简介

Hadoop的HDFS集群非常容易出现机器与机器之间磁盘利用率不平衡的情况,比如集群中添
加新的数据节点。当HDFS出现不平衡状况的时候,将引发很多问题,比如MR程序无法很好
地利用本地计算的优势,机器之间无法达到更好的网络带宽使用率,机器磁盘无法利用等
等。可见,保证HDFS中的数据平衡是非常重要的。
在Hadoop中,包含一个Balancer程序,通过运行这个程序,可以使得HDFS集群达到一个平
衡的状态

Step1: balancer command

1
2
3
4
5
6
7
8
9
10
11
12
HDFS Balance
>bin/start-balancer.sh -threshold 15
>bin/stop-balancer.sh
>各个DataNode存储不均匀时使用
-不同机器之间
-同机器不同磁盘之间无法通过这种方法balance
> dfs.balance.bandwidthPerSec(限制balance带宽)

影响hadoop balance工具的几个参数:
-threshold 默认设置:10,参数取值范围:0-100,参数含义:判断集群是否平衡的目标参数,每一个 datanode 存储使用率和集群总存储使用率的差值都应该小于这个阀值 ,理论上,该参数设置的越小,整个集群就越平衡,但是在线上环境中,hadoop集群在进行balance时,还在并发的进行数据的写入和删除,所以有可能无法到达设定的平衡参数值。

dfs.balance.bandwