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    * Enable a basic on-heap cache for these jobs. Any BlockCache implementation based on
277    * direct memory will likely cause the map tasks to OOM when opening the region. This
278    * is done here instead of in TableSnapshotRegionRecordReader in case an advanced user
279    * wants to override this behavior in their job.
280    */
281   public static void resetCacheConfig(Configuration conf) {
282     conf.setFloat(
283       HConstants.HFILE_BLOCK_CACHE_SIZE_KEY, HConstants.HFILE_BLOCK_CACHE_SIZE_DEFAULT);
284     conf.setFloat("hbase.offheapcache.percentage", 0f);
285     conf.setFloat("hbase.bucketcache.size", 0f);
286   }
287 
288   /**
289    * Sets up the job for reading from a table snapshot. It bypasses hbase servers
290    * and read directly from snapshot files.
291    *
292    * @param snapshotName The name of the snapshot (of a table) to read from.
293    * @param scan  The scan instance with the columns, time range etc.
294    * @param mapper  The mapper class to use.
295    * @param outputKeyClass  The class of the output key.
296    * @param outputValueClass  The class of the output value.
297    * @param job  The current job to adjust.  Make sure the passed job is
298    * carrying all necessary HBase configuration.
299    * @param addDependencyJars upload HBase jars and jars for any of the configured
300    *           job classes via the distributed cache (tmpjars).
301    *
302    * @param tmpRestoreDir a temporary directory to copy the snapshot files into. Current user should
303    * have write permissions to this directory, and this should not be a subdirectory of rootdir.
304    * After the job is finished, restore directory can be deleted.
305    * @throws IOException When setting up the details fails.
306    * @see TableSnapshotInputFormat
307    */
308   public static void initTableSnapshotMapperJob(String snapshotName, Scan scan,
309       Class<? extends TableMapper> mapper,
310       Class<?> outputKeyClass,
311       Class<?> outputValueClass, Job job,
312       boolean addDependencyJars, Path tmpRestoreDir)
313   throws IOException {
314     TableSnapshotInputFormat.setInput(job, snapshotName, tmpRestoreDir);
315     initTableMapperJob(snapshotName, scan, mapper, outputKeyClass,
316         outputValueClass, job, addDependencyJars, false, TableSnapshotInputFormat.class);
317 
318     resetCacheConfig(job.getConfiguration());
319     // We would need even more libraries that hbase-server depends on
320     TableMapReduceUtil.addDependencyJars(job.getConfiguration(), Counter.class);
321   }
322 
323   /**
324    * Use this before submitting a Multi TableMap job. It will appropriately set
325    * up the job.
326    *
327    * @param scans The list of {@link Scan} objects to read from.
328    * @param mapper The mapper class to use.
329    * @param outputKeyClass The class of the output key.
330    * @param outputValueClass The class of the output value.
331    * @param job The current job to adjust. Make sure the passed job is carrying
332    *          all necessary HBase configuration.
333    * @throws IOException When setting up the details fails.
334    */
335   public static void initTableMapperJob(List<Scan> scans,
336       Class<? extends TableMapper> mapper,
337       Class<? extends WritableComparable> outputKeyClass,
338       Class<? extends Writable> outputValueClass, Job job) throws IOException {
339     initTableMapperJob(scans, mapper, outputKeyClass, outputValueClass, job,
340         true);
341   }
342 
343   /**
344    * Use this before submitting a Multi TableMap job. It will appropriately set
345    * up the job.
346    *
347    * @param scans The list of {@link Scan} objects to read from.
348    * @param mapper The mapper class to use.
349    * @param outputKeyClass The class of the output key.
350    * @param outputValueClass The class of the output value.
351    * @param job The current job to adjust. Make sure the passed job is carrying
352    *          all necessary HBase configuration.
353    * @param addDependencyJars upload HBase jars and jars for any of the
354    *          configured job classes via the distributed cache (tmpjars).
355    * @throws IOException When setting up the details fails.
