boolean useRowColBloom; 084 /** 085 * A flag that enables StoreFileScanner parallel-seeking 086 */ 087 private boolean parallelSeekEnabled = false; 088 private ExecutorService executor; 089 private final Scan scan; 090 private final long oldestUnexpiredTS; 091 private final long now; 092 private final int minVersions; 093 private final long maxRowSize; 094 private final long cellsPerHeartbeatCheck; 095 long memstoreOnlyReads; 096 long mixedReads; 097 098 // 1) Collects all the KVHeap that are eagerly getting closed during the 099 // course of a scan 100 // 2) Collects the unused memstore scanners. If we close the memstore scanners 101 // before sending data to client, the chunk may be reclaimed by other 102 // updates and the data will be corrupt. 103 private final List<KeyValueScanner> scannersForDelayedClose = new ArrayList<>(); 104 105 /** 106 * The number of KVs seen by the scanner. Includes explicitly skipped KVs, but not KVs skipped via 107 * seeking to next row/column. TODO: estimate them? 108 */ 109 private long kvsScanned = 0; 110 private Cell prevCell = null; 111 112 private final long preadMaxBytes; 113 private long bytesRead; 114 115 /** We don't ever expect to change this, the constant is just for clarity. */ 116 static final boolean LAZY_SEEK_ENABLED_BY_DEFAULT = true; 117 public static final String STORESCANNER_PARALLEL_SEEK_ENABLE = 118 "hbase.storescanner.parallel.seek.enable"; 119 120 /** Used during unit testing to ensure that lazy seek does save seek ops */ 121 private static boolean lazySeekEnabledGlobally = LAZY_SEEK_ENABLED_BY_DEFAULT; 122 123 /** 124 * The number of cells scanned in between timeout checks. Specifying a larger value means that 125 * timeout checks will occur less frequently. Specifying a small value will lead to more frequent 126 * timeout checks. 127 */ 128 public static final String HBASE_CELLS_SCANNED_PER_HEARTBEAT_CHECK = 129 "hbase.cells.scanned.per.heartbeat.check"; 130 131 /** 132 * Default value of {@link #HBASE_CELLS_SCANNED_PER_HEARTBEAT_CHECK}. 133 */ 134 public static final long DEFAULT_HBASE_CELLS_SCANNED_PER_HEARTBEAT_CHECK = 10000; 135 136 /** 137 * If the read type is Scan.ReadType.DEFAULT, we will start with pread, and if the kvs we scanned 138 * reaches this limit, we will reopen the scanner with stream. The default value is 4 times of 139 * block size for this store. If configured with a value <0, for all scans with ReadType DEFAULT, 140 * we will open scanner with stream mode itself. 141 */ 142 public static final String STORESCANNER_PREAD_MAX_BYTES = "hbase.storescanner.pread.max.bytes"; 143 144 private final Scan.ReadType readType; 145 146 // A flag whether use pread for scan 147 // it maybe changed if we use Scan.ReadType.DEFAULT and we have read lots of data. 148 private boolean scanUsePread; 149 // Indicates whether there was flush during the course of the scan 150 private volatile boolean flushed = false; 151 // generally we get one file from a flush 152 private final List<KeyValueScanner> flushedstoreFileScanners = new ArrayList<>(1); 153 // Since CompactingMemstore is now default, we get three memstore scanners from a flush 154 private final List<KeyValueScanner> memStoreScannersAfterFlush = new ArrayList<>(3); 155 // The current list of scanners 156 final List<KeyValueScanner> currentScanners = new ArrayList<>(); 157 // flush update lock 158 private final ReentrantLock flushLock = new ReentrantLock(); 159 // lock for closing. 160 private final ReentrantLock closeLock = new ReentrantLock(); 161 162 protected final long readPt; 163 private boolean topChanged = false; 164 165 /** An internal constructor. */ 166 private StoreScanner(HStore store, Scan scan, ScanInfo scanInfo, int numColumns, long readPt, 167 boolean cacheBlocks, ScanType scanType) { 168 this.readPt = readPt; 169 this.store = store; 170 this.cacheBlocks = cacheBlocks; 171 this.comparator = Preconditions.checkNotNull(scanInfo.getComparator()); 172 get = scan.isGetScan(); 173 explicitColumnQuery = numColumns > 0; 174 this.scan = scan; 175 this.now = EnvironmentEdgeManager.currentTime(); 176 this.oldestUnexpiredTS = scan.isRaw() ? 0L : now - scanInfo.getTtl(); 177 this.minVersions = scanInfo.getMinVersions(); 178 179 // We look up row-column Bloom filters for multi-column queries as part of 180 // the seek operation. However, we also look the row-column Bloom filter 181 // for multi-row (non-"get") scans because this is not done in 182 // StoreFile.passesBloomFilter(Scan, SortedSet<byte[]>). 183 this.useRowColBloom = numColumns > 1 || (!get && numColumns == 1) && (store == null 184 || store.getColumnFamilyDescriptor().getBloomFilterType() == BloomType.ROWCOL); 185 this.maxRowSize = scanInfo.getTableMaxRowSize(); 186 this.preadMaxBytes = scanInfo.getPreadMaxBytes(); 187 if (get) { 188 this.readType = Scan.ReadType.PREAD; 189 this.scanUsePread = true; 190 } else if (scanType != ScanType.USER_SCAN) { 191 // For compaction scanners never use Pread as already we have stream based scanners on the 192 // store files to be compacted 193 this.readType = Scan.ReadType.STREAM; 194 this.scanUsePread = false; 195 } else { 196 if (scan.getReadType() == Scan.ReadType.DEFAULT) { 197 if (scanInfo.isUsePread()) { 198 this.readType = Scan.ReadType.PREAD; 199 } else if (this.preadMaxBytes < 0) { 200 this.readType = Scan.ReadType.STREAM; 201 } else { 202 this.readType = Scan.ReadType.DEFAULT; 203 } 204 } else { 205 this.readType = scan.getReadType(); 206 } 207 // Always start with pread unless user specific stream. Will change to stream later if 208 // readType is default if the scan keeps running for a long time. 209 this.scanUsePread = this.readType != Scan.ReadType.STREAM; 210 } 211 this.cellsPerHeartbeatCheck = scanInfo.getCellsPerTimeoutCheck(); 212 // Parallel seeking is on if the config allows and more there is more than one store file. 213 if (store != null && store.getStorefilesCount() > 1) { 214 RegionServerServices rsService = store.getHRegion().getRegionServerServices(); 215 if (rsService != null && scanInfo.isParallelSeekEnabled()) { 216 this.parallelSeekEnabled = true; 217 this.executor = rsService.getExecutorService(); 218 } 219 } 220 } 221 222 private void addCurrentScanners(List<? extends KeyValueScanner> scanners) { 223 this.currentScanners.addAll(scanners); 224 } 225 226 /** 227 * Opens a scanner across memstore, snapshot, and all StoreFiles. Assumes we are not in a 228 * compaction. 