View Javadoc

1   /**
2    *
3    * Licensed to the Apache Software Foundation (ASF) under one
4    * or more contributor license agreements.  See the NOTICE file
5    * distributed with this work for additional information
6    * regarding copyright ownership.  The ASF licenses this file
7    * to you under the Apache License, Version 2.0 (the
8    * "License"); you may not use this file except in compliance
9    * with the License.  You may obtain a copy of the License at
10   *
11   *     http://www.apache.org/licenses/LICENSE-2.0
12   *
13   * Unless required by applicable law or agreed to in writing, software
14   * distributed under the License is distributed on an "AS IS" BASIS,
15   * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
16   * See the License for the specific language governing permissions and
17   * limitations under the License.
18   */
19  package org.apache.hadoop.hbase.mapreduce;
20  
21  import java.io.File;
22  import java.io.IOException;
23  import java.lang.reflect.InvocationTargetException;
24  import java.lang.reflect.Method;
25  import java.net.URL;
26  import java.net.URLDecoder;
27  import java.util.ArrayList;
28  import java.util.Enumeration;
29  import java.util.HashMap;
30  import java.util.HashSet;
31  import java.util.List;
32  import java.util.Map;
33  import java.util.Set;
34  import java.util.zip.ZipEntry;
35  import java.util.zip.ZipFile;
36  
37  import org.apache.commons.logging.Log;
38  import org.apache.commons.logging.LogFactory;
39  import org.apache.hadoop.classification.InterfaceAudience;
40  import org.apache.hadoop.classification.InterfaceStability;
41  import org.apache.hadoop.conf.Configuration;
42  import org.apache.hadoop.fs.FileSystem;
43  import org.apache.hadoop.fs.Path;
44  import org.apache.hadoop.hbase.HBaseConfiguration;
45  import org.apache.hadoop.hbase.HConstants;
46  import org.apache.hadoop.hbase.catalog.MetaReader;
47  import org.apache.hadoop.hbase.client.Put;
48  import org.apache.hadoop.hbase.client.Scan;
49  import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
50  import org.apache.hadoop.hbase.mapreduce.hadoopbackport.JarFinder;
51  import org.apache.hadoop.hbase.protobuf.ProtobufUtil;
52  import org.apache.hadoop.hbase.protobuf.generated.ClientProtos;
53  import org.apache.hadoop.hbase.security.User;
54  import org.apache.hadoop.hbase.security.UserProvider;
55  import org.apache.hadoop.hbase.security.token.AuthenticationTokenIdentifier;
56  import org.apache.hadoop.hbase.security.token.AuthenticationTokenSelector;
57  import org.apache.hadoop.hbase.util.Base64;
58  import org.apache.hadoop.hbase.util.Bytes;
59  import org.apache.hadoop.hbase.zookeeper.ZKClusterId;
60  import org.apache.hadoop.hbase.zookeeper.ZKUtil;
61  import org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher;
62  import org.apache.hadoop.io.Text;
63  import org.apache.hadoop.io.Writable;
64  import org.apache.hadoop.io.WritableComparable;
65  import org.apache.hadoop.mapreduce.InputFormat;
66  import org.apache.hadoop.mapreduce.Job;
67  import org.apache.hadoop.security.token.Token;
68  import org.apache.hadoop.util.StringUtils;
69  import org.apache.zookeeper.KeeperException;
70  import org.cliffc.high_scale_lib.Counter;
71  
72  import com.google.protobuf.InvalidProtocolBufferException;
73  
74  /**
75   * Utility for {@link TableMapper} and {@link TableReducer}
76   */
77  @SuppressWarnings({ "rawtypes", "unchecked" })
78  @InterfaceAudience.Public
79  @InterfaceStability.Stable
80  public class TableMapReduceUtil {
81    static Log LOG = LogFactory.getLog(TableMapReduceUtil.class);
82  
83    /**
84     * Use this before submitting a TableMap job. It will appropriately set up
85     * the job.
86     *
87     * @param table  The table name to read from.
88     * @param scan  The scan instance with the columns, time range etc.
89     * @param mapper  The mapper class to use.
90     * @param outputKeyClass  The class of the output key.
91     * @param outputValueClass  The class of the output value.
92     * @param job  The current job to adjust.  Make sure the passed job is
93     * carrying all necessary HBase configuration.
94     * @throws IOException When setting up the details fails.
95     */
96    public static void initTableMapperJob(String table, Scan scan,
97        Class<? extends TableMapper> mapper,
98        Class<?> outputKeyClass,
99        Class<?> outputValueClass, Job job)
100   throws IOException {
101     initTableMapperJob(table, scan, mapper, outputKeyClass, outputValueClass,
102         job, true);
103   }
104 
105   /**
106    * Use this before submitting a TableMap job. It will appropriately set up
107    * the job.
108    *
109    * @param table Binary representation of the table name to read from.
110    * @param scan  The scan instance with the columns, time range etc.
111    * @param mapper  The mapper class to use.
112    * @param outputKeyClass  The class of the output key.
113    * @param outputValueClass  The class of the output value.
114    * @param job  The current job to adjust.  Make sure the passed job is
115    * carrying all necessary HBase configuration.
116    * @throws IOException When setting up the details fails.
117    */
118    public static void initTableMapperJob(byte[] table, Scan scan,
119       Class<? extends TableMapper> mapper,
120       Class<?> outputKeyClass,
121       Class<?> outputValueClass, Job job)
122   throws IOException {
123       initTableMapperJob(Bytes.toString(table), scan, mapper, outputKeyClass, outputValueClass,
124               job, true);
125   }
126 
127    /**
128     * Use this before submitting a TableMap job. It will appropriately set up
129     * the job.
130     *
131     * @param table  The table name to read from.
132     * @param scan  The scan instance with the columns, time range etc.
133     * @param mapper  The mapper class to use.
134     * @param outputKeyClass  The class of the output key.
135     * @param outputValueClass  The class of the output value.
136     * @param job  The current job to adjust.  Make sure the passed job is
137     * carrying all necessary HBase configuration.
138     * @param addDependencyJars upload HBase jars and jars for any of the configured
139     *           job classes via the distributed cache (tmpjars).
140     * @throws IOException When setting up the details fails.
141     */
142    public static void initTableMapperJob(String table, Scan scan,
143        Class<? extends TableMapper> mapper,
144        Class<?> outputKeyClass,
145        Class<?> outputValueClass, Job job,
146        boolean addDependencyJars, Class<? extends InputFormat> inputFormatClass)
147    throws IOException {
148      initTableMapperJob(table, scan, mapper, outputKeyClass, outputValueClass, job,
149          addDependencyJars, true, inputFormatClass);
150    }
151 
152 
153   /**
154    * Use this before submitting a TableMap job. It will appropriately set up
155    * the job.
156    *
157    * @param table  The table name to read from.
158    * @param scan  The scan instance with the columns, time range etc.
159    * @param mapper  The mapper class to use.
160    * @param outputKeyClass  The class of the output key.
161    * @param outputValueClass  The class of the output value.
162    * @param job  The current job to adjust.  Make sure the passed job is
163    * carrying all necessary HBase configuration.
164    * @param addDependencyJars upload HBase jars and jars for any of the configured
165    *           job classes via the distributed cache (tmpjars).
166    * @param initCredentials whether to initialize hbase auth credentials for the job
167    * @param inputFormatClass the input format
168    * @throws IOException When setting up the details fails.
169    */
170   public static void initTableMapperJob(String table, Scan scan,
171       Class<? extends TableMapper> mapper,
172       Class<?> outputKeyClass,
173       Class<?> outputValueClass, Job job,
174       boolean addDependencyJars, boolean initCredentials,
175       Class<? extends InputFormat> inputFormatClass)
176   throws IOException {
177     job.setInputFormatClass(inputFormatClass);
178     if (outputValueClass != null) job.setMapOutputValueClass(outputValueClass);
179     if (outputKeyClass != null) job.setMapOutputKeyClass(outputKeyClass);
180     job.setMapperClass(mapper);
181     if (Put.class.equals(outputValueClass)) {
182       job.setCombinerClass(PutCombiner.class);
183     }
184     Configuration conf = job.getConfiguration();
185     HBaseConfiguration.merge(conf, HBaseConfiguration.create(conf));
186     conf.set(TableInputFormat.INPUT_TABLE, table);
187     conf.set(TableInputFormat.SCAN, convertScanToString(scan));
188     conf.setStrings("io.serializations", conf.get("io.serializations"),
189         MutationSerialization.class.getName(), ResultSerialization.class.getName(),
190         KeyValueSerialization.class.getName());
191     if (addDependencyJars) {
192       addDependencyJars(job);
193     }
194     if (initCredentials) {
195       initCredentials(job);
196     }
197   }
198 
199   /**
200    * Use this before submitting a TableMap job. It will appropriately set up
201    * the job.
202    *
203    * @param table Binary representation of the table name to read from.
204    * @param scan  The scan instance with the columns, time range etc.
205    * @param mapper  The mapper class to use.
206    * @param outputKeyClass  The class of the output key.
207    * @param outputValueClass  The class of the output value.
208    * @param job  The current job to adjust.  Make sure the passed job is
209    * carrying all necessary HBase configuration.
210    * @param addDependencyJars upload HBase jars and jars for any of the configured
211    *           job classes via the distributed cache (tmpjars).
212    * @param inputFormatClass The class of the input format
213    * @throws IOException When setting up the details fails.
214    */
215   public static void initTableMapperJob(byte[] table, Scan scan,
216       Class<? extends TableMapper> mapper,
217       Class<?> outputKeyClass,
218       Class<?> outputValueClass, Job job,
219       boolean addDependencyJars, Class<? extends InputFormat> inputFormatClass)
220   throws IOException {
221       initTableMapperJob(Bytes.toString(table), scan, mapper, outputKeyClass,
222               outputValueClass, job, addDependencyJars, inputFormatClass);
223   }
224 
225   /**
226    * Use this before submitting a TableMap job. It will appropriately set up
227    * the job.
228    *
229    * @param table Binary representation of the table name to read from.
230    * @param scan  The scan instance with the columns, time range etc.
231    * @param mapper  The mapper class to use.
232    * @param outputKeyClass  The class of the output key.
233    * @param outputValueClass  The class of the output value.
234    * @param job  The current job to adjust.  Make sure the passed job is
235    * carrying all necessary HBase configuration.
236    * @param addDependencyJars upload HBase jars and jars for any of the configured
237    *           job classes via the distributed cache (tmpjars).
238    * @throws IOException When setting up the details fails.
239    */
240   public static void initTableMapperJob(byte[] table, Scan scan,
241       Class<? extends TableMapper> mapper,
242       Class<?> outputKeyClass,
243       Class<?> outputValueClass, Job job,
244       boolean addDependencyJars)
245   throws IOException {
246       initTableMapperJob(Bytes.toString(table), scan, mapper, outputKeyClass,
247               outputValueClass, job, addDependencyJars, TableInputFormat.class);
248   }
249 
250   /**
251    * Use this before submitting a TableMap job. It will appropriately set up
252    * the job.
253    *
254    * @param table The table name to read from.
255    * @param scan  The scan instance with the columns, time range etc.
256    * @param mapper  The mapper class to use.
257    * @param outputKeyClass  The class of the output key.
258    * @param outputValueClass  The class of the output value.
259    * @param job  The current job to adjust.  Make sure the passed job is
260    * carrying all necessary HBase configuration.
261    * @param addDependencyJars upload HBase jars and jars for any of the configured
262    *           job classes via the distributed cache (tmpjars).
263    * @throws IOException When setting up the details fails.
264    */
265   public static void initTableMapperJob(String table, Scan scan,
266       Class<? extends TableMapper> mapper,
267       Class<?> outputKeyClass,
268       Class<?> outputValueClass, Job job,
269       boolean addDependencyJars)
270   throws IOException {
271       initTableMapperJob(table, scan, mapper, outputKeyClass,
272               outputValueClass, job, addDependencyJars, TableInputFormat.class);
273   }
274 
275   /**
276    * Sets up the job for reading from a table snapshot. It bypasses hbase servers
277    * and read directly from snapshot files.
278    *
279    * @param snapshotName The name of the snapshot (of a table) to read from.
280    * @param scan  The scan instance with the columns, time range etc.
281    * @param mapper  The mapper class to use.
282    * @param outputKeyClass  The class of the output key.
283    * @param outputValueClass  The class of the output value.
284    * @param job  The current job to adjust.  Make sure the passed job is
285    * carrying all necessary HBase configuration.
286    * @param addDependencyJars upload HBase jars and jars for any of the configured
287    *           job classes via the distributed cache (tmpjars).
288    *
289    * @param tmpRestoreDir a temporary directory to copy the snapshot files into. Current user should
290    * have write permissions to this directory, and this should not be a subdirectory of rootdir.
291    * After the job is finished, restore directory can be deleted.
292    * @throws IOException When setting up the details fails.
293    * @see TableSnapshotInputFormat
294    */
295   public static void initTableSnapshotMapperJob(String snapshotName, Scan scan,
296       Class<? extends TableMapper> mapper,
297       Class<?> outputKeyClass,
298       Class<?> outputValueClass, Job job,
299       boolean addDependencyJars, Path tmpRestoreDir)
300   throws IOException {
301     TableSnapshotInputFormat.setInput(job, snapshotName, tmpRestoreDir);
302     initTableMapperJob(snapshotName, scan, mapper, outputKeyClass,
303         outputValueClass, job, addDependencyJars, false, TableSnapshotInputFormat.class);
304 
305     /*
306      * Enable a basic on-heap cache for these jobs. Any BlockCache implementation based on
307      * direct memory will likely cause the map tasks to OOM when opening the region. This
308      * is done here instead of in TableSnapshotRegionRecordReader in case an advanced user
309      * wants to override this behavior in their job.
310      */
311     job.getConfiguration().setFloat(
312       HConstants.HFILE_BLOCK_CACHE_SIZE_KEY, HConstants.HFILE_BLOCK_CACHE_SIZE_DEFAULT);
313     job.getConfiguration().setFloat("hbase.offheapcache.percentage", 0f);
314     job.getConfiguration().setFloat("hbase.bucketcache.size", 0f);
315 
316     // We would need even more libraries that hbase-server depends on
317     TableMapReduceUtil.addDependencyJars(job.getConfiguration(), Counter.class);
318   }
319 
320   /**
321    * Use this before submitting a Multi TableMap job. It will appropriately set
322    * up the job.
323    *
324    * @param scans The list of {@link Scan} objects to read from.
325    * @param mapper The mapper class to use.
326    * @param outputKeyClass The class of the output key.
