001/*
002 * Licensed to the Apache Software Foundation (ASF) under one or more
003 * contributor license agreements. See the NOTICE file distributed with this
004 * work for additional information regarding copyright ownership. The ASF
005 * licenses this file to you under the Apache License, Version 2.0 (the
006 * "License"); you may not use this file except in compliance with the License.
007 * You may obtain a copy of the License at
008 *
009 * http://www.apache.org/licenses/LICENSE-2.0
010 *
011 * Unless required by applicable law or agreed to in writing, software
012 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
013 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
014 * License for the specific language governing permissions and limitations
015 * under the License.
016 */
017package org.apache.hadoop.hbase.util;
018
019import java.io.IOException;
020import java.io.InterruptedIOException;
021import java.lang.reflect.Constructor;
022import java.security.SecureRandom;
023import java.util.ArrayList;
024import java.util.Arrays;
025import java.util.List;
026import java.util.Properties;
027import java.util.concurrent.atomic.AtomicReference;
028
029import javax.crypto.spec.SecretKeySpec;
030
031import org.apache.hadoop.conf.Configuration;
032import org.apache.hadoop.hbase.HBaseConfiguration;
033import org.apache.hadoop.hbase.HBaseInterfaceAudience;
034import org.apache.hadoop.hbase.HBaseTestingUtility;
035import org.apache.hadoop.hbase.HConstants;
036import org.apache.hadoop.hbase.TableName;
037import org.apache.hadoop.hbase.client.ColumnFamilyDescriptor;
038import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;
039import org.apache.yetus.audience.InterfaceAudience;
040import org.slf4j.Logger;
041import org.slf4j.LoggerFactory;
042import org.apache.hadoop.hbase.client.Admin;
043import org.apache.hadoop.hbase.client.Connection;
044import org.apache.hadoop.hbase.client.ConnectionFactory;
045import org.apache.hadoop.hbase.client.Durability;
046import org.apache.hadoop.hbase.client.TableDescriptor;
047import org.apache.hadoop.hbase.io.compress.Compression;
048import org.apache.hadoop.hbase.io.crypto.Cipher;
049import org.apache.hadoop.hbase.io.crypto.Encryption;
050import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding;
051import org.apache.hadoop.hbase.log.HBaseMarkers;
052import org.apache.hadoop.hbase.regionserver.BloomType;
053import org.apache.hadoop.hbase.security.EncryptionUtil;
054import org.apache.hadoop.hbase.security.HBaseKerberosUtils;
055import org.apache.hadoop.hbase.security.User;
056import org.apache.hadoop.hbase.security.access.AccessControlClient;
057import org.apache.hadoop.hbase.security.access.Permission;
058import org.apache.hadoop.hbase.util.test.LoadTestDataGenerator;
059import org.apache.hadoop.hbase.util.test.LoadTestDataGeneratorWithACL;
060import org.apache.hadoop.util.ToolRunner;
061
062import org.apache.hbase.thirdparty.org.apache.commons.cli.AlreadySelectedException;
063import org.apache.hbase.thirdparty.org.apache.commons.cli.CommandLine;
064import org.apache.hbase.thirdparty.org.apache.commons.cli.CommandLineParser;
065import org.apache.hbase.thirdparty.org.apache.commons.cli.DefaultParser;
066import org.apache.hbase.thirdparty.org.apache.commons.cli.MissingOptionException;
067import org.apache.hbase.thirdparty.org.apache.commons.cli.Options;
068import org.apache.hbase.thirdparty.org.apache.commons.cli.ParseException;
069
070/**
071 * A command-line utility that reads, writes, and verifies data. Unlike
072 * {@link org.apache.hadoop.hbase.PerformanceEvaluation}, this tool validates the data written,
073 * and supports simultaneously writing and reading the same set of keys.
074 */
075@InterfaceAudience.LimitedPrivate(HBaseInterfaceAudience.TOOLS)
076public class LoadTestTool extends AbstractHBaseTool {
077
078  private static final Logger LOG = LoggerFactory.getLogger(LoadTestTool.class);
079  private static final String COLON = ":";
080
081  /** Table name for the test */
082  private TableName tableName;
083
084  /** Column families for the test */
085  private byte[][] families;
086
087  /** Table name to use of not overridden on the command line */
088  protected static final String DEFAULT_TABLE_NAME = "cluster_test";
089
090  /** The default data size if not specified */
091  protected static final int DEFAULT_DATA_SIZE = 64;
092
093  /** The number of reader/writer threads if not specified */
094  protected static final int DEFAULT_NUM_THREADS = 20;
095
096  /** Usage string for the load option */
097  protected static final String OPT_USAGE_LOAD =
098      "<avg_cols_per_key>:<avg_data_size>" +
099      "[:<#threads=" + DEFAULT_NUM_THREADS + ">]";
100
101  /** Usage string for the read option */
102  protected static final String OPT_USAGE_READ =
103      "<verify_percent>[:<#threads=" + DEFAULT_NUM_THREADS + ">]";
104
105  /** Usage string for the update option */
106  protected static final String OPT_USAGE_UPDATE =
107      "<update_percent>[:<#threads=" + DEFAULT_NUM_THREADS
108      + ">][:<#whether to ignore nonce collisions=0>]";
109
110  protected static final String OPT_USAGE_BLOOM = "Bloom filter type, one of " +
111      Arrays.toString(BloomType.values());
112
113  protected static final String OPT_USAGE_COMPRESSION = "Compression type, " +
114      "one of " + Arrays.toString(Compression.Algorithm.values());
115
116  public static final String OPT_BLOOM = "bloom";
117  public static final String OPT_COMPRESSION = "compression";
118  public static final String OPT_DEFERRED_LOG_FLUSH = "deferredlogflush";
119  public static final String OPT_DEFERRED_LOG_FLUSH_USAGE = "Enable deferred log flush.";
120
121  public static final String OPT_INMEMORY = "in_memory";
122  public static final String OPT_USAGE_IN_MEMORY = "Tries to keep the HFiles of the CF " +
123      "inmemory as far as possible.  Not guaranteed that reads are always served from inmemory";
124
125  public static final String OPT_GENERATOR = "generator";
126  public static final String OPT_GENERATOR_USAGE = "The class which generates load for the tool."