229 * @param store who we scan 230 * @param scan the spec 231 * @param columns which columns we are scanning 232 */ 233 public StoreScanner(HStore store, ScanInfo scanInfo, Scan scan, NavigableSet<byte[]> columns, 234 long readPt) throws IOException { 235 this(store, scan, scanInfo, columns != null ? columns.size() : 0, readPt, scan.getCacheBlocks(), 236 ScanType.USER_SCAN); 237 if (columns != null && scan.isRaw()) { 238 throw new DoNotRetryIOException("Cannot specify any column for a raw scan"); 239 } 240 matcher = UserScanQueryMatcher.create(scan, scanInfo, columns, oldestUnexpiredTS, now, 241 store.getCoprocessorHost()); 242 243 store.addChangedReaderObserver(this); 244 245 List<KeyValueScanner> scanners = null; 246 try { 247 // Pass columns to try to filter out unnecessary StoreFiles. 248 scanners = selectScannersFrom(store, 249 store.getScanners(cacheBlocks, scanUsePread, false, matcher, scan.getStartRow(), 250 scan.includeStartRow(), scan.getStopRow(), scan.includeStopRow(), this.readPt)); 251 252 // Seek all scanners to the start of the Row (or if the exact matching row 253 // key does not exist, then to the start of the next matching Row). 254 // Always check bloom filter to optimize the top row seek for delete 255 // family marker. 256 seekScanners(scanners, matcher.getStartKey(), explicitColumnQuery && lazySeekEnabledGlobally, 257 parallelSeekEnabled); 258 259 // set storeLimit 260 this.storeLimit = scan.getMaxResultsPerColumnFamily(); 261 262 // set rowOffset 263 this.storeOffset = scan.getRowOffsetPerColumnFamily(); 264 addCurrentScanners(scanners); 265 // Combine all seeked scanners with a heap 266 resetKVHeap(scanners, comparator); 267 } catch (IOException e) { 268 clearAndClose(scanners); 269 // remove us from the HStore#changedReaderObservers here or we'll have no chance to 270 // and might cause memory leak 271 store.deleteChangedReaderObserver(this); 272 throw e; 273 } 274 } 275 276 // a dummy scan instance for compaction. 277 private static final Scan SCAN_FOR_COMPACTION = new Scan(); 278 279 /** 280 * Used for store file compaction and memstore compaction. 281 * <p> 282 * Opens a scanner across specified StoreFiles/MemStoreSegments. 283 * @param store who we scan 284 * @param scanners ancillary scanners 285 * @param smallestReadPoint the readPoint that we should use for tracking versions 286 */ 287 public StoreScanner(HStore store, ScanInfo scanInfo, List<? extends KeyValueScanner> scanners, 288 ScanType scanType, long smallestReadPoint, long earliestPutTs) throws IOException { 289 this(store, scanInfo, scanners, scanType, smallestReadPoint, earliestPutTs, null, null); 290 } 291 292 /** 293 * Used for compactions that drop deletes from a limited range of rows. 294 * <p> 295 * Opens a scanner across specified StoreFiles. 296 * @param store who we scan 297 * @param scanners ancillary scanners 298 * @param smallestReadPoint the readPoint that we should use for tracking versions 299 * @param dropDeletesFromRow The inclusive left bound of the range; can be EMPTY_START_ROW. 300 * @param dropDeletesToRow The exclusive right bound of the range; can be EMPTY_END_ROW. 301 */ 302 public StoreScanner(HStore store, ScanInfo scanInfo, List<? extends KeyValueScanner> scanners, 303 long smallestReadPoint, long earliestPutTs, byte[] dropDeletesFromRow, byte[] dropDeletesToRow) 304 throws IOException { 305 this(store, scanInfo, scanners, ScanType.COMPACT_RETAIN_DELETES, smallestReadPoint, 306 earliestPutTs, dropDeletesFromRow, dropDeletesToRow); 307 } 308 309 private StoreScanner(HStore store, ScanInfo scanInfo, List<? extends KeyValueScanner> scanners, 310 ScanType scanType, long smallestReadPoint, long earliestPutTs, byte[] dropDeletesFromRow, 311 byte[] dropDeletesToRow) throws IOException { 312 this(store, SCAN_FOR_COMPACTION, scanInfo, 0, 313 store.getHRegion().getReadPoint(IsolationLevel.READ_COMMITTED), false, scanType); 314 assert scanType != ScanType.USER_SCAN; 315 matcher = 316 CompactionScanQueryMatcher.create(scanInfo, scanType, smallestReadPoint, earliestPutTs, 317 oldestUnexpiredTS, now, dropDeletesFromRow, dropDeletesToRow, store.getCoprocessorHost()); 318 319 // Filter the list of scanners using Bloom filters, time range, TTL, etc. 320 scanners = selectScannersFrom(store, scanners); 321 322 // Seek all scanners to the initial key 323 seekScanners(scanners, matcher.getStartKey(), false, parallelSeekEnabled); 324 addCurrentScanners(scanners); 325 // Combine all seeked scanners with a heap 326 resetKVHeap(scanners, comparator); 327 } 328 329 private void seekAllScanner(ScanInfo scanInfo, List<? extends KeyValueScanner> scanners) 330 throws IOException { 331 // Seek all scanners to the initial key 332 seekScanners(scanners, matcher.getStartKey(), false, parallelSeekEnabled); 333 addCurrentScanners(scanners); 334 resetKVHeap(scanners, comparator); 335 } 336 337 // For mob compaction only as we do not have a Store instance when doing mob compaction. 