327    * @param outputValueClass The class of the output value.
328    * @param job The current job to adjust. Make sure the passed job is carrying
329    *          all necessary HBase configuration.
330    * @throws IOException When setting up the details fails.
331    */
332   public static void initTableMapperJob(List<Scan> scans,
333       Class<? extends TableMapper> mapper,
334       Class<? extends WritableComparable> outputKeyClass,
335       Class<? extends Writable> outputValueClass, Job job) throws IOException {
336     initTableMapperJob(scans, mapper, outputKeyClass, outputValueClass, job,
337         true);
338   }
339 
340   /**
341    * Use this before submitting a Multi TableMap job. It will appropriately set
342    * up the job.
343    *
344    * @param scans The list of {@link Scan} objects to read from.
345    * @param mapper The mapper class to use.
346    * @param outputKeyClass The class of the output key.
347    * @param outputValueClass The class of the output value.
348    * @param job The current job to adjust. Make sure the passed job is carrying
349    *          all necessary HBase configuration.
350    * @param addDependencyJars upload HBase jars and jars for any of the
351    *          configured job classes via the distributed cache (tmpjars).
352    * @throws IOException When setting up the details fails.
353    */
354   public static void initTableMapperJob(List<Scan> scans,
355       Class<? extends TableMapper> mapper,
356       Class<? extends WritableComparable> outputKeyClass,
357       Class<? extends Writable> outputValueClass, Job job,
358       boolean addDependencyJars) throws IOException {
359     job.setInputFormatClass(MultiTableInputFormat.class);
360     if (outputValueClass != null) {
361       job.setMapOutputValueClass(outputValueClass);
362     }
363     if (outputKeyClass != null) {
364       job.setMapOutputKeyClass(outputKeyClass);
365     }
366     job.setMapperClass(mapper);
367     HBaseConfiguration.addHbaseResources(job.getConfiguration());
368     List<String> scanStrings = new ArrayList<String>();
369 
370     for (Scan scan : scans) {
371       scanStrings.add(convertScanToString(scan));
372     }
373     job.getConfiguration().setStrings(MultiTableInputFormat.SCANS,
374       scanStrings.toArray(new String[scanStrings.size()]));
375 
376     if (addDependencyJars) {
377       addDependencyJars(job);
378     }
379   }
380 
381   public static void initCredentials(Job job) throws IOException {
382     UserProvider userProvider = UserProvider.instantiate(job.getConfiguration());
383     if (userProvider.isHadoopSecurityEnabled()) {
384       // propagate delegation related props from launcher job to MR job
385       if (System.getenv("HADOOP_TOKEN_FILE_LOCATION") != null) {
386         job.getConfiguration().set("mapreduce.job.credentials.binary",
387                                    System.getenv("HADOOP_TOKEN_FILE_LOCATION"));
388       }
389     }
390 
391     if (userProvider.isHBaseSecurityEnabled()) {
392       try {
393         // init credentials for remote cluster
394         String quorumAddress = job.getConfiguration().get(TableOutputFormat.QUORUM_ADDRESS);
395         User user = userProvider.getCurrent();
396         if (quorumAddress != null) {
397           Configuration peerConf = HBaseConfiguration.create(job.getConfiguration());
398           ZKUtil.applyClusterKeyToConf(peerConf, quorumAddress);
399           obtainAuthTokenForJob(job, peerConf, user);
400         }
401 
402         obtainAuthTokenForJob(job, job.getConfiguration(), user);
403       } catch (InterruptedException ie) {
404         LOG.info("Interrupted obtaining user authentication token");
405         Thread.currentThread().interrupt();
406       }
407     }
408   }
409 
410   /**
411    * Obtain an authentication token, for the specified cluster, on behalf of the current user
412    * and add it to the credentials for the given map reduce job.
413    *
414    * The quorumAddress is the key to the ZK ensemble, which contains:
415    * hbase.zookeeper.quorum, hbase.zookeeper.client.port and zookeeper.znode.parent
416    *
417    * @param job The job that requires the permission.
418    * @param quorumAddress string that contains the 3 required configuratins
419    * @throws IOException When the authentication token cannot be obtained.
420    */
421   public static void initCredentialsForCluster(Job job, String quorumAddress)
422       throws IOException {
423     UserProvider userProvider = UserProvider.instantiate(job.getConfiguration());
424     if (userProvider.isHBaseSecurityEnabled()) {
425       try {
426         Configuration peerConf = HBaseConfiguration.create(job.getConfiguration());
427         ZKUtil.applyClusterKeyToConf(peerConf, quorumAddress);
428         obtainAuthTokenForJob(job, peerConf, userProvider.getCurrent());
429       } catch (InterruptedException e) {
430         LOG.info("Interrupted obtaining user authentication token");
431         Thread.interrupted();
432       }
433     }
434   }
435 
436   private static void obtainAuthTokenForJob(Job job, Configuration conf, User user)
437       throws IOException, InterruptedException {
438     Token<AuthenticationTokenIdentifier> authToken = getAuthToken(conf, user);
439     if (authToken == null) {
440       user.obtainAuthTokenForJob(conf, job);
441     } else {
442       job.getCredentials().addToken(authToken.getService(), authToken);
443     }
444   }
445 
446   /**
447    * Get the authentication token of the user for the cluster specified in the configuration
448    * @return null if the user does not have the token, otherwise the auth token for the cluster.
449    */
450   private static Token<AuthenticationTokenIdentifier> getAuthToken(Configuration conf, User user)
451       throws IOException, InterruptedException {
452     ZooKeeperWatcher zkw = new ZooKeeperWatcher(conf, "mr-init-credentials", null);
453     try {
454       String clusterId = ZKClusterId.readClusterIdZNode(zkw);
455       return new AuthenticationTokenSelector().selectToken(new Text(clusterId), user.getUGI().getTokens());
456     } catch (KeeperException e) {
457       throw new IOException(e);
458     } finally {
459       zkw.close();
460     }
461   }
462 
463   /**
464    * Writes the given scan into a Base64 encoded string.
465    *
466    * @param scan  The scan to write out.
467    * @return The scan saved in a Base64 encoded string.
468    * @throws IOException When writing the scan fails.
469    */
470   static String convertScanToString(Scan scan) throws IOException {
471     ClientProtos.Scan proto = ProtobufUtil.toScan(scan);
472     return Base64.encodeBytes(proto.toByteArray());
473   }
474 
475   /**
476    * Converts the given Base64 string back into a Scan instance.
477    *
478    * @param base64  The scan details.
479    * @return The newly created Scan instance.
480    * @throws IOException When reading the scan instance fails.
481    */
482   static Scan convertStringToScan(String base64) throws IOException {
483     byte [] decoded = Base64.decode(base64);
484     ClientProtos.Scan scan;
485     try {
486       scan = ClientProtos.Scan.parseFrom(decoded);
487     } catch (InvalidProtocolBufferException ipbe) {
488       throw new IOException(ipbe);
489     }
490 
491     return ProtobufUtil.toScan(scan);
492   }
493 
494   /**
495    * Use this before submitting a TableReduce job. It will
496    * appropriately set up the JobConf.
497    *
498    * @param table  The output table.
499    * @param reducer  The reducer class to use.
500    * @param job  The current job to adjust.
501    * @throws IOException When determining the region count fails.
502    */
503   public static void initTableReducerJob(String table,
504     Class<? extends TableReducer> reducer, Job job)
505   throws IOException {
506     initTableReducerJob(table, reducer, job, null);
507   }
508 
509   /**
510    * Use this before submitting a TableReduce job. It will
511    * appropriately set up the JobConf.
512    *
513    * @param table  The output table.
514    * @param reducer  The reducer class to use.
515    * @param job  The current job to adjust.
516    * @param partitioner  Partitioner to use. Pass <code>null</code> to use
517    * default partitioner.
518    * @throws IOException When determining the region count fails.
519    */
520   public static void initTableReducerJob(String table,
521     Class<? extends TableReducer> reducer, Job job,
522     Class partitioner) throws IOException {
523     initTableReducerJob(table, reducer, job, partitioner, null, null, null);
524   }
525 
526   /**
527    * Use this before submitting a TableReduce job. It will
528    * appropriately set up the JobConf.
529    *
530    * @param table  The output table.
531    * @param reducer  The reducer class to use.
532    * @param job  The current job to adjust.  Make sure the passed job is
533    * carrying all necessary HBase configuration.
534    * @param partitioner  Partitioner to use. Pass <code>null</code> to use
535    * default partitioner.
536    * @param quorumAddress Distant cluster to write to; default is null for
537    * output to the cluster that is designated in <code>hbase-site.xml</code>.
538    * Set this String to the zookeeper ensemble of an alternate remote cluster
539    * when you would have the reduce write a cluster that is other than the
540    * default; e.g. copying tables between clusters, the source would be
541    * designated by <code>hbase-site.xml</code> and this param would have the
542    * ensemble address of the remote cluster.  The format to pass is particular.
543    * Pass <code> &lt;hbase.zookeeper.quorum>:&lt;hbase.zookeeper.client.port>:&lt;zookeeper.znode.parent>
544    * </code> such as <code>server,server2,server3:2181:/hbase</code>.
545    * @param serverClass redefined hbase.regionserver.class
546    * @param serverImpl redefined hbase.regionserver.impl
547    * @throws IOException When determining the region count fails.
548    */
549   public static void initTableReducerJob(String table,
550     Class<? extends TableReducer> reducer, Job job,
551     Class partitioner, String quorumAddress, String serverClass,
552     String serverImpl) throws IOException {
553     initTableReducerJob(table, reducer, job, partitioner, quorumAddress,
554         serverClass, serverImpl, true);
555   }
556 
557   /**
558    * Use this before submitting a TableReduce job. It will
559    * appropriately set up the JobConf.
560    *
561    * @param table  The output table.
562    * @param reducer  The reducer class to use.
563    * @param job  The current job to adjust.  Make sure the passed job is
564    * carrying all necessary HBase configuration.
565    * @param partitioner  Partitioner to use. Pass <code>null</code> to use
566    * default partitioner.
567    * @param quorumAddress Distant cluster to write to; default is null for
568    * output to the cluster that is designated in <code>hbase-site.xml</code>.
569    * Set this String to the zookeeper ensemble of an alternate remote cluster
570    * when you would have the reduce write a cluster that is other than the
571    * default; e.g. copying tables between clusters, the source would be
572    * designated by <code>hbase-site.xml</code> and this param would have the
573    * ensemble address of the remote cluster.  The format to pass is particular.
574    * Pass <code> &lt;hbase.zookeeper.quorum>:&lt;hbase.zookeeper.client.port>:&lt;zookeeper.znode.parent>
575    * </code> such as <code>server,server2,server3:2181:/hbase</code>.
576    * @param serverClass redefined hbase.regionserver.class
577    * @param serverImpl redefined hbase.regionserver.impl
578    * @param addDependencyJars upload HBase jars and jars for any of the configured
579    *           job classes via the distributed cache (tmpjars).
580    * @throws IOException When determining the region count fails.
581    */
582   public static void initTableReducerJob(String table,
583     Class<? extends TableReducer> reducer, Job job,
584     Class partitioner, String quorumAddress, String serverClass,
585     String serverImpl, boolean addDependencyJars) throws IOException {
586 
587     Configuration conf = job.getConfiguration();
588     HBaseConfiguration.merge(conf, HBaseConfiguration.create(conf));
589     job.setOutputFormatClass(TableOutputFormat.class);
590     if (reducer != null) job.setReducerClass(reducer);
591     conf.set(TableOutputFormat.OUTPUT_TABLE, table);
592     conf.setStrings("io.serializations", conf.get("io.serializations"),
593         MutationSerialization.class.getName(), ResultSerialization.class.getName());
594     // If passed a quorum/ensemble address, pass it on to TableOutputFormat.
595     if (quorumAddress != null) {
596       // Calling this will validate the format
597       ZKUtil.transformClusterKey(quorumAddress);
598       conf.set(TableOutputFormat.QUORUM_ADDRESS,quorumAddress);
599     }
600     if (serverClass != null && serverImpl != null) {
601       conf.set(TableOutputFormat.REGION_SERVER_CLASS, serverClass);
602       conf.set(TableOutputFormat.REGION_SERVER_IMPL, serverImpl);
603     }
604     job.setOutputKeyClass(ImmutableBytesWritable.class);
605     job.setOutputValueClass(Writable.class);
606     if (partitioner == HRegionPartitioner.class) {
607       job.setPartitionerClass(HRegionPartitioner.class);
608       int regions = MetaReader.getRegionCount(conf, table);
609       if (job.getNumReduceTasks() > regions) {
610         job.setNumReduceTasks(regions);
611       }
612     } else if (partitioner != null) {
613       job.setPartitionerClass(partitioner);
614     }
615 
616     if (addDependencyJars) {
617       addDependencyJars(job);
618     }
619 
620     initCredentials(job);
621   }
622 
623   /**
624    * Ensures that the given number of reduce tasks for the given job
625    * configuration does not exceed the number of regions for the given table.
626    *
627    * @param table  The table to get the region count for.
628    * @param job  The current job to adjust.
629    * @throws IOException When retrieving the table details fails.
630    */
631   public static void limitNumReduceTasks(String table, Job job)
632   throws IOException {
633     int regions = MetaReader.getRegionCount(job.getConfiguration(), table);
634     if (job.getNumReduceTasks() > regions)
635       job.setNumReduceTasks(regions);
636   }
637 
638   /**
639    * Sets the number of reduce tasks for the given job configuration to the
640    * number of regions the given table has.
641    *
642    * @param table  The table to get the region count for.
643    * @param job  The current job to adjust.
644    * @throws IOException When retrieving the table details fails.
645    */
646   public static void setNumReduceTasks(String table, Job job)
647   throws IOException {
648     job.setNumReduceTasks(MetaReader.getRegionCount(job.getConfiguration(), table));
649   }
650 
651   /**
652    * Sets the number of rows to return and cache with each scanner iteration.
653    * Higher caching values will enable faster mapreduce jobs at the expense of
654    * requiring more heap to contain the cached rows.
655    *
656    * @param job The current job to adjust.
657    * @param batchSize The number of rows to return in batch with each scanner
658    * iteration.
659    */
660   public static void setScannerCaching(Job job, int batchSize) {
661     job.getConfiguration().setInt("hbase.client.scanner.caching", batchSize);
662   }
663 
664   /**
665    * Add HBase and its dependencies (only) to the job configuration.
666    * <p>
667    * This is intended as a low-level API, facilitating code reuse between this
668    * class and its mapred counterpart. It also of use to extenral tools that
669    * need to build a MapReduce job that interacts with HBase but want
670    * fine-grained control over the jars shipped to the cluster.
671    * </p>
672    * @param conf The Configuration object to extend with dependencies.
673    * @see org.apache.hadoop.hbase.mapred.TableMapReduceUtil
674    * @see <a href="https://issues.apache.org/jira/browse/PIG-3285">PIG-3285</a>
675    */
676   public static void addHBaseDependencyJars(Configuration conf) throws IOException {
677     addDependencyJars(conf,
678       // explicitly pull a class from each module
679       org.apache.hadoop.hbase.HConstants.class,                      // hbase-common
680       org.apache.hadoop.hbase.protobuf.generated.ClientProtos.class, // hbase-protocol
681       org.apache.hadoop.hbase.client.Put.class,                      // hbase-client
682       org.apache.hadoop.hbase.CompatibilityFactory.class,            // hbase-hadoop-compat
683       org.apache.hadoop.hbase.mapreduce.TableMapper.class,           // hbase-server
684       // pull necessary dependencies
685       org.apache.zookeeper.ZooKeeper.class,
686       org.jboss.netty.channel.ChannelFactory.class,
687       com.google.protobuf.Message.class,
688       com.google.common.collect.Lists.class,
689       org.cloudera.htrace.Trace.class);
690   }
691 
692   /**
693    * Returns a classpath string built from the content of the "tmpjars" value in {@code conf}.
694    * Also exposed to shell scripts via `bin/hbase mapredcp`.
695    */
696   public static String buildDependencyClasspath(Configuration conf) {
697     if (conf == null) {
698       throw new IllegalArgumentException("Must provide a configuration object.");
699     }
700     Set<String> paths = new HashSet<String>(conf.getStringCollection("tmpjars"));
701     if (paths.size() == 0) {
702       throw new IllegalArgumentException("Configuration contains no tmpjars.");
703     }
704     StringBuilder sb = new StringBuilder();
705     for (String s : paths) {
706       // entries can take the form 'file:/path/to/file.jar'.
707       int idx = s.indexOf(":");
708       if (idx != -1) s = s.substring(idx + 1);
709       if (sb.length() > 0) sb.append(File.pathSeparator);
710       sb.append(s);
711     }
712     return sb.toString();
713   }
714 
715   /**
716    * Add the HBase dependency jars as well as jars for any of the configured
717    * job classes to the job configuration, so that JobClient will ship them
718    * to the cluster and add them to the DistributedCache.
719    */
720   public static void addDependencyJars(Job job) throws IOException {
721     addHBaseDependencyJars(job.getConfiguration());
722     try {
723       addDependencyJars(job.getConfiguration(),
724           // when making changes here, consider also mapred.TableMapReduceUtil
725           // pull job classes
726           job.getMapOutputKeyClass(),
727           job.getMapOutputValueClass(),
728           job.getInputFormatClass(),
729           job.getOutputKeyClass(),
730           job.getOutputValueClass(),
731           job.getOutputFormatClass(),
732           job.getPartitionerClass(),
733           job.getCombinerClass());
734     } catch (ClassNotFoundException e) {
735       throw new IOException(e);
736     }
737   }
738 
739   /**
740    * Add the jars containing the given classes to the job's configuration
741    * such that JobClient will ship them to the cluster and add them to
742    * the DistributedCache.
743    */
744   public static void addDependencyJars(Configuration conf,
745       Class<?>... classes) throws IOException {
746 
747     FileSystem localFs = FileSystem.getLocal(conf);
748     Set<String> jars = new HashSet<String>();
749     // Add jars that are already in the tmpjars variable
750     jars.addAll(conf.getStringCollection("tmpjars"));
751 
752     // add jars as we find them to a map of contents jar name so that we can avoid
753     // creating new jars for classes that have already been packaged.
754     Map<String, String> packagedClasses = new HashMap<String, String>();
755 
756     // Add jars containing the specified classes
757     for (Class<?> clazz : classes) {
758       if (clazz == null) continue;
759 
760       Path path = findOrCreateJar(clazz, localFs, packagedClasses);
761       if (path == null) {
762         LOG.warn("Could not find jar for class " + clazz +
763                  " in order to ship it to the cluster.");
764         continue;
765       }
766       if (!localFs.exists(path)) {
767         LOG.warn("Could not validate jar file " + path + " for class "
768                  + clazz);
769         continue;
770       }
771       jars.add(path.toString());
772     }
773     if (jars.isEmpty()) return;
774 
775     conf.set("tmpjars", StringUtils.arrayToString(jars.toArray(new String[jars.size()])));
776   }
777 
778   /**
779    * If org.apache.hadoop.util.JarFinder is available (0.23+ hadoop), finds
780    * the Jar for a class or creates it if it doesn't exist. If the class is in
781    * a directory in the classpath, it creates a Jar on the fly with the
782    * contents of the directory and returns the path to that Jar. If a Jar is
783    * created, it is created in the system temporary directory. Otherwise,
784    * returns an existing jar that contains a class of the same name. Maintains
785    * a mapping from jar contents to the tmp jar created.