356    */
357   public static void initTableMapperJob(List<Scan> scans,
358       Class<? extends TableMapper> mapper,
359       Class<? extends WritableComparable> outputKeyClass,
360       Class<? extends Writable> outputValueClass, Job job,
361       boolean addDependencyJars) throws IOException {
362     initTableMapperJob(scans, mapper, outputKeyClass, outputValueClass, job,
363       addDependencyJars, true);
364   }
365 
366   /**
367    * Use this before submitting a Multi TableMap job. It will appropriately set
368    * up the job.
369    *
370    * @param scans The list of {@link Scan} objects to read from.
371    * @param mapper The mapper class to use.
372    * @param outputKeyClass The class of the output key.
373    * @param outputValueClass The class of the output value.
374    * @param job The current job to adjust. Make sure the passed job is carrying
375    *          all necessary HBase configuration.
376    * @param addDependencyJars upload HBase jars and jars for any of the
377    *          configured job classes via the distributed cache (tmpjars).
378    * @param initCredentials whether to initialize hbase auth credentials for the job
379    * @throws IOException When setting up the details fails.
380    */
381   public static void initTableMapperJob(List<Scan> scans,
382       Class<? extends TableMapper> mapper,
383       Class<? extends WritableComparable> outputKeyClass,
384       Class<? extends Writable> outputValueClass, Job job,
385       boolean addDependencyJars, 
386       boolean initCredentials) throws IOException {
387     job.setInputFormatClass(MultiTableInputFormat.class);
388     if (outputValueClass != null) {
389       job.setMapOutputValueClass(outputValueClass);
390     }
391     if (outputKeyClass != null) {
392       job.setMapOutputKeyClass(outputKeyClass);
393     }
394     job.setMapperClass(mapper);
395     Configuration conf = job.getConfiguration();
396     HBaseConfiguration.merge(conf, HBaseConfiguration.create(conf));
397     List<String> scanStrings = new ArrayList<String>();
398 
399     for (Scan scan : scans) {
400       scanStrings.add(convertScanToString(scan));
401     }
402     job.getConfiguration().setStrings(MultiTableInputFormat.SCANS,
403       scanStrings.toArray(new String[scanStrings.size()]));
404 
405     if (addDependencyJars) {
406       addDependencyJars(job);
407     }
408 
409     if (initCredentials) {
410       initCredentials(job);
411     }
412   }
413 
414   public static void initCredentials(Job job) throws IOException {
415     UserProvider userProvider = UserProvider.instantiate(job.getConfiguration());
416     if (userProvider.isHadoopSecurityEnabled()) {
417       // propagate delegation related props from launcher job to MR job
418       if (System.getenv("HADOOP_TOKEN_FILE_LOCATION") != null) {
419         job.getConfiguration().set("mapreduce.job.credentials.binary",
420                                    System.getenv("HADOOP_TOKEN_FILE_LOCATION"));
421       }
422     }
423 
424     if (userProvider.isHBaseSecurityEnabled()) {
425       try {
426         // init credentials for remote cluster
427         String quorumAddress = job.getConfiguration().get(TableOutputFormat.QUORUM_ADDRESS);
428         User user = userProvider.getCurrent();
429         if (quorumAddress != null) {
430           Configuration peerConf = HBaseConfiguration.create(job.getConfiguration());
431           ZKUtil.applyClusterKeyToConf(peerConf, quorumAddress);
432           obtainAuthTokenForJob(job, peerConf, user);
433         }
434 
435         obtainAuthTokenForJob(job, job.getConfiguration(), user);
436       } catch (InterruptedException ie) {
437         LOG.info("Interrupted obtaining user authentication token");
438         Thread.currentThread().interrupt();
439       }
440     }
441   }
442 
443   /**
444    * Obtain an authentication token, for the specified cluster, on behalf of the current user
445    * and add it to the credentials for the given map reduce job.
446    *
447    * The quorumAddress is the key to the ZK ensemble, which contains:
448    * hbase.zookeeper.quorum, hbase.zookeeper.client.port and zookeeper.znode.parent
449    *
450    * @param job The job that requires the permission.
451    * @param quorumAddress string that contains the 3 required configuratins
452    * @throws IOException When the authentication token cannot be obtained.