338 public StoreScanner(ScanInfo scanInfo, ScanType scanType, 339 List<? extends KeyValueScanner> scanners) throws IOException { 340 this(null, SCAN_FOR_COMPACTION, scanInfo, 0, Long.MAX_VALUE, false, scanType); 341 assert scanType != ScanType.USER_SCAN; 342 this.matcher = CompactionScanQueryMatcher.create(scanInfo, scanType, Long.MAX_VALUE, 0L, 343 oldestUnexpiredTS, now, null, null, null); 344 seekAllScanner(scanInfo, scanners); 345 } 346 347 // Used to instantiate a scanner for user scan in test 348 StoreScanner(Scan scan, ScanInfo scanInfo, NavigableSet<byte[]> columns, 349 List<? extends KeyValueScanner> scanners) throws IOException { 350 // 0 is passed as readpoint because the test bypasses Store 351 this(null, scan, scanInfo, columns != null ? columns.size() : 0, 0L, scan.getCacheBlocks(), 352 ScanType.USER_SCAN); 353 this.matcher = 354 UserScanQueryMatcher.create(scan, scanInfo, columns, oldestUnexpiredTS, now, null); 355 seekAllScanner(scanInfo, scanners); 356 } 357 358 // Used to instantiate a scanner for user scan in test 359 StoreScanner(Scan scan, ScanInfo scanInfo, NavigableSet<byte[]> columns, 360 List<? extends KeyValueScanner> scanners, ScanType scanType) throws IOException { 361 // 0 is passed as readpoint because the test bypasses Store 362 this(null, scan, scanInfo, columns != null ? columns.size() : 0, 0L, scan.getCacheBlocks(), 363 scanType); 364 if (scanType == ScanType.USER_SCAN) { 365 this.matcher = 366 UserScanQueryMatcher.create(scan, scanInfo, columns, oldestUnexpiredTS, now, null); 367 } else { 368 this.matcher = CompactionScanQueryMatcher.create(scanInfo, scanType, Long.MAX_VALUE, 369 HConstants.OLDEST_TIMESTAMP, oldestUnexpiredTS, now, null, null, null); 370 } 371 seekAllScanner(scanInfo, scanners); 372 } 373 374 // Used to instantiate a scanner for compaction in test 375 StoreScanner(ScanInfo scanInfo, int maxVersions, ScanType scanType, 376 List<? extends KeyValueScanner> scanners) throws IOException { 377 // 0 is passed as readpoint because the test bypasses Store 378 this(null, maxVersions > 0 ? new Scan().readVersions(maxVersions) : SCAN_FOR_COMPACTION, 379 scanInfo, 0, 0L, false, scanType); 380 this.matcher = CompactionScanQueryMatcher.create(scanInfo, scanType, Long.MAX_VALUE, 381 HConstants.OLDEST_TIMESTAMP, oldestUnexpiredTS, now, null, null, null); 382 seekAllScanner(scanInfo, scanners); 383 } 384 385 boolean isScanUsePread() { 386 return this.scanUsePread; 387 } 388 389 /** 390 * Seek the specified scanners with the given key 391 * @param isLazy true if using lazy seek 392 * @param isParallelSeek true if using parallel seek 393 */ 394 protected void seekScanners(List<? extends KeyValueScanner> scanners, Cell seekKey, 395 boolean isLazy, boolean isParallelSeek) throws IOException { 396 // Seek all scanners to the start of the Row (or if the exact matching row 397 // key does not exist, then to the start of the next matching Row). 398 // Always check bloom filter to optimize the top row seek for delete 399 // family marker. 400 if (isLazy) { 401 for (KeyValueScanner scanner : scanners) { 402 scanner.requestSeek(seekKey, false, true); 403 } 404 } else { 405 if (!isParallelSeek) { 406 long totalScannersSoughtBytes = 0; 407 for (KeyValueScanner scanner : scanners) { 408 if (matcher.isUserScan() && totalScannersSoughtBytes >= maxRowSize) { 409 throw new RowTooBigException( 410 "Max row size allowed: " + maxRowSize + ", but row is bigger than that"); 411 } 412 scanner.seek(seekKey); 413 Cell c = scanner.peek(); 414 if (c != null) { 415 totalScannersSoughtBytes += PrivateCellUtil.estimatedSerializedSizeOf(c); 416 } 417 } 418 } else { 419 parallelSeek(scanners, seekKey); 420 } 421 } 422 } 423 424 protected void resetKVHeap(List<? extends KeyValueScanner> scanners, CellComparator comparator) 425 throws IOException { 426 // Combine all seeked scanners with a heap 427 heap = newKVHeap(scanners, comparator); 428 } 429 430 protected KeyValueHeap newKVHeap(List<? extends KeyValueScanner> scanners, 431 CellComparator comparator) throws IOException { 432 return new KeyValueHeap(scanners, comparator); 433 } 434 435 /** 436 * Filters the given list of scanners using Bloom filter, time range, and TTL. 437 * <p> 438 * Will be overridden by testcase so declared as protected. 439 */ 440 protected List<KeyValueScanner> selectScannersFrom(HStore store, 441 List<? extends KeyValueScanner> allScanners) { 442 boolean memOnly; 443 boolean filesOnly; 444 if (scan instanceof InternalScan) { 445 InternalScan iscan = (InternalScan) scan; 446 memOnly = iscan.isCheckOnlyMemStore(); 447 filesOnly = iscan.isCheckOnlyStoreFiles(); 448 } else { 449 memOnly = false; 450 filesOnly = false; 451 } 452 453 List<KeyValueScanner> scanners = new ArrayList<>(allScanners.size()); 454 455 // We can only exclude store files based on TTL if minVersions is set to 0. 456 // Otherwise, we might have to return KVs that have technically expired. 457 long expiredTimestampCutoff = minVersions == 0 ? oldestUnexpiredTS : Long.MIN_VALUE; 458 459 // include only those scan files which pass all filters 460 for (KeyValueScanner kvs : allScanners) { 461 boolean isFile = kvs.isFileScanner(); 462 if ((!isFile && filesOnly) || (isFile && memOnly)) { 463 kvs.close(); 464 continue; 465 } 466 467 if (kvs.shouldUseScanner(scan, store, expiredTimestampCutoff)) { 468 scanners.add(kvs); 469 } else { 470 kvs.close(); 471 } 472 } 473 return scanners; 474 } 475 476 @Override 477 public Cell peek() { 478 return heap != null ? heap.peek() : null; 479 } 480 481 @Override 482 public KeyValue next() { 483 // throw runtime exception perhaps? 484 throw new RuntimeException("Never call StoreScanner.next()"); 485 } 486 487 @Override 488 public void close() { 489 close(true); 490 } 491 492 private void close(boolean withDelayedScannersClose) { 493 closeLock.lock(); 494 // If the closeLock is acquired then any subsequent updateReaders() 495 // call is ignored. 496 try { 497 if (this.closing) { 498 return; 499 } 500 if (withDelayedScannersClose) { 501 this.closing = true; 502 } 503 // For mob compaction, we do not have a store. 