127      + " Any args for this class can be passed as colon separated after class name";
128
129  public static final String OPT_WRITER = "writer";
130  public static final String OPT_WRITER_USAGE = "The class for executing the write requests";
131
132  public static final String OPT_UPDATER = "updater";
133  public static final String OPT_UPDATER_USAGE = "The class for executing the update requests";
134
135  public static final String OPT_READER = "reader";
136  public static final String OPT_READER_USAGE = "The class for executing the read requests";
137
138  protected static final String OPT_KEY_WINDOW = "key_window";
139  protected static final String OPT_WRITE = "write";
140  protected static final String OPT_MAX_READ_ERRORS = "max_read_errors";
141  public static final String OPT_MULTIPUT = "multiput";
142  public static final String OPT_MULTIGET = "multiget_batchsize";
143  protected static final String OPT_NUM_KEYS = "num_keys";
144  protected static final String OPT_READ = "read";
145  protected static final String OPT_START_KEY = "start_key";
146  public static final String OPT_TABLE_NAME = "tn";
147  public static final String OPT_COLUMN_FAMILIES = "families";
148  protected static final String OPT_ZK_QUORUM = "zk";
149  protected static final String OPT_ZK_PARENT_NODE = "zk_root";
150  protected static final String OPT_SKIP_INIT = "skip_init";
151  protected static final String OPT_INIT_ONLY = "init_only";
152  protected static final String NUM_TABLES = "num_tables";
153  protected static final String OPT_BATCHUPDATE = "batchupdate";
154  protected static final String OPT_UPDATE = "update";
155
156  public static final String OPT_ENCRYPTION = "encryption";
157  protected static final String OPT_ENCRYPTION_USAGE =
158    "Enables transparent encryption on the test table, one of " +
159    Arrays.toString(Encryption.getSupportedCiphers());
160
161  public static final String OPT_NUM_REGIONS_PER_SERVER = "num_regions_per_server";
162  protected static final String OPT_NUM_REGIONS_PER_SERVER_USAGE
163    = "Desired number of regions per region server. Defaults to 5.";
164  public static int DEFAULT_NUM_REGIONS_PER_SERVER = 5;
165
166  public static final String OPT_REGION_REPLICATION = "region_replication";
167  protected static final String OPT_REGION_REPLICATION_USAGE =
168      "Desired number of replicas per region";
169
170  public static final String OPT_REGION_REPLICA_ID = "region_replica_id";
171  protected static final String OPT_REGION_REPLICA_ID_USAGE =
172      "Region replica id to do the reads from";
173
174  public static final String OPT_MOB_THRESHOLD = "mob_threshold";
175  protected static final String OPT_MOB_THRESHOLD_USAGE =
176      "Desired cell size to exceed in bytes that will use the MOB write path";
177
178  protected static final long DEFAULT_START_KEY = 0;
179
180  /** This will be removed as we factor out the dependency on command line */
181  protected CommandLine cmd;
182
183  protected MultiThreadedWriter writerThreads = null;
184  protected MultiThreadedReader readerThreads = null;
185  protected MultiThreadedUpdater updaterThreads = null;
186
187  protected long startKey, endKey;
188
189  protected boolean isWrite, isRead, isUpdate;
190  protected boolean deferredLogFlush;
191
192  // Column family options
193  protected DataBlockEncoding dataBlockEncodingAlgo;
194  protected Compression.Algorithm compressAlgo;
195  protected BloomType bloomType;
196  private boolean inMemoryCF;
197
198  private User userOwner;
199  // Writer options
200  protected int numWriterThreads = DEFAULT_NUM_THREADS;
201  protected int minColsPerKey, maxColsPerKey;
202  protected int minColDataSize = DEFAULT_DATA_SIZE, maxColDataSize = DEFAULT_DATA_SIZE;
203  protected boolean isMultiPut;
204
205  // Updater options
206  protected int numUpdaterThreads = DEFAULT_NUM_THREADS;
207  protected int updatePercent;
208  protected boolean ignoreConflicts = false;
209  protected boolean isBatchUpdate;
210
211  // Reader options
212  private int numReaderThreads = DEFAULT_NUM_THREADS;
213  private int keyWindow = MultiThreadedReader.DEFAULT_KEY_WINDOW;
214  private int multiGetBatchSize = MultiThreadedReader.DEFAULT_BATCH_SIZE;
215  private int maxReadErrors = MultiThreadedReader.DEFAULT_MAX_ERRORS;
216  private int verifyPercent;
217
218  private int numTables = 1;
219
220  private String superUser;
221
222  private String userNames;
223  //This file is used to read authentication information in secure clusters.
224  private String authnFileName;
225
226  private int numRegionsPerServer = DEFAULT_NUM_REGIONS_PER_SERVER;
227  private int regionReplication = -1; // not set
228  private int regionReplicaId = -1; // not set
229
230  private int mobThreshold = -1; // not set
231
232  // TODO: refactor LoadTestToolImpl somewhere to make the usage from tests less bad,
233  //       console tool itself should only be used from console.
234  protected boolean isSkipInit = false;
235  protected boolean isInitOnly = false;
236
237  protected Cipher cipher = null;
238
239  protected String[] splitColonSeparated(String option,
240      int minNumCols, int maxNumCols) {
241    String optVal = cmd.getOptionValue(option);
242    String[] cols = optVal.split(COLON);
243    if (cols.length < minNumCols || cols.length > maxNumCols) {
244      throw new IllegalArgumentException("Expected at least "
245          + minNumCols + " columns but no more than " + maxNumCols +
246          " in the colon-separated value '" + optVal + "' of the " +
247          "-" + option + " option");
248    }
249    return cols;
250  }
251
252  protected int getNumThreads(String numThreadsStr) {
253    return parseInt(numThreadsStr, 1, Short.MAX_VALUE);
254  }
255
256  public byte[][] getColumnFamilies() {
257    return families;
258  }
259
260  /**
261   * Apply column family options such as Bloom filters, compression, and data
262   * block encoding.
263   */
264  protected void applyColumnFamilyOptions(TableName tableName,
265      byte[][] columnFamilies) throws IOException {
266    try (Connection conn = ConnectionFactory.createConnection(conf);
267        Admin admin = conn.getAdmin()) {
268      TableDescriptor tableDesc = admin.getDescriptor(tableName);
269      LOG.info("Disabling table " + tableName);
270      admin.disableTable(tableName);
271      for (byte[] cf : columnFamilies) {
272        ColumnFamilyDescriptor columnDesc = tableDesc.getColumnFamily(cf);
273        boolean isNewCf = columnDesc == null;
274        ColumnFamilyDescriptorBuilder columnDescBuilder = isNewCf ?
275            ColumnFamilyDescriptorBuilder.newBuilder(cf) :
276            ColumnFamilyDescriptorBuilder.newBuilder(columnDesc);
277        if (bloomType != null) {
278          columnDescBuilder.setBloomFilterType(bloomType);
279        }
280        if (compressAlgo != null) {
281          columnDescBuilder.setCompressionType(compressAlgo);
282        }
283        if (dataBlockEncodingAlgo != null) {
284          columnDescBuilder.setDataBlockEncoding(dataBlockEncodingAlgo);
285        }
286        if (inMemoryCF) {
287          columnDescBuilder.setInMemory(inMemoryCF);
288        }
289        if (cipher != null) {
290          byte[] keyBytes = new byte[cipher.getKeyLength()];
291          new SecureRandom().nextBytes(keyBytes);
292          columnDescBuilder.setEncryptionType(cipher.getName());
293          columnDescBuilder.setEncryptionKey(
294              EncryptionUtil.wrapKey(conf,
295                  User.getCurrent().getShortName(),
296                  new SecretKeySpec(keyBytes,
297                      cipher.getName())));
298        }
299        if (mobThreshold >= 0) {
300          columnDescBuilder.setMobEnabled(true);
301          columnDescBuilder.setMobThreshold(mobThreshold);
302        }
303
304        if (isNewCf) {
305          admin.addColumnFamily(tableName, columnDescBuilder.build());
306        } else {
307          admin.modifyColumnFamily(tableName, columnDescBuilder.build());
308        }
309      }
310      LOG.info("Enabling table " + tableName);
311      admin.enableTable(tableName);
312    }
313  }
314
315  @Override
316  protected void addOptions() {
317    addOptWithArg(OPT_ZK_QUORUM, "ZK quorum as comma-separated host names " +
318        "without port numbers");
319    addOptWithArg(OPT_ZK_PARENT_NODE, "name of parent znode in zookeeper");
320    addOptWithArg(OPT_TABLE_NAME, "The name of the table to read or write");
321    addOptWithArg(OPT_COLUMN_FAMILIES, "The name of the column families to use separated by comma");
322    addOptWithArg(OPT_WRITE, OPT_USAGE_LOAD);
323    addOptWithArg(OPT_READ, OPT_USAGE_READ);
324    addOptWithArg(OPT_UPDATE, OPT_USAGE_UPDATE);
325    addOptNoArg(OPT_INIT_ONLY, "Initialize the test table only, don't do any loading");
326    addOptWithArg(OPT_BLOOM, OPT_USAGE_BLOOM);
327    addOptWithArg(OPT_COMPRESSION, OPT_USAGE_COMPRESSION);
328    addOptWithArg(HFileTestUtil.OPT_DATA_BLOCK_ENCODING, HFileTestUtil.OPT_DATA_BLOCK_ENCODING_USAGE);
329    addOptWithArg(OPT_MAX_READ_ERRORS, "The maximum number of read errors " +
330        "to tolerate before terminating all reader threads. The default is " +
331        MultiThreadedReader.DEFAULT_MAX_ERRORS + ".");
332    addOptWithArg(OPT_MULTIGET, "Whether to use multi-gets as opposed to " +
333        "separate gets for every column in a row");
334    addOptWithArg(OPT_KEY_WINDOW, "The 'key window' to maintain between " +
335        "reads and writes for concurrent write/read workload. The default " +
336        "is " + MultiThreadedReader.DEFAULT_KEY_WINDOW + ".");
337
338    addOptNoArg(OPT_MULTIPUT, "Whether to use multi-puts as opposed to " +
339        "separate puts for every column in a row");
340    addOptNoArg(OPT_BATCHUPDATE, "Whether to use batch as opposed to " +
341        "separate updates for every column in a row");
342    addOptNoArg(OPT_INMEMORY, OPT_USAGE_IN_MEMORY);
343    addOptWithArg(OPT_GENERATOR, OPT_GENERATOR_USAGE);
344    addOptWithArg(OPT_WRITER, OPT_WRITER_USAGE);
345    addOptWithArg(OPT_UPDATER, OPT_UPDATER_USAGE);
346    addOptWithArg(OPT_READER, OPT_READER_USAGE);
347
348    addOptWithArg(OPT_NUM_KEYS, "The number of keys to read/write");
349    addOptWithArg(OPT_START_KEY, "The first key to read/write " +
350        "(a 0-based index). The default value is " +
351        DEFAULT_START_KEY + ".");
352    addOptNoArg(OPT_SKIP_INIT, "Skip the initialization; assume test table "
353        + "already exists");
354
355    addOptWithArg(NUM_TABLES,
356      "A positive integer number. When a number n is specified, load test "
357          + "tool  will load n table parallely. -tn parameter value becomes "
358          + "table name prefix. Each table name is in format <tn>_1...<tn>_n");
359
360    addOptWithArg(OPT_ENCRYPTION, OPT_ENCRYPTION_USAGE);
361    addOptNoArg(OPT_DEFERRED_LOG_FLUSH, OPT_DEFERRED_LOG_FLUSH_USAGE);
362    addOptWithArg(OPT_NUM_REGIONS_PER_SERVER, OPT_NUM_REGIONS_PER_SERVER_USAGE);
363    addOptWithArg(OPT_REGION_REPLICATION, OPT_REGION_REPLICATION_USAGE);
364    addOptWithArg(OPT_REGION_REPLICA_ID, OPT_REGION_REPLICA_ID_USAGE);
365    addOptWithArg(OPT_MOB_THRESHOLD, OPT_MOB_THRESHOLD_USAGE);
366  }
367
368  @Override
369  protected CommandLineParser newParser() {
370    // Commons-CLI lacks the capability to handle combinations of options, so we do it ourselves
371    // Validate in parse() to get helpful error messages instead of exploding in processOptions()
372    return new DefaultParser() {
373      @Override
374      public CommandLine parse(Options opts, String[] args, Properties props, boolean stop)
375          throws ParseException {
376        CommandLine cl = super.parse(opts, args, props, stop);
377
378        boolean isReadWriteUpdate = cmd.hasOption(OPT_READ)
379            || cmd.hasOption(OPT_WRITE)
380            || cmd.hasOption(OPT_UPDATE);
381        boolean isInitOnly = cmd.hasOption(OPT_INIT_ONLY);
382
383        if (!isInitOnly && !isReadWriteUpdate) {
384          throw new MissingOptionException("Must specify either -" + OPT_INIT_ONLY
385              + " or at least one of -" + OPT_READ + ", -" + OPT_WRITE + ", -" + OPT_UPDATE);
386        }
387
388        if (isInitOnly && isReadWriteUpdate) {
389          throw new AlreadySelectedException(OPT_INIT_ONLY + " cannot be specified with any of -"
390              + OPT_READ + ", -" + OPT_WRITE + ", -" + OPT_UPDATE);
391        }
392
393        if (isReadWriteUpdate && !cmd.hasOption(OPT_NUM_KEYS)) {
394          throw new MissingOptionException(OPT_NUM_KEYS + " must be specified in read/write mode.");
395        }
396
397        return cl;
398      }
399    };
400  }
401
402  @Override
403  protected void processOptions(CommandLine cmd) {
404    this.cmd = cmd;
405
406    tableName = TableName.valueOf(cmd.getOptionValue(OPT_TABLE_NAME,
407        DEFAULT_TABLE_NAME));
408
409    if (cmd.hasOption(OPT_COLUMN_FAMILIES)) {
410      String[] list = cmd.getOptionValue(OPT_COLUMN_FAMILIES).split(",");
411      families = new byte[list.length][];
412      for (int i = 0; i < list.length; i++) {
413        families[i] = Bytes.toBytes(list[i]);
414      }
415    } else {
416      families = HFileTestUtil.DEFAULT_COLUMN_FAMILIES;
417    }
418
419    isWrite = cmd.hasOption(OPT_WRITE);
420    isRead = cmd.hasOption(OPT_READ);
421    isUpdate = cmd.hasOption(OPT_UPDATE);
422    isInitOnly = cmd.hasOption(OPT_INIT_ONLY);
423    deferredLogFlush = cmd.hasOption(OPT_DEFERRED_LOG_FLUSH);
424
425    if (!isInitOnly) {
426      startKey = parseLong(cmd.getOptionValue(OPT_START_KEY,
427          String.valueOf(DEFAULT_START_KEY)), 0, Long.MAX_VALUE);
428      long numKeys = parseLong(cmd.getOptionValue(OPT_NUM_KEYS), 1,
429          Long.MAX_VALUE - startKey);
430      endKey = startKey + numKeys;
431      isSkipInit = cmd.hasOption(OPT_SKIP_INIT);
432      System.out.println("Key range: [" + startKey + ".." + (endKey - 1) + "]");
433    }
434
435    parseColumnFamilyOptions(cmd);
436
437    if (isWrite) {
438      String[] writeOpts = splitColonSeparated(OPT_WRITE, 2, 3);
439
440      int colIndex = 0;
441      minColsPerKey = 1;
442      maxColsPerKey = 2 * Integer.parseInt(writeOpts[colIndex++]);
443      int avgColDataSize =
444          parseInt(writeOpts[colIndex++], 1, Integer.MAX_VALUE);
445      minColDataSize = avgColDataSize / 2;
446      maxColDataSize = avgColDataSize * 3 / 2;
447
448      if (colIndex < writeOpts.length) {
449        numWriterThreads = getNumThreads(writeOpts[colIndex++]);
450      }
451
452      isMultiPut = cmd.hasOption(OPT_MULTIPUT);
453
454      mobThreshold = -1;
455      if (cmd.hasOption(OPT_MOB_THRESHOLD)) {
456        mobThreshold = Integer.parseInt(cmd.getOptionValue(OPT_MOB_THRESHOLD));
457      }
458
459      System.out.println("Multi-puts: " + isMultiPut);
460      System.out.println("Columns per key: " + minColsPerKey + ".."
461          + maxColsPerKey);
462      System.out.println("Data size per column: " + minColDataSize + ".."
463          + maxColDataSize);
464    }
465
466    if (isUpdate) {
467      String[] mutateOpts = splitColonSeparated(OPT_UPDATE, 1, 3);
468      int colIndex = 0;
469      updatePercent = parseInt(mutateOpts[colIndex++], 0, 100);
470      if (colIndex < mutateOpts.length) {
471        numUpdaterThreads = getNumThreads(mutateOpts[colIndex++]);
472      }
473      if (colIndex < mutateOpts.length) {
474        ignoreConflicts = parseInt(mutateOpts[colIndex++], 0, 1) == 1;
475      }
476
477      isBatchUpdate = cmd.hasOption(OPT_BATCHUPDATE);
478
479      System.out.println("Batch updates: " + isBatchUpdate);
480      System.out.println("Percent of keys to update: " + updatePercent);
481      System.out.println("Updater threads: " + numUpdaterThreads);
482      System.out.println("Ignore nonce conflicts: " + ignoreConflicts);
483    }
484
485    if (isRead) {
486      String[] readOpts = splitColonSeparated(OPT_READ, 1, 2);
487      int colIndex = 0;
488      verifyPercent = parseInt(readOpts[colIndex++], 0, 100);
489      if (colIndex < readOpts.length) {
490        numReaderThreads = getNumThreads(readOpts[colIndex++]);
491      }
492
493      if (cmd.hasOption(OPT_MAX_READ_ERRORS)) {
494        maxReadErrors = parseInt(cmd.getOptionValue(OPT_MAX_READ_ERRORS),
495            0, Integer.MAX_VALUE);
496      }
497
498      if (cmd.hasOption(OPT_KEY_WINDOW)) {
499        keyWindow = parseInt(cmd.getOptionValue(OPT_KEY_WINDOW),
500            0, Integer.MAX_VALUE);
501      }
502
503      if (cmd.hasOption(OPT_MULTIGET)) {
504        multiGetBatchSize = parseInt(cmd.getOptionValue(OPT_MULTIGET),
505            0, Integer.MAX_VALUE);
506      }
507
508      System.out.println("Multi-gets (value of 1 means no multigets): " + multiGetBatchSize);
509      System.out.println("Percent of keys to verify: " + verifyPercent);
510      System.out.println("Reader threads: " + numReaderThreads);
511    }
512
513    numTables = 1;
514    if (cmd.hasOption(NUM_TABLES)) {
515      numTables = parseInt(cmd.getOptionValue(NUM_TABLES), 1, Short.MAX_VALUE);
516    }
517
518    numRegionsPerServer = DEFAULT_NUM_REGIONS_PER_SERVER;
519    if (cmd.hasOption(OPT_NUM_REGIONS_PER_SERVER)) {
520      numRegionsPerServer = Integer.parseInt(cmd.getOptionValue(OPT_NUM_REGIONS_PER_SERVER));
521    }
522
523    regionReplication = 1;
524    if (cmd.hasOption(OPT_REGION_REPLICATION)) {
525      regionReplication = Integer.parseInt(cmd.getOptionValue(OPT_REGION_REPLICATION));
526    }
527
528    regionReplicaId = -1;
529    if (cmd.hasOption(OPT_REGION_REPLICA_ID)) {
530      regionReplicaId = Integer.parseInt(cmd.getOptionValue(OPT_REGION_REPLICA_ID));
531    }
532  }
533
534  private void parseColumnFamilyOptions(CommandLine cmd) {
535    String dataBlockEncodingStr = cmd.getOptionValue(HFileTestUtil.OPT_DATA_BLOCK_ENCODING);
536    dataBlockEncodingAlgo = dataBlockEncodingStr == null ? null :
537        DataBlockEncoding.valueOf(dataBlockEncodingStr);
538
539    String compressStr = cmd.getOptionValue(OPT_COMPRESSION);
540    compressAlgo = compressStr == null ? Compression.Algorithm.NONE :
541        Compression.Algorithm.valueOf(compressStr);
542
543    String bloomStr = cmd.getOptionValue(OPT_BLOOM);
544    bloomType = bloomStr == null ? BloomType.ROW :
545        BloomType.valueOf(bloomStr);
546
547    inMemoryCF = cmd.hasOption(OPT_INMEMORY);
548    if (cmd.hasOption(OPT_ENCRYPTION)) {
549      cipher = Encryption.getCipher(conf, cmd.getOptionValue(OPT_ENCRYPTION));
550    }
551
552  }
553
554  public void initTestTable() throws IOException {
555    Durability durability = Durability.USE_DEFAULT;
556    if (deferredLogFlush) {
557      durability = Durability.ASYNC_WAL;
558    }
559
560    HBaseTestingUtility.createPreSplitLoadTestTable(conf, tableName,
561      getColumnFamilies(), compressAlgo, dataBlockEncodingAlgo, numRegionsPerServer,
562        regionReplication, durability);
563    applyColumnFamilyOptions(tableName, getColumnFamilies());
564  }
565
566  @Override
567  protected int doWork() throws IOException {
568    if (numTables > 1) {
569      return parallelLoadTables();
570    } else {
571      return loadTable();
572    }
573  }
574
575  protected int loadTable() throws IOException {
576    if (cmd.hasOption(OPT_ZK_QUORUM)) {
577      conf.set(HConstants.ZOOKEEPER_QUORUM, cmd.getOptionValue(OPT_ZK_QUORUM));
578    }
579    if (cmd.hasOption(OPT_ZK_PARENT_NODE)) {
580      conf.set(HConstants.ZOOKEEPER_ZNODE_PARENT, cmd.getOptionValue(OPT_ZK_PARENT_NODE));
581    }
582
583    if (isInitOnly) {
584      LOG.info("Initializing only; no reads or writes");
585      initTestTable();
586      return 0;
587    }
588
589    if (!isSkipInit) {
590      initTestTable();
591    }
592    LoadTestDataGenerator dataGen = null;
593    if (cmd.hasOption(OPT_GENERATOR)) {
594      String[] clazzAndArgs = cmd.getOptionValue(OPT_GENERATOR).split(COLON);
595      dataGen = getLoadGeneratorInstance(clazzAndArgs[0]);
596      String[] args;
597      if (dataGen instanceof LoadTestDataGeneratorWithACL) {
598        LOG.info("Using LoadTestDataGeneratorWithACL");
599        if (User.isHBaseSecurityEnabled(conf)) {
600          LOG.info("Security is enabled");
601          authnFileName = clazzAndArgs[1];
602          superUser = clazzAndArgs[2];
603          userNames = clazzAndArgs[3];
604          args = Arrays.copyOfRange(clazzAndArgs, 2, clazzAndArgs.length);
605          Properties authConfig = new Properties();
606          authConfig.load(this.getClass().getClassLoader().getResourceAsStream(authnFileName));
607          try {
608            addAuthInfoToConf(authConfig, conf, superUser, userNames);
609          } catch (IOException exp) {
610            LOG.error(exp.toString(), exp);
611            return EXIT_FAILURE;
612          }
613          userOwner = User.create(HBaseKerberosUtils.loginAndReturnUGI(conf, superUser));
614        } else {
615          superUser = clazzAndArgs[1];
616          userNames = clazzAndArgs[2];
617          args = Arrays.copyOfRange(clazzAndArgs, 1, clazzAndArgs.length);
618          userOwner = User.createUserForTesting(conf, superUser, new String[0]);
619        }
620      } else {
621        args = clazzAndArgs.length == 1 ? new String[0] : Arrays.copyOfRange(clazzAndArgs, 1,
622            clazzAndArgs.length);
623      }
624      dataGen.initialize(args);
625    } else {
626      // Default DataGenerator is MultiThreadedAction.DefaultDataGenerator
627      dataGen = new MultiThreadedAction.DefaultDataGenerator(minColDataSize, maxColDataSize,
628          minColsPerKey, maxColsPerKey, families);
629    }
630
631    if (userOwner != null) {
632      LOG.info("Granting permissions for user " + userOwner.getShortName());
633      Permission.Action[] actions = {
634        Permission.Action.ADMIN, Permission.Action.CREATE,
635        Permission.Action.READ, Permission.Action.WRITE };
636      try {
637        AccessControlClient.grant(ConnectionFactory.createConnection(conf),
638            tableName, userOwner.getShortName(), null, null, actions);
639      } catch (Throwable e) {
640        LOG.error(HBaseMarkers.FATAL, "Error in granting permission for the user " +
641            userOwner.getShortName(), e);
642        return EXIT_FAILURE;
643      }
644    }
645
646    if (userNames != null) {
647      // This will be comma separated list of expressions.
648      String users[] = userNames.split(",");
649      User user = null;
650      for (String userStr : users) {
651        if (User.isHBaseSecurityEnabled(conf)) {
652          user = User.create(HBaseKerberosUtils.loginAndReturnUGI(conf, userStr));
653        } else {
654          user = User.createUserForTesting(conf, userStr, new String[0]);
655        }
656      }
657    }
658
659    if (isWrite) {
660      if (userOwner != null) {
661        writerThreads = new MultiThreadedWriterWithACL(dataGen, conf, tableName, userOwner);
662      } else {
663        String writerClass = null;
664        if (cmd.hasOption(OPT_WRITER)) {
665          writerClass = cmd.getOptionValue(OPT_WRITER);
666        } else {
667          writerClass = MultiThreadedWriter.class.getCanonicalName();
668        }
669
670        writerThreads = getMultiThreadedWriterInstance(writerClass, dataGen);
671      }
672      writerThreads.setMultiPut(isMultiPut);
673    }
674
675    if (isUpdate) {
676      if (userOwner != null) {
677        updaterThreads = new MultiThreadedUpdaterWithACL(dataGen, conf, tableName, updatePercent,
678            userOwner, userNames);
679      } else {
680        String updaterClass = null;
681        if (cmd.hasOption(OPT_UPDATER)) {
682          updaterClass = cmd.getOptionValue(OPT_UPDATER);
683        } else {
684          updaterClass = MultiThreadedUpdater.class.getCanonicalName();
685        }
686        updaterThreads = getMultiThreadedUpdaterInstance(updaterClass, dataGen);
687      }
688      updaterThreads.setBatchUpdate(isBatchUpdate);
689      updaterThreads.setIgnoreNonceConflicts(ignoreConflicts);
690    }
691
692    if (isRead) {
693      if (userOwner != null) {
694        readerThreads = new MultiThreadedReaderWithACL(dataGen, conf, tableName, verifyPercent,
695            userNames);
696      } else {
697        String readerClass = null;
698        if (cmd.hasOption(OPT_READER)) {
699          readerClass = cmd.getOptionValue(OPT_READER);
700        } else {
701          readerClass = MultiThreadedReader.class.getCanonicalName();
702        }
703        readerThreads = getMultiThreadedReaderInstance(readerClass, dataGen);
704      }
705      readerThreads.setMaxErrors(maxReadErrors);
706      readerThreads.setKeyWindow(keyWindow);
707      readerThreads.setMultiGetBatchSize(multiGetBatchSize);
708      readerThreads.setRegionReplicaId(regionReplicaId);
709    }
710
711    if (isUpdate && isWrite) {
712      LOG.info("Concurrent write/update workload: making updaters aware of the " +
713        "write point");
714      updaterThreads.linkToWriter(writerThreads);
715    }
716
717    if (isRead && (isUpdate || isWrite)) {
718      LOG.info("Concurrent write/read workload: making readers aware of the " +
719        "write point");
720      readerThreads.linkToWriter(isUpdate ? updaterThreads : writerThreads);
721    }
722
723    if (isWrite) {
724      System.out.println("Starting to write data...");
725      writerThreads.start(startKey, endKey, numWriterThreads);
726    }
727
728    if (isUpdate) {
729      LOG.info("Starting to mutate data...");
730      System.out.println("Starting to mutate data...");
731      // TODO : currently append and increment operations not tested with tags
732      // Will update this after it is done
733      updaterThreads.start(startKey, endKey, numUpdaterThreads);
734    }
735
736    if (isRead) {
737      System.out.println("Starting to read data...");
738      readerThreads.start(startKey, endKey, numReaderThreads);
739    }
740
741    if (isWrite) {
742      writerThreads.waitForFinish();
743    }
744
745    if (isUpdate) {
746      updaterThreads.waitForFinish();
747    }
748
749    if (isRead) {
750      readerThreads.waitForFinish();
751    }
752
753    boolean success = true;
754    if (isWrite) {
755      success = success && writerThreads.getNumWriteFailures() == 0;
756    }
757    if (isUpdate) {
758      success = success && updaterThreads.getNumWriteFailures() == 0;
759    }
760    if (isRead) {
761      success = success && readerThreads.getNumReadErrors() == 0
762          && readerThreads.getNumReadFailures() == 0;
763    }
764    return success ? EXIT_SUCCESS : EXIT_FAILURE;
765  }
766
767  private LoadTestDataGenerator getLoadGeneratorInstance(String clazzName) throws IOException {
768    try {
769      Class<?> clazz = Class.forName(clazzName);
770      Constructor<?> constructor = clazz.getConstructor(int.class, int.class, int.class, int.class,
771          byte[][].class);
772      return (LoadTestDataGenerator) constructor.newInstance(minColDataSize, maxColDataSize,
773          minColsPerKey, maxColsPerKey, families);
774    } catch (Exception e) {
775      throw new IOException(e);
776    }
777  }
778
779  private MultiThreadedWriter getMultiThreadedWriterInstance(String clazzName
780      , LoadTestDataGenerator dataGen) throws IOException {
781    try {
782      Class<?> clazz = Class.forName(clazzName);
783      Constructor<?> constructor = clazz.getConstructor(
784        LoadTestDataGenerator.class, Configuration.class, TableName.class);
785      return (MultiThreadedWriter) constructor.newInstance(dataGen, conf, tableName);
786    } catch (Exception e) {
787      throw new IOException(e);
788    }
789  }
790
791  private MultiThreadedUpdater getMultiThreadedUpdaterInstance(String clazzName
792      , LoadTestDataGenerator dataGen) throws IOException {
793    try {
794      Class<?> clazz = Class.forName(clazzName);
795      Constructor<?> constructor = clazz.getConstructor(
796        LoadTestDataGenerator.class, Configuration.class, TableName.class, double.class);
797      return (MultiThreadedUpdater) constructor.newInstance(
798        dataGen, conf, tableName, updatePercent);
799    } catch (Exception e) {
800      throw new IOException(e);
801    }
802  }
803
804  private MultiThreadedReader getMultiThreadedReaderInstance(String clazzName
805      , LoadTestDataGenerator dataGen) throws IOException {
806    try {
807      Class<?> clazz = Class.forName(clazzName);
808      Constructor<?> constructor = clazz.getConstructor(
809        LoadTestDataGenerator.class, Configuration.class, TableName.class, double.class);
810      return (MultiThreadedReader) constructor.newInstance(dataGen, conf, tableName, verifyPercent);
811    } catch (Exception e) {
812      throw new IOException(e);
813    }
814  }
815
816  public static void main(String[] args) {
817    new LoadTestTool().doStaticMain(args);
818  }
819
820  /**
821   * When NUM_TABLES is specified, the function starts multiple worker threads
822   * which individually start a LoadTestTool instance to load a table. Each
823   * table name is in format <tn>_<index>. For example, "-tn test -num_tables 2"
824   * , table names will be "test_1", "test_2"
825   *
826   * @throws IOException
827   */
828  private int parallelLoadTables()
829      throws IOException {
830    // create new command args
831    String tableName = cmd.getOptionValue(OPT_TABLE_NAME, DEFAULT_TABLE_NAME);
832    String[] newArgs = null;
833    if (!cmd.hasOption(LoadTestTool.OPT_TABLE_NAME)) {
834      newArgs = new String[cmdLineArgs.length + 2];
835      newArgs[0] = "-" + LoadTestTool.OPT_TABLE_NAME;
836      newArgs[1] = LoadTestTool.DEFAULT_TABLE_NAME;
837      System.arraycopy(cmdLineArgs, 0, newArgs, 2, cmdLineArgs.length);
838    } else {
839      newArgs = cmdLineArgs;
840    }
841
842    int tableNameValueIndex = -1;
843    for (int j = 0; j < newArgs.length; j++) {
844      if (newArgs[j].endsWith(OPT_TABLE_NAME)) {
845        tableNameValueIndex = j + 1;
846      } else if (newArgs[j].endsWith(NUM_TABLES)) {
847        // change NUM_TABLES to 1 so that each worker loads one table
848        newArgs[j + 1] = "1";
849      }
850    }
851
852    // starting to load multiple tables
853    List<WorkerThread> workers = new ArrayList<>();
854    for (int i = 0; i < numTables; i++) {
855      String[] workerArgs = newArgs.clone();
856      workerArgs[tableNameValueIndex] = tableName + "_" + (i+1);
857      WorkerThread worker = new WorkerThread(i, workerArgs);
858      workers.add(worker);
859      LOG.info(worker + " starting");
860      worker.start();
861    }
862
863    // wait for all workers finish
864    LOG.info("Waiting for worker threads to finish");
865    for (WorkerThread t : workers) {
866      try {
867        t.join();
868      } catch (InterruptedException ie) {
869        IOException iie = new InterruptedIOException();
870        iie.initCause(ie);
871        throw iie;
872      }
873      checkForErrors();
874    }
875
876    return EXIT_SUCCESS;
877  }
878
879  // If an exception is thrown by one of worker threads, it will be
880  // stored here.
881  protected AtomicReference<Throwable> thrown = new AtomicReference<>();
882
883  private void workerThreadError(Throwable t) {
884    thrown.compareAndSet(null, t);
885  }
886
887  /**
888   * Check for errors in the writer threads. If any is found, rethrow it.
889   */
890  private void checkForErrors() throws IOException {
891    Throwable thrown = this.thrown.get();
892    if (thrown == null) return;
893    if (thrown instanceof IOException) {
894      throw (IOException) thrown;
895    } else {
896      throw new RuntimeException(thrown);
897    }
898  }
899
900  class WorkerThread extends Thread {
901    private String[] workerArgs;
902
903    WorkerThread(int i, String[] args) {
904      super("WorkerThread-" + i);
905      workerArgs = args;
906    }
907
908    @Override
909    public void run() {
910      try {
911        int ret = ToolRunner.run(HBaseConfiguration.create(), new LoadTestTool(), workerArgs);
912        if (ret != 0) {
913          throw new RuntimeException("LoadTestTool exit with non-zero return code.");
914        }
915      } catch (Exception ex) {
916        LOG.error("Error in worker thread", ex);
917        workerThreadError(ex);
918      }
919    }
920  }
921
922  private void addAuthInfoToConf(Properties authConfig, Configuration conf, String owner,
923      String userList) throws IOException {
924    List<String> users = new ArrayList<>(Arrays.asList(userList.split(",")));
925    users.add(owner);
926    for (String user : users) {
927      String keyTabFileConfKey = "hbase." + user + ".keytab.file";
928      String principalConfKey = "hbase." + user + ".kerberos.principal";
929      if (!authConfig.containsKey(keyTabFileConfKey) || !authConfig.containsKey(principalConfKey)) {
930        throw new IOException("Authentication configs missing for user : " + user);
931      }
932    }
933    for (String key : authConfig.stringPropertyNames()) {
934      conf.set(key, authConfig.getProperty(key));
935    }
936    LOG.debug("Added authentication properties to config successfully.");
937  }
938
939}