453    */
454   public static void initCredentialsForCluster(Job job, String quorumAddress)
455       throws IOException {
456     UserProvider userProvider = UserProvider.instantiate(job.getConfiguration());
457     if (userProvider.isHBaseSecurityEnabled()) {
458       try {
459         Configuration peerConf = HBaseConfiguration.create(job.getConfiguration());
460         ZKUtil.applyClusterKeyToConf(peerConf, quorumAddress);
461         obtainAuthTokenForJob(job, peerConf, userProvider.getCurrent());
462       } catch (InterruptedException e) {
463         LOG.info("Interrupted obtaining user authentication token");
464         Thread.interrupted();
465       }
466     }
467   }
468 
469   private static void obtainAuthTokenForJob(Job job, Configuration conf, User user)
470       throws IOException, InterruptedException {
471     Token<AuthenticationTokenIdentifier> authToken = getAuthToken(conf, user);
472     if (authToken == null) {
473       user.obtainAuthTokenForJob(conf, job);
474     } else {
475       job.getCredentials().addToken(authToken.getService(), authToken);
476     }
477   }
478 
479   /**
480    * Get the authentication token of the user for the cluster specified in the configuration
481    * @return null if the user does not have the token, otherwise the auth token for the cluster.
482    */
483   private static Token<AuthenticationTokenIdentifier> getAuthToken(Configuration conf, User user)
484       throws IOException, InterruptedException {
485     ZooKeeperWatcher zkw = new ZooKeeperWatcher(conf, "mr-init-credentials", null);
486     try {
487       String clusterId = ZKClusterId.readClusterIdZNode(zkw);
488       return new AuthenticationTokenSelector().selectToken(new Text(clusterId), user.getUGI().getTokens());
489     } catch (KeeperException e) {
490       throw new IOException(e);
491     } finally {
492       zkw.close();
493     }
494   }
495 
496   /**
497    * Writes the given scan into a Base64 encoded string.
498    *
499    * @param scan  The scan to write out.
500    * @return The scan saved in a Base64 encoded string.
501    * @throws IOException When writing the scan fails.
502    */
503   static String convertScanToString(Scan scan) throws IOException {
504     ClientProtos.Scan proto = ProtobufUtil.toScan(scan);
505     return Base64.encodeBytes(proto.toByteArray());
506   }
507 
508   /**
509    * Converts the given Base64 string back into a Scan instance.
510    *
511    * @param base64  The scan details.
512    * @return The newly created Scan instance.
513    * @throws IOException When reading the scan instance fails.
514    */
515   static Scan convertStringToScan(String base64) throws IOException {
516     byte [] decoded = Base64.decode(base64);
517     ClientProtos.Scan scan;
518     try {
519       scan = ClientProtos.Scan.parseFrom(decoded);
520     } catch (InvalidProtocolBufferException ipbe) {
521       throw new IOException(ipbe);
522     }
523 
524     return ProtobufUtil.toScan(scan);
525   }
526 
527   /**
528    * Use this before submitting a TableReduce job. It will
529    * appropriately set up the JobConf.
530    *
531    * @param table  The output table.
532    * @param reducer  The reducer class to use.
533    * @param job  The current job to adjust.
534    * @throws IOException When determining the region count fails.
535    */
536   public static void initTableReducerJob(String table,
537     Class<? extends TableReducer> reducer, Job job)
538   throws IOException {
539     initTableReducerJob(table, reducer, job, null);
540   }
541 
542   /**
543    * Use this before submitting a TableReduce job. It will
544    * appropriately set up the JobConf.
545    *
546    * @param table  The output table.
547    * @param reducer  The reducer class to use.
548    * @param job  The current job to adjust.
549    * @param partitioner  Partitioner to use. Pass <code>null</code> to use
550    * default partitioner.
551    * @throws IOException When determining the region count fails.
552    */
553   public static void initTableReducerJob(String table,
554     Class<? extends TableReducer> reducer, Job job,
555     Class partitioner) throws IOException {
556     initTableReducerJob(table, reducer, job, partitioner, null, null, null);
557   }
558 
559   /**
560    * Use this before submitting a TableReduce job. It will
561    * appropriately set up the JobConf.