504 if (this.store != null) { 505 this.store.deleteChangedReaderObserver(this); 506 } 507 if (withDelayedScannersClose) { 508 clearAndClose(scannersForDelayedClose); 509 clearAndClose(memStoreScannersAfterFlush); 510 clearAndClose(flushedstoreFileScanners); 511 if (this.heap != null) { 512 this.heap.close(); 513 this.currentScanners.clear(); 514 this.heap = null; // CLOSED! 515 } 516 } else { 517 if (this.heap != null) { 518 this.scannersForDelayedClose.add(this.heap); 519 this.currentScanners.clear(); 520 this.heap = null; 521 } 522 } 523 } finally { 524 closeLock.unlock(); 525 } 526 } 527 528 @Override 529 public boolean seek(Cell key) throws IOException { 530 if (checkFlushed()) { 531 reopenAfterFlush(); 532 } 533 return this.heap.seek(key); 534 } 535 536 /** 537 * Get the next row of values from this Store. 538 * @return true if there are more rows, false if scanner is done 539 */ 540 @Override 541 public boolean next(List<Cell> outResult, ScannerContext scannerContext) throws IOException { 542 if (scannerContext == null) { 543 throw new IllegalArgumentException("Scanner context cannot be null"); 544 } 545 if (checkFlushed() && reopenAfterFlush()) { 546 return scannerContext.setScannerState(NextState.MORE_VALUES).hasMoreValues(); 547 } 548 549 // if the heap was left null, then the scanners had previously run out anyways, close and 550 // return. 551 if (this.heap == null) { 552 // By this time partial close should happened because already heap is null 553 close(false);// Do all cleanup except heap.close() 554 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 555 } 556 557 Cell cell = this.heap.peek(); 558 if (cell == null) { 559 close(false);// Do all cleanup except heap.close() 560 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 561 } 562 563 // only call setRow if the row changes; avoids confusing the query matcher 564 // if scanning intra-row 565 566 // If no limits exists in the scope LimitScope.Between_Cells then we are sure we are changing 567 // rows. Else it is possible we are still traversing the same row so we must perform the row 568 // comparison. 569 if (!scannerContext.hasAnyLimit(LimitScope.BETWEEN_CELLS) || matcher.currentRow() == null) { 570 this.countPerRow = 0; 571 matcher.setToNewRow(cell); 572 } 573 574 // Clear progress away unless invoker has indicated it should be kept. 575 if (!scannerContext.getKeepProgress() && !scannerContext.getSkippingRow()) { 576 scannerContext.clearProgress(); 577 } 578 579 Optional<RpcCall> rpcCall = 580 matcher.isUserScan() ? RpcServer.getCurrentCall() : Optional.empty(); 581 582 int count = 0; 583 long totalBytesRead = 0; 584 boolean onlyFromMemstore = matcher.isUserScan(); 585 try { 586 LOOP: do { 587 // Update and check the time limit based on the configured value of cellsPerTimeoutCheck 588 // Or if the preadMaxBytes is reached and we may want to return so we can switch to stream 589 // in 590 // the shipped method below. 591 if ( 592 kvsScanned % cellsPerHeartbeatCheck == 0 593 || (scanUsePread && readType == Scan.ReadType.DEFAULT && bytesRead > preadMaxBytes) 594 ) { 595 if (scannerContext.checkTimeLimit(LimitScope.BETWEEN_CELLS)) { 596 return scannerContext.setScannerState(NextState.TIME_LIMIT_REACHED).hasMoreValues(); 597 } 598 } 599 // Do object compare - we set prevKV from the same heap. 600 if (prevCell != cell) { 601 ++kvsScanned; 602 } 603 checkScanOrder(prevCell, cell, comparator); 604 int cellSize = PrivateCellUtil.estimatedSerializedSizeOf(cell); 605 bytesRead += cellSize; 606 if (scanUsePread && readType == Scan.ReadType.DEFAULT && bytesRead > preadMaxBytes) { 607 // return immediately if we want to switch from pread to stream. We need this because we 608 // can 609 // only switch in the shipped method, if user use a filter to filter out everything and 610 // rpc 611 // timeout is very large then the shipped method will never be called until the whole scan 612 // is finished, but at that time we have already scan all the data... 613 // See HBASE-20457 for more details. 614 // And there is still a scenario that can not be handled. If we have a very large row, 615 // which 616 // have millions of qualifiers, and filter.filterRow is used, then even if we set the flag 617 // here, we still need to scan all the qualifiers before returning... 618 scannerContext.returnImmediately(); 619 } 620 621 heap.recordBlockSize(blockSize -> { 622 if (rpcCall.isPresent()) { 623 rpcCall.get().incrementBlockBytesScanned(blockSize); 624 } 625 scannerContext.incrementBlockProgress(blockSize); 626 }); 627 628 prevCell = cell; 629 scannerContext.setLastPeekedCell(cell); 630 topChanged = false; 631 ScanQueryMatcher.MatchCode qcode = matcher.match(cell); 632 switch (qcode) { 633 case INCLUDE: 634 case INCLUDE_AND_SEEK_NEXT_ROW: 635 case INCLUDE_AND_SEEK_NEXT_COL: 636 Filter f = matcher.getFilter(); 637 if (f != null) { 638 cell = f.transformCell(cell); 639 } 640 this.countPerRow++; 641 642 // add to results only if we have skipped #storeOffset kvs 643 // also update metric accordingly 644 if (this.countPerRow > storeOffset) { 645 outResult.add(cell); 646 647 // Update local tracking information 648 count++; 649 totalBytesRead += cellSize; 650 651 /** 652 * Increment the metric if all the cells are from memstore. If not we will account it 653 * for mixed reads 654 */ 655 onlyFromMemstore = onlyFromMemstore && heap.isLatestCellFromMemstore(); 656 // Update the progress of the scanner context 657 scannerContext.incrementSizeProgress(cellSize, cell.heapSize()); 658 scannerContext.incrementBatchProgress(1); 659 660 if (matcher.isUserScan() && totalBytesRead > maxRowSize) { 661 String message = "Max row size allowed: " + maxRowSize 662 + ", but the row is bigger than that, the row info: " 663 + CellUtil.toString(cell, false) + ", already