786    * @param my_class the class to find.
787    * @param fs the FileSystem with which to qualify the returned path.
788    * @param packagedClasses a map of class name to path.
789    * @return a jar file that contains the class.
790    * @throws IOException
791    */
792   private static Path findOrCreateJar(Class<?> my_class, FileSystem fs,
793       Map<String, String> packagedClasses)
794   throws IOException {
795     // attempt to locate an existing jar for the class.
796     String jar = findContainingJar(my_class, packagedClasses);
797     if (null == jar || jar.isEmpty()) {
798       jar = getJar(my_class);
799       updateMap(jar, packagedClasses);
800     }
801 
802     if (null == jar || jar.isEmpty()) {
803       return null;
804     }
805 
806     LOG.debug(String.format("For class %s, using jar %s", my_class.getName(), jar));
807     return new Path(jar).makeQualified(fs);
808   }
809 
810   /**
811    * Add entries to <code>packagedClasses</code> corresponding to class files
812    * contained in <code>jar</code>.
813    * @param jar The jar who's content to list.
814    * @param packagedClasses map[class -> jar]
815    */
816   private static void updateMap(String jar, Map<String, String> packagedClasses) throws IOException {
817     if (null == jar || jar.isEmpty()) {
818       return;
819     }
820     ZipFile zip = null;
821     try {
822       zip = new ZipFile(jar);
823       for (Enumeration<? extends ZipEntry> iter = zip.entries(); iter.hasMoreElements();) {
824         ZipEntry entry = iter.nextElement();
825         if (entry.getName().endsWith("class")) {
826           packagedClasses.put(entry.getName(), jar);
827         }
828       }
829     } finally {
830       if (null != zip) zip.close();
831     }
832   }
833 
834   /**
835    * Find a jar that contains a class of the same name, if any. It will return
836    * a jar file, even if that is not the first thing on the class path that
837    * has a class with the same name. Looks first on the classpath and then in
838    * the <code>packagedClasses</code> map.
839    * @param my_class the class to find.
840    * @return a jar file that contains the class, or null.
841    * @throws IOException
842    */
843   private static String findContainingJar(Class<?> my_class, Map<String, String> packagedClasses)
844       throws IOException {
845     ClassLoader loader = my_class.getClassLoader();
846     String class_file = my_class.getName().replaceAll("\\.", "/") + ".class";
847 
848     // first search the classpath
849     for (Enumeration<URL> itr = loader.getResources(class_file); itr.hasMoreElements();) {
850       URL url = itr.nextElement();
851       if ("jar".equals(url.getProtocol())) {
852         String toReturn = url.getPath();
853         if (toReturn.startsWith("file:")) {
854           toReturn = toReturn.substring("file:".length());
855         }
856         // URLDecoder is a misnamed class, since it actually decodes
857         // x-www-form-urlencoded MIME type rather than actual
858         // URL encoding (which the file path has). Therefore it would
859         // decode +s to ' 's which is incorrect (spaces are actually
860         // either unencoded or encoded as "%20"). Replace +s first, so
861         // that they are kept sacred during the decoding process.
862         toReturn = toReturn.replaceAll("\\+", "%2B");
863         toReturn = URLDecoder.decode(toReturn, "UTF-8");
864         return toReturn.replaceAll("!.*$", "");
865       }
866     }
867 
868     // now look in any jars we've packaged using JarFinder. Returns null when
869     // no jar is found.
870     return packagedClasses.get(class_file);
871   }
872 
873   /**
874    * Invoke 'getJar' on a JarFinder implementation. Useful for some job
875    * configuration contexts (HBASE-8140) and also for testing on MRv2. First
876    * check if we have HADOOP-9426. Lacking that, fall back to the backport.
877    * @param my_class the class to find.
878    * @return a jar file that contains the class, or null.
879    */
880   private static String getJar(Class<?> my_class) {
881     String ret = null;
882     String hadoopJarFinder = "org.apache.hadoop.util.JarFinder";
883     Class<?> jarFinder = null;
884     try {
885       LOG.debug("Looking for " + hadoopJarFinder + ".");
886       jarFinder = Class.forName(hadoopJarFinder);
887       LOG.debug(hadoopJarFinder + " found.");
888       Method getJar = jarFinder.getMethod("getJar", Class.class);
889       ret = (String) getJar.invoke(null, my_class);
890     } catch (ClassNotFoundException e) {
891       LOG.debug("Using backported JarFinder.");
892       ret = JarFinder.getJar(my_class);
893     } catch (InvocationTargetException e) {
894       // function was properly called, but threw it's own exception. Unwrap it
895       // and pass it on.
896       throw new RuntimeException(e.getCause());
897     } catch (Exception e) {
898       // toss all other exceptions, related to reflection failure
899       throw new RuntimeException("getJar invocation failed.", e);
900     }
901 
902     return ret;
903   }
904 }