562    *
563    * @param table  The output table.
564    * @param reducer  The reducer class to use.
565    * @param job  The current job to adjust.  Make sure the passed job is
566    * carrying all necessary HBase configuration.
567    * @param partitioner  Partitioner to use. Pass <code>null</code> to use
568    * default partitioner.
569    * @param quorumAddress Distant cluster to write to; default is null for
570    * output to the cluster that is designated in <code>hbase-site.xml</code>.
571    * Set this String to the zookeeper ensemble of an alternate remote cluster
572    * when you would have the reduce write a cluster that is other than the
573    * default; e.g. copying tables between clusters, the source would be
574    * designated by <code>hbase-site.xml</code> and this param would have the
575    * ensemble address of the remote cluster.  The format to pass is particular.
576    * Pass <code> &lt;hbase.zookeeper.quorum>:&lt;hbase.zookeeper.client.port>:&lt;zookeeper.znode.parent>
577    * </code> such as <code>server,server2,server3:2181:/hbase</code>.
578    * @param serverClass redefined hbase.regionserver.class
579    * @param serverImpl redefined hbase.regionserver.impl
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) throws IOException {
586     initTableReducerJob(table, reducer, job, partitioner, quorumAddress,
587         serverClass, serverImpl, true);
588   }
589 
590   /**
591    * Use this before submitting a TableReduce job. It will
592    * appropriately set up the JobConf.
593    *
594    * @param table  The output table.
595    * @param reducer  The reducer class to use.
596    * @param job  The current job to adjust.  Make sure the passed job is
597    * carrying all necessary HBase configuration.
598    * @param partitioner  Partitioner to use. Pass <code>null</code> to use
599    * default partitioner.
600    * @param quorumAddress Distant cluster to write to; default is null for
601    * output to the cluster that is designated in <code>hbase-site.xml</code>.
602    * Set this String to the zookeeper ensemble of an alternate remote cluster
603    * when you would have the reduce write a cluster that is other than the
604    * default; e.g. copying tables between clusters, the source would be
605    * designated by <code>hbase-site.xml</code> and this param would have the
606    * ensemble address of the remote cluster.  The format to pass is particular.
607    * Pass <code> &lt;hbase.zookeeper.quorum>:&lt;hbase.zookeeper.client.port>:&lt;zookeeper.znode.parent>
608    * </code> such as <code>server,server2,server3:2181:/hbase</code>.
609    * @param serverClass redefined hbase.regionserver.class
610    * @param serverImpl redefined hbase.regionserver.impl
611    * @param addDependencyJars upload HBase jars and jars for any of the configured
612    *           job classes via the distributed cache (tmpjars).
613    * @throws IOException When determining the region count fails.
614    */
615   public static void initTableReducerJob(String table,
616     Class<? extends TableReducer> reducer, Job job,
617     Class partitioner, String quorumAddress, String serverClass,
618     String serverImpl, boolean addDependencyJars) throws IOException {
619 
620     Configuration conf = job.getConfiguration();
621     HBaseConfiguration.merge(conf, HBaseConfiguration.create(conf));
622     job.setOutputFormatClass(TableOutputFormat.class);
623     if (reducer != null) job.setReducerClass(reducer);
624     conf.set(TableOutputFormat.OUTPUT_TABLE, table);
625     conf.setStrings("io.serializations", conf.get("io.serializations"),
626         MutationSerialization.class.getName(), ResultSerialization.class.getName());
627     // If passed a quorum/ensemble address, pass it on to TableOutputFormat.
628     if (quorumAddress != null) {
629       // Calling this will validate the format
630       ZKUtil.transformClusterKey(quorumAddress);
631       conf.set(TableOutputFormat.QUORUM_ADDRESS,quorumAddress);
632     }
633     if (serverClass != null && serverImpl != null) {
634       conf.set(TableOutputFormat.REGION_SERVER_CLASS, serverClass);
635       conf.set(TableOutputFormat.REGION_SERVER_IMPL, serverImpl);
636     }
637     job.setOutputKeyClass(ImmutableBytesWritable.class);
638     job.setOutputValueClass(Writable.class);
639     if (partitioner == HRegionPartitioner.class) {
640       job.setPartitionerClass(HRegionPartitioner.class);
641       int regions = MetaReader.getRegionCount(conf, table);
642       if (job.getNumReduceTasks() > regions) {
643         job.setNumReduceTasks(regions);
644       }
645     } else if (partitioner != null) {
646       job.setPartitionerClass(partitioner);
647     }
648 
649     if (addDependencyJars) {
650       addDependencyJars(job);
651     }
652 
653     initCredentials(job);
654   }
655 
656   /**
657    * Ensures that the given number of reduce tasks for the given job
658    * configuration does not exceed the number of regions for the given table.
659    *
660    * @param table  The table to get the region count for.
661    * @param job  The current job to adjust.
662    * @throws IOException When retrieving the table details fails.
663    */
664   public static void limitNumReduceTasks(String table, Job job)
665   throws IOException {
666     int regions = MetaReader.getRegionCount(job.getConfiguration(), table);
667     if (job.getNumReduceTasks() > regions)
668       job.setNumReduceTasks(regions);
669   }
670 
671   /**
672    * Sets the number of reduce tasks for the given job configuration to the
673    * number of regions the given table has.
674    *
675    * @param table  The table to get the region count for.
676    * @param job  The current job to adjust.
677    * @throws IOException When retrieving the table details fails.
678    */
679   public static void setNumReduceTasks(String table, Job job)
680   throws IOException {
681     job.setNumReduceTasks(MetaReader.getRegionCount(job.getConfiguration(), table));
682   }
683 
684   /**
685    * Sets the number of rows to return and cache with each scanner iteration.
686    * Higher caching values will enable faster mapreduce jobs at the expense of
687    * requiring more heap to contain the cached rows.
688    *
689    * @param job The current job to adjust.
690    * @param batchSize The number of rows to return in batch with each scanner
691    * iteration.
692    */
693   public static void setScannerCaching(Job job, int batchSize) {
694     job.getConfiguration().setInt("hbase.client.scanner.caching", batchSize);
695   }
696 
697   /**
698    * Add HBase and its dependencies (only) to the job configuration.
699    * <p>
700    * This is intended as a low-level API, facilitating code reuse between this
701    * class and its mapred counterpart. It also of use to extenral tools that
702    * need to build a MapReduce job that interacts with HBase but want
703    * fine-grained control over the jars shipped to the cluster.
704    * </p>
705    * @param conf The Configuration object to extend with dependencies.
706    * @see org.apache.hadoop.hbase.mapred.TableMapReduceUtil
707    * @see <a href="https://issues.apache.org/jira/browse/PIG-3285">PIG-3285</a>
708    */
709   public static void addHBaseDependencyJars(Configuration conf) throws IOException {
710     addDependencyJars(conf,
711       // explicitly pull a class from each module
712       org.apache.hadoop.hbase.HConstants.class,                      // hbase-common
713       org.apache.hadoop.hbase.protobuf.generated.ClientProtos.class, // hbase-protocol
714       org.apache.hadoop.hbase.client.Put.class,                      // hbase-client
715       org.apache.hadoop.hbase.CompatibilityFactory.class,            // hbase-hadoop-compat
716       org.apache.hadoop.hbase.mapreduce.TableMapper.class,           // hbase-server
717       // pull necessary dependencies
718       org.apache.zookeeper.ZooKeeper.class,
719       org.jboss.netty.channel.ChannelFactory.class,
720       com.google.protobuf.Message.class,
721       com.google.common.collect.Lists.class,
722       org.cloudera.htrace.Trace.class);
723   }
724 
725   /**
726    * Returns a classpath string built from the content of the "tmpjars" value in {@code conf}.
727    * Also exposed to shell scripts via `bin/hbase mapredcp`.
728    */
729   public static String buildDependencyClasspath(Configuration conf) {
730     if (conf == null) {
731       throw new IllegalArgumentException("Must provide a configuration object.");
732     }
733     Set<String> paths = new HashSet<String>(conf.getStringCollection("tmpjars"));
734     if (paths.size() == 0) {
735       throw new IllegalArgumentException("Configuration contains no tmpjars.");
736     }
737     StringBuilder sb = new StringBuilder();
738     for (String s : paths) {
739       // entries can take the form 'file:/path/to/file.jar'.
740       int idx = s.indexOf(":");
741       if (idx != -1) s = s.substring(idx + 1);
742       if (sb.length() > 0) sb.append(File.pathSeparator);
743       sb.append(s);
744     }
745     return sb.toString();
746   }
747 
748   /**
749    * Add the HBase dependency jars as well as jars for any of the configured
750    * job classes to the job configuration, so that JobClient will ship them
751    * to the cluster and add them to the DistributedCache.
752    */
753   public static void addDependencyJars(Job job) throws IOException {
754     addHBaseDependencyJars(job.getConfiguration());
755     try {
756       addDependencyJars(job.getConfiguration(),
757           // when making changes here, consider also mapred.TableMapReduceUtil
758           // pull job classes
759           job.getMapOutputKeyClass(),
760           job.getMapOutputValueClass(),
761           job.getInputFormatClass(),
762           job.getOutputKeyClass(),
763           job.getOutputValueClass(),
764           job.getOutputFormatClass(),
765           job.getPartitionerClass(),
766           job.getCombinerClass());
767     } catch (ClassNotFoundException e) {
768       throw new IOException(e);
769     }
770   }
771 
772   /**
773    * Add the jars containing the given classes to the job's configuration
774    * such that JobClient will ship them to the cluster and add them to
775    * the DistributedCache.
776    */
777   public static void addDependencyJars(Configuration conf,
778       Class<?>... classes) throws IOException {
779 
780     FileSystem localFs = FileSystem.getLocal(conf);
781     Set<String> jars = new HashSet<String>();
782     // Add jars that are already in the tmpjars variable
783     jars.addAll(conf.getStringCollection("tmpjars"));
784 
785     // add jars as we find them to a map of contents jar name so that we can avoid
786     // creating new jars for classes that have already been packaged.
787     Map<String, String> packagedClasses = new HashMap<String, String>();
788 
789     // Add jars containing the specified classes
790     for (Class<?> clazz : classes) {
791       if (clazz == null) continue;
792 
793       Path path = findOrCreateJar(clazz, localFs, packagedClasses);
794       if (path == null) {
795         LOG.warn("Could not find jar for class " + clazz +
796                  " in order to ship it to the cluster.");
797         continue;
798       }
799       if (!localFs.exists(path)) {
800         LOG.warn("Could not validate jar file " + path + " for class "
801                  + clazz);
802         continue;
803       }
804       jars.add(path.toString());
805     }
806     if (jars.isEmpty()) return;
807 
808     conf.set("tmpjars", StringUtils.arrayToString(jars.toArray(new String[jars.size()])));
809   }
810 
811   /**
812    * If org.apache.hadoop.util.JarFinder is available (0.23+ hadoop), finds
813    * the Jar for a class or creates it if it doesn't exist. If the class is in
814    * a directory in the classpath, it creates a Jar on the fly with the
815    * contents of the directory and returns the path to that Jar. If a Jar is
816    * created, it is created in the system temporary directory. Otherwise,
817    * returns an existing jar that contains a class of the same name. Maintains
818    * a mapping from jar contents to the tmp jar created.
819    * @param my_class the class to find.
820    * @param fs the FileSystem with which to qualify the returned path.
821    * @param packagedClasses a map of class name to path.
822    * @return a jar file that contains the class.
823    * @throws IOException
824    */
825   private static Path findOrCreateJar(Class<?> my_class, FileSystem fs,
826       Map<String, String> packagedClasses)
827   throws IOException {
828     // attempt to locate an existing jar for the class.
829     String jar = findContainingJar(my_class, packagedClasses);
830     if (null == jar || jar.isEmpty()) {
831       jar = getJar(my_class);
832       updateMap(jar, packagedClasses);
833     }
834 
835     if (null == jar || jar.isEmpty()) {
836       return null;
837     }
838 
839     LOG.debug(String.format("For class %s, using jar %s", my_class.getName(), jar));
840     return new Path(jar).makeQualified(fs);
841   }
842 
843   /**
844    * Add entries to <code>packagedClasses</code> corresponding to class files
845    * contained in <code>jar</code>.
846    * @param jar The jar who's content to list.
847    * @param packagedClasses map[class -> jar]
848    */
849   private static void updateMap(String jar, Map<String, String> packagedClasses) throws IOException {
850     if (null == jar || jar.isEmpty()) {
851       return;
852     }
853     ZipFile zip = null;
854     try {
855       zip = new ZipFile(jar);
856       for (Enumeration<? extends ZipEntry> iter = zip.entries(); iter.hasMoreElements();) {
857         ZipEntry entry = iter.nextElement();
858         if (entry.getName().endsWith("class")) {
859           packagedClasses.put(entry.getName(), jar);
860         }
861       }
862     } finally {
863       if (null != zip) zip.close();
864     }
865   }
866 
867   /**
868    * Find a jar that contains a class of the same name, if any. It will return
869    * a jar file, even if that is not the first thing on the class path that
870    * has a class with the same name. Looks first on the classpath and then in
871    * the <code>packagedClasses</code> map.
872    * @param my_class the class to find.
873    * @return a jar file that contains the class, or null.
874    * @throws IOException
875    */
876   private static String findContainingJar(Class<?> my_class, Map<String, String> packagedClasses)
877       throws IOException {
878     ClassLoader loader = my_class.getClassLoader();
879     String class_file = my_class.getName().replaceAll("\\.", "/") + ".class";
880 
881     // first search the classpath
882     for (Enumeration<URL> itr = loader.getResources(class_file); itr.hasMoreElements();) {
883       URL url = itr.nextElement();
884       if ("jar".equals(url.getProtocol())) {
885         String toReturn = url.getPath();
886         if (toReturn.startsWith("file:")) {
887           toReturn = toReturn.substring("file:".length());
888         }
889         // URLDecoder is a misnamed class, since it actually decodes
890         // x-www-form-urlencoded MIME type rather than actual
891         // URL encoding (which the file path has). Therefore it would
892         // decode +s to ' 's which is incorrect (spaces are actually
893         // either unencoded or encoded as "%20"). Replace +s first, so
894         // that they are kept sacred during the decoding process.
895         toReturn = toReturn.replaceAll("\\+", "%2B");
896         toReturn = URLDecoder.decode(toReturn, "UTF-8");
897         return toReturn.replaceAll("!.*$", "");
898       }
899     }
900 
901     // now look in any jars we've packaged using JarFinder. Returns null when
902     // no jar is found.
903     return packagedClasses.get(class_file);
904   }
905 
906   /**
907    * Invoke 'getJar' on a JarFinder implementation. Useful for some job
908    * configuration contexts (HBASE-8140) and also for testing on MRv2. First
909    * check if we have HADOOP-9426. Lacking that, fall back to the backport.
910    * @param my_class the class to find.
911    * @return a jar file that contains the class, or null.
912    */
913   private static String getJar(Class<?> my_class) {
914     String ret = null;
915     String hadoopJarFinder = "org.apache.hadoop.util.JarFinder";
916     Class<?> jarFinder = null;
917     try {
918       LOG.debug("Looking for " + hadoopJarFinder + ".");
919       jarFinder = Class.forName(hadoopJarFinder);
920       LOG.debug(hadoopJarFinder + " found.");
921       Method getJar = jarFinder.getMethod("getJar", Class.class);
922       ret = (String) getJar.invoke(null, my_class);
923     } catch (ClassNotFoundException e) {
924       LOG.debug("Using backported JarFinder.");
925       ret = JarFinder.getJar(my_class);
926     } catch (InvocationTargetException e) {
927       // function was properly called, but threw it's own exception. Unwrap it
928       // and pass it on.
929       throw new RuntimeException(e.getCause());
930     } catch (Exception e) {
931       // toss all other exceptions, related to reflection failure
932       throw new RuntimeException("getJar invocation failed.", e);
933     }
934 
935     return ret;
936   }
937 }