have process row cells = " 664 + outResult.size() + ", it belong to region = " 665 + store.getHRegion().getRegionInfo().getRegionNameAsString(); 666 LOG.warn(message); 667 throw new RowTooBigException(message); 668 } 669 670 if (storeLimit > -1 && this.countPerRow >= (storeLimit + storeOffset)) { 671 // do what SEEK_NEXT_ROW does. 672 if (!matcher.moreRowsMayExistAfter(cell)) { 673 close(false);// Do all cleanup except heap.close() 674 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 675 } 676 matcher.clearCurrentRow(); 677 seekToNextRow(cell); 678 break LOOP; 679 } 680 } 681 682 if (qcode == ScanQueryMatcher.MatchCode.INCLUDE_AND_SEEK_NEXT_ROW) { 683 if (!matcher.moreRowsMayExistAfter(cell)) { 684 close(false);// Do all cleanup except heap.close() 685 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 686 } 687 matcher.clearCurrentRow(); 688 seekOrSkipToNextRow(cell); 689 } else if (qcode == ScanQueryMatcher.MatchCode.INCLUDE_AND_SEEK_NEXT_COL) { 690 seekOrSkipToNextColumn(cell); 691 } else { 692 this.heap.next(); 693 } 694 695 if (scannerContext.checkBatchLimit(LimitScope.BETWEEN_CELLS)) { 696 break LOOP; 697 } 698 if (scannerContext.checkSizeLimit(LimitScope.BETWEEN_CELLS)) { 699 break LOOP; 700 } 701 continue; 702 703 case DONE: 704 // Optimization for Gets! If DONE, no more to get on this row, early exit! 705 if (get) { 706 // Then no more to this row... exit. 707 close(false);// Do all cleanup except heap.close() 708 // update metric 709 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 710 } 711 matcher.clearCurrentRow(); 712 return scannerContext.setScannerState(NextState.MORE_VALUES).hasMoreValues(); 713 714 case DONE_SCAN: 715 close(false);// Do all cleanup except heap.close() 716 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 717 718 case SEEK_NEXT_ROW: 719 // This is just a relatively simple end of scan fix, to short-cut end 720 // us if there is an endKey in the scan. 721 if (!matcher.moreRowsMayExistAfter(cell)) { 722 close(false);// Do all cleanup except heap.close() 723 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 724 } 725 matcher.clearCurrentRow(); 726 seekOrSkipToNextRow(cell); 727 NextState stateAfterSeekNextRow = needToReturn(outResult); 728 if (stateAfterSeekNextRow != null) { 729 return scannerContext.setScannerState(stateAfterSeekNextRow).hasMoreValues(); 730 } 731 break; 732 733 case SEEK_NEXT_COL: 734 seekOrSkipToNextColumn(cell); 735 NextState stateAfterSeekNextColumn = needToReturn(outResult); 736 if (stateAfterSeekNextColumn != null) { 737 return scannerContext.setScannerState(stateAfterSeekNextColumn).hasMoreValues(); 738 } 739 break; 740 741 case SKIP: 742 this.heap.next(); 743 break; 744 745 case SEEK_NEXT_USING_HINT: 746 Cell nextKV = matcher.getNextKeyHint(cell); 747 if (nextKV != null) { 748 int difference = comparator.compare(nextKV, cell); 749 if ( 750 ((!scan.isReversed() && difference > 0) || (scan.isReversed() && difference < 0)) 751 ) { 752 seekAsDirection(nextKV); 753 NextState stateAfterSeekByHint = needToReturn(outResult); 754 if (stateAfterSeekByHint != null) { 755 return scannerContext.setScannerState(stateAfterSeekByHint).hasMoreValues(); 756 } 757 break; 758 } 759 } 760 heap.next(); 761 break; 762 763 default: 764 throw new RuntimeException("UNEXPECTED"); 765 } 766 767 // One last chance to break due to size limit. The INCLUDE* cases above already check 768 // limit and continue. For the various filtered cases, we need to check because block 769 // size limit may have been exceeded even if we don't add cells to result list. 770 if (scannerContext.checkSizeLimit(LimitScope.BETWEEN_CELLS)) { 771 return scannerContext.setScannerState(NextState.MORE_VALUES).hasMoreValues(); 772 } 773 } while ((cell = this.heap.peek()) != null); 774 775 if (count > 0) { 776 return scannerContext.setScannerState(NextState.MORE_VALUES).hasMoreValues(); 777 } 778 779 // No more keys 780 close(false);// Do all cleanup except heap.close() 781 return scannerContext.setScannerState(NextState.NO_MORE_VALUES).hasMoreValues(); 782 } finally { 783 // increment only if we have some result 784 if (count > 0 && matcher.isUserScan()) { 785 // if true increment memstore metrics, if not the mixed one 786 updateMetricsStore(onlyFromMemstore); 787 } 788 } 789 } 790 791 private void updateMetricsStore(boolean memstoreRead) { 792 if (store != null) { 793 store.updateMetricsStore(memstoreRead); 794 } else { 795 // for testing. 796 if (memstoreRead) { 797 memstoreOnlyReads++; 798 } else { 799 mixedReads++; 800 } 801 } 802 } 803 804 /** 805 * If the top cell won't be flushed into disk, the new top cell may be changed after 806 * #reopenAfterFlush. Because the older top cell only exist in the memstore scanner but the 807 * memstore scanner is replaced by hfile scanner after #reopenAfterFlush. If the row of top cell 808 * is changed, we should return the current cells. Otherwise, we may return the cells across 809 * different rows. 810 * @param outResult the cells which are visible for user scan 811 * @return null is the top cell doesn't change. Otherwise, the NextState to return 812 */ 813 private NextState needToReturn(List<Cell> outResult) { 814 if (!outResult.isEmpty() && topChanged) { 815 return heap.peek() == null ? NextState.NO_MORE_VALUES : NextState.MORE_VALUES; 816 } 817 return null; 818 } 819 820 private void seekOrSkipToNextRow(Cell cell) throws IOException { 821 // If it is a Get Scan, then we know that we are done with this row; there are no more 822 // rows beyond the current one: don't try to optimize. 823 if (!get) { 824 if (trySkipToNextRow(cell)) { 825 return; 826 } 827 } 828 seekToNextRow(cell); 829 } 830 831 private void seekOrSkipToNextColumn(Cell cell) throws IOException { 832 if (!trySkipToNextColumn(cell)) { 833 seekAsDirection(matcher.getKeyForNextColumn(cell)); 834 } 835 } 836 837 /** 838 * See if we should actually SEEK or rather just SKIP to the next Cell (see HBASE-13109). 839 * ScanQueryMatcher may issue SEEK hints, such as seek to next column, next row, or seek to an 840 * arbitrary seek key. This method decides whether a seek is the most efficient _actual_ way to 841 * get us to the requested cell (SEEKs are more expensive than SKIP, SKIP, SKIP inside the 842 * current, loaded block). It does this by looking at the next indexed key of the current HFile. 843 * This key is then compared with the _SEEK_ key, where a SEEK key is an artificial 'last possible 844 * key on the row' (only in here, we avoid actually creating a SEEK key; in the compare we work 845 * with the current Cell but compare as though it were a seek key; see down in 846 * matcher.compareKeyForNextRow, etc). If the compare gets us onto the next block we *_SEEK, 847 * otherwise we just SKIP to the next requested cell. 848 * <p> 849 * Other notes: 850 * <ul> 851 * <li>Rows can straddle block boundaries</li> 852 * <li>Versions of columns can straddle block boundaries (i.e. column C1 at T1 might be in a 853 * different block than column C1 at T2)</li> 854 * <li>We want to SKIP if the chance is high that we'll find the desired Cell after a few 855 * SKIPs...</li> 856 * <li>We want to SEEK when the chance is high that we'll be able to seek past many Cells, 857 * especially if we know we need to go to the next block.</li> 858 * </ul> 859 * <p> 860 * A good proxy (best effort) to determine whether SKIP is better than SEEK is whether we'll 861 * likely end up seeking to the next block (or past the next block) to get our next column. 862 * Example: 863 * 864 * <pre> 865 * | BLOCK 1 | BLOCK 2 | 866 * | r1/c1, r1/c2, r1/c3 | r1/c4, r1/c5, r2/c1 | 867 * ^ ^ 868 * | | 869 * Next Index Key SEEK_NEXT_ROW (before r2/c1) 870 * 871 * 872 * | BLOCK 1 | BLOCK 2 | 873 * | r1/c1/t5, r1/c1/t4, r1/c1/t3 | r1/c1/t2, r1/c1/T1, r1/c2/T3 | 874 * ^ ^ 875 * | | 876 * Next Index Key SEEK_NEXT_COL 877 * </pre> 878 * 879 * Now imagine we want columns c1 and c3 (see first diagram above), the 'Next Index Key' of r1/c4 880 * is > r1/c3 so we should seek to get to the c1 on the next row, r2. In second case, say we only 881 * want one version of c1, after we have it, a SEEK_COL will be issued to get to c2. Looking at 882 * the 'Next Index Key', it would land us in the next block, so we should SEEK. In other scenarios 883 * where the SEEK will not land us in the next block, it is very likely better to issues a series 884 * of SKIPs. 885 * @param cell current cell 886 * @return true means skip to next row, false means not 887 */ 888 protected boolean trySkipToNextRow(Cell cell) throws IOException { 889 Cell nextCell = null; 890 // used to guard against a changed next indexed key by doing a identity comparison 891 // when the identity changes we need to compare the bytes again 892 Cell previousIndexedKey = null; 893 do { 894 Cell nextIndexedKey = getNextIndexedKey(); 895 if ( 896 nextIndexedKey != null && nextIndexedKey != KeyValueScanner.NO_NEXT_INDEXED_KEY 897 && (nextIndexedKey == previousIndexedKey 898 || matcher.compareKeyForNextRow(nextIndexedKey, cell) >= 0) 899 ) { 900 this.heap.next(); 901 ++kvsScanned; 902 previousIndexedKey = nextIndexedKey; 903 } else { 904 return false; 905 } 906 } while ((nextCell = this.heap.peek()) != null && CellUtil.matchingRows(cell, nextCell)); 907 return true; 908 } 909 910 /** 911 * See {@link org.apache.hadoop.hbase.regionserver.StoreScanner#trySkipToNextRow(Cell)} 912 * @param cell current cell 913 * @return true means skip to next column, false means not 914 */ 915 protected boolean trySkipToNextColumn(Cell cell) throws IOException { 916 Cell nextCell = null; 917 // used to guard against a changed next indexed key by doing a identity comparison 918 // when the identity changes we need to compare the bytes again 919 Cell previousIndexedKey = null; 920 do { 921 Cell nextIndexedKey = getNextIndexedKey(); 922 if ( 923 nextIndexedKey != null && nextIndexedKey != KeyValueScanner.NO_NEXT_INDEXED_KEY 924 && (nextIndexedKey == previousIndexedKey 925 || matcher.compareKeyForNextColumn(nextIndexedKey, cell) >= 0) 926 ) { 927 this.heap.next(); 928 ++kvsScanned; 929 previousIndexedKey = nextIndexedKey; 930 } else { 931 return false; 932 } 933 } while ((nextCell = this.heap.peek()) != null && CellUtil.matchingRowColumn(cell, nextCell)); 934 // We need this check because it may happen that the new scanner that we get 935 // during heap.next() is requiring reseek due of fake KV previously generated for 936 // ROWCOL bloom filter optimization. See HBASE-19863 for more details 937 if (useRowColBloom && nextCell != null && cell.getTimestamp() == HConstants.OLDEST_TIMESTAMP) { 938 return false; 939 } 940 return true; 941 } 942 943 @Override 944 public long getReadPoint() { 945 return this.readPt; 946 } 947 948 private static void clearAndClose(List<KeyValueScanner> scanners) { 949 if (scanners == null) { 950 return; 951 } 952 for (KeyValueScanner s : scanners) { 953 s.close(); 954 } 955 scanners.clear(); 956 } 957 958 // Implementation of ChangedReadersObserver 959 @Override 960 public void updateReaders(List<HStoreFile> sfs, List<KeyValueScanner> memStoreScanners) 961 throws IOException { 962 if (CollectionUtils.isEmpty(sfs) && CollectionUtils.isEmpty(memStoreScanners)) { 963 return; 964 } 965 boolean updateReaders = false; 966 flushLock.lock(); 967 try { 968 if (!closeLock.tryLock()) { 969 // The reason for doing this is that when the current store scanner does not retrieve 970 // any new cells, then the scanner is considered to be done. The heap of this scanner 971 // is not closed till the shipped() call is completed. Hence in that case if at all 972 // the partial close (close (false)) has been called before updateReaders(), there is no 973 // need for the updateReaders() to happen. 974 LOG.debug("StoreScanner already has the close lock. There is no need to updateReaders"); 975 // no lock acquired. 976 clearAndClose(memStoreScanners); 977 return; 978 } 979 // lock acquired 980 updateReaders = true; 981 if (this.closing) { 982 LOG.debug("StoreScanner already closing. There is no need to updateReaders"); 983 clearAndClose(memStoreScanners); 984 return; 985 } 986 flushed = true; 987 final boolean isCompaction = false; 988 boolean usePread = get || scanUsePread; 989 // SEE HBASE-19468 where the flushed files are getting compacted even before a scanner 990 // calls next(). So its better we create scanners here rather than next() call. Ensure 991 // these scanners are properly closed() whether or not the scan is completed successfully 992 // Eagerly creating scanners so that we have the ref counting ticking on the newly created 993 // store files. In case of stream scanners this eager creation does not induce performance 994 // penalty because in scans (that uses stream scanners) the next() call is bound to happen. 995 List<KeyValueScanner> scanners = store.getScanners(sfs, cacheBlocks, get, usePread, 996 isCompaction, matcher, scan.getStartRow(), scan.getStopRow(), this.readPt, false); 997 flushedstoreFileScanners.addAll(scanners); 998 if (!CollectionUtils.isEmpty(memStoreScanners)) { 999 clearAndClose(memStoreScannersAfterFlush); 1000 memStoreScannersAfterFlush.addAll(memStoreScanners); 1001 } 1002 } finally { 1003 flushLock.unlock(); 1004 if (updateReaders) { 1005 closeLock.unlock(); 1006 } 1007 } 1008 // Let the next() call handle re-creating and seeking 1009 } 1010 1011 /** Returns if top of heap has changed (and KeyValueHeap has to try the next KV) */ 1012 protected final boolean reopenAfterFlush() throws IOException { 1013 // here we can make sure that we have a Store instance so no null check on store. 1014 Cell lastTop = heap.peek(); 1015 // When we have the scan object, should we not pass it to getScanners() to get a limited set of 1016 // scanners? We did so in the constructor and we could have done it now by storing the scan 1017 // object from the constructor 1018 List<KeyValueScanner> scanners; 1019 flushLock.lock(); 1020 try { 1021 List<KeyValueScanner> allScanners = 1022 new ArrayList<>(flushedstoreFileScanners.size() + memStoreScannersAfterFlush.size()); 1023 allScanners.addAll(flushedstoreFileScanners); 1024 allScanners.addAll(memStoreScannersAfterFlush); 1025 scanners = selectScannersFrom(store, allScanners); 1026 // Clear the current set of flushed store files scanners so that they don't get added again 1027 flushedstoreFileScanners.clear(); 1028 memStoreScannersAfterFlush.clear(); 1029 } finally { 1030 flushLock.unlock(); 1031 } 1032 1033 // Seek the new scanners to the last key 1034 seekScanners(scanners, lastTop, false, parallelSeekEnabled); 1035 // remove the older memstore scanner 1036 for (int i = currentScanners.size() - 1; i >= 0; i--) { 1037 if (!currentScanners.get(i).isFileScanner()) { 1038 scannersForDelayedClose.add(currentScanners.remove(i)); 1039 } else { 1040 // we add the memstore scanner to the end of currentScanners 1041 break; 1042 } 1043 } 1044 // add the newly created scanners on the flushed files and the current active memstore scanner 1045 addCurrentScanners(scanners); 1046 // Combine all seeked scanners with a heap 1047 resetKVHeap(this.currentScanners, store.getComparator()); 1048 resetQueryMatcher(lastTop); 1049 if (heap.peek() == null || store.getComparator().compareRows(lastTop, this.heap.peek()) != 0) { 1050 LOG.info("Storescanner.peek() is changed where before = " + lastTop.toString() 1051 + ",and after = " + heap.peek()); 1052 topChanged = true; 1053 } else { 1054 topChanged = false; 1055 } 1056 return topChanged; 1057 } 1058 1059 private void resetQueryMatcher(Cell lastTopKey) { 1060 // Reset the state of the Query Matcher and set to top row. 1061 // Only reset and call setRow if the row changes; avoids confusing the 1062 // query matcher if scanning intra-row. 1063 Cell cell = heap.peek(); 1064 if (cell == null) { 1065 cell = lastTopKey; 1066 } 1067 if ((matcher.currentRow() == null) || !CellUtil.matchingRows(cell, matcher.currentRow())) { 1068 this.countPerRow = 0; 1069 // The setToNewRow will call reset internally 1070 matcher.setToNewRow(cell); 1071 } 1072 } 1073 1074 /** 1075 * Check whether scan as expected order 1076 */ 1077 protected void checkScanOrder(Cell prevKV, Cell kv, CellComparator comparator) 1078 throws IOException { 1079 // Check that the heap gives us KVs in an increasing order. 1080 assert prevKV == null || comparator == null || comparator.compare(prevKV, kv) <= 0 1081 : "Key " + prevKV + " followed by a smaller key " + kv + " in cf " + store; 1082 } 1083 1084 protected boolean seekToNextRow(Cell c) throws IOException { 1085 return reseek(PrivateCellUtil.createLastOnRow(c)); 1086 } 1087 1088 /** 1089 * Do a reseek in a normal StoreScanner(scan forward) 1090 * @return true if scanner has values left, false if end of scanner 1091 */ 1092 protected boolean seekAsDirection(Cell kv) throws IOException { 1093 return reseek(kv); 1094 } 1095 1096 @Override 1097 public boolean reseek(Cell kv) throws IOException { 1098 if (checkFlushed()) { 1099 reopenAfterFlush(); 1100 } 1101 if (explicitColumnQuery && lazySeekEnabledGlobally) { 1102 return heap.requestSeek(kv, true, useRowColBloom); 1103 } 1104 return heap.reseek(kv); 1105 } 1106 1107 void trySwitchToStreamRead() { 1108 if ( 1109 readType != Scan.ReadType.DEFAULT || !scanUsePread || closing || heap.peek() == null 1110 || bytesRead < preadMaxBytes 1111 ) { 1112 return; 1113 } 1114 LOG.debug("Switch to stream read (scanned={} bytes) of {}", bytesRead, 1115 this.store.getColumnFamilyName()); 1116 scanUsePread = false; 1117 Cell lastTop = heap.peek(); 1118 List<KeyValueScanner> memstoreScanners = new ArrayList<>(); 1119 List<KeyValueScanner> scannersToClose = new ArrayList<>(); 1120 for (KeyValueScanner kvs : currentScanners) { 1121 if (!kvs.isFileScanner()) { 1122 // collect memstorescanners here 1123 memstoreScanners.add(kvs); 1124 } else { 1125 scannersToClose.add(kvs); 1126 } 1127 } 1128 List<KeyValueScanner> fileScanners = null; 1129 List<KeyValueScanner> newCurrentScanners; 1130 KeyValueHeap newHeap; 1131 try { 1132 // We must have a store instance here so no null check 1133 // recreate the scanners on the current file scanners 1134 fileScanners = store.recreateScanners(scannersToClose, cacheBlocks, false, false, matcher, 1135 scan.getStartRow(), scan.includeStartRow(), scan.getStopRow(), scan.includeStopRow(), 1136 readPt, false); 1137 if (fileScanners == null) { 1138 return; 1139 } 1140 seekScanners(fileScanners, lastTop, false, parallelSeekEnabled); 1141 newCurrentScanners = new ArrayList<>(fileScanners.size() + memstoreScanners.size()); 1142 newCurrentScanners.addAll(fileScanners); 1143 newCurrentScanners.addAll(memstoreScanners); 1144 newHeap = newKVHeap(newCurrentScanners, comparator); 1145 } catch (Exception e) { 1146 LOG.warn("failed to switch to stream read", e); 1147 if (fileScanners != null) { 1148 fileScanners.forEach(KeyValueScanner::close); 1149 } 1150 return; 1151 } 1152 currentScanners.clear(); 1153 addCurrentScanners(newCurrentScanners); 1154 this.heap = newHeap; 1155 resetQueryMatcher(lastTop); 1156 scannersToClose.forEach(KeyValueScanner::close); 1157 } 1158 1159 protected final boolean checkFlushed() { 1160 // check the var without any lock. Suppose even if we see the old 1161 // value here still it is ok to continue because we will not be resetting 1162 // the heap but will continue with the referenced memstore's snapshot. For compactions 1163 // any way we don't need the updateReaders at all to happen as we still continue with 1164 // the older files 1165 if (flushed) { 1166 // If there is a flush and the current scan is notified on the flush ensure that the 1167 // scan's heap gets reset and we do a seek on the newly flushed file. 1168 if (this.closing) { 1169 return false; 1170 } 1171 // reset the flag 1172 flushed = false; 1173 return true; 1174 } 1175 return false; 1176 } 1177 1178 /** 1179 * Seek storefiles in parallel to optimize IO latency as much as possible 1180 * @param scanners the list {@link KeyValueScanner}s to be read from 1181 * @param kv the KeyValue on which the operation is being requested 1182 */ 1183 private void parallelSeek(final List<? extends KeyValueScanner> scanners, final Cell kv) 1184 throws IOException { 1185 if (scanners.isEmpty()) return; 1186 int storeFileScannerCount = scanners.size(); 1187 CountDownLatch latch = new CountDownLatch(storeFileScannerCount); 1188 List<ParallelSeekHandler> handlers = new ArrayList<>(storeFileScannerCount); 1189 for (KeyValueScanner scanner : scanners) { 1190 if (scanner instanceof StoreFileScanner) { 1191 ParallelSeekHandler seekHandler = new ParallelSeekHandler(scanner, kv, this.readPt, latch); 1192 executor.submit(seekHandler); 1193 handlers.add(seekHandler); 1194 } else { 1195 scanner.seek(kv); 1196 latch.countDown(); 1197 } 1198 } 1199 1200 try { 1201 latch.await(); 1202 } catch (InterruptedException ie) { 1203 throw (InterruptedIOException) new InterruptedIOException().initCause(ie); 1204 } 1205 1206 for (ParallelSeekHandler handler : handlers) { 1207 if (handler.getErr() != null) { 1208 throw new IOException(handler.getErr()); 1209 } 1210 } 1211 } 1212 1213 /** 1214 * Used in testing. 1215 * @return all scanners in no particular order 1216 */ 1217 List<KeyValueScanner> getAllScannersForTesting() { 1218 List<KeyValueScanner> allScanners = new ArrayList<>(); 1219 KeyValueScanner current = heap.getCurrentForTesting(); 1220 if (current != null) allScanners.add(current); 1221 for (KeyValueScanner scanner : heap.getHeap()) 1222 allScanners.add(scanner); 1223 return allScanners; 1224 } 1225 1226 static void enableLazySeekGlobally(boolean enable) { 1227 lazySeekEnabledGlobally = enable; 1228 } 1229 1230 /** Returns The estimated number of KVs seen by this scanner (includes some skipped KVs). */ 1231 public long getEstimatedNumberOfKvsScanned() { 1232 return this.kvsScanned; 1233 } 1234 1235 @Override 1236 public Cell getNextIndexedKey() { 1237 return this.heap.getNextIndexedKey(); 1238 } 1239 1240 @Override 1241 public void shipped() throws IOException { 1242 if (prevCell != null) { 1243 // Do the copy here so that in case the prevCell ref is pointing to the previous 1244 // blocks we can safely release those blocks. 1245 // This applies to blocks that are got from Bucket cache, L1 cache and the blocks 1246 // fetched from HDFS. Copying this would ensure that we let go the references to these 1247 // blocks so that they can be GCed safely(in case of bucket cache) 1248 prevCell = KeyValueUtil.toNewKeyCell(this.prevCell); 1249 } 1250 matcher.beforeShipped(); 1251 // There wont be further fetch of Cells from these scanners. Just close. 1252 clearAndClose(scannersForDelayedClose); 1253 if (this.heap != null) { 1254 this.heap.shipped(); 1255 // When switching from pread to stream, we will open a new scanner for each store file, but 1256 // the old scanner may still track the HFileBlocks we have scanned but not sent back to client 1257 // yet. If we close the scanner immediately then the HFileBlocks may be messed up by others 1258 // before we serialize and send it back to client. The HFileBlocks will be released in shipped 1259 // method, so we here will also open new scanners and close old scanners in shipped method. 1260 // See HBASE-18055 for more details. 1261 trySwitchToStreamRead(); 1262 } 1263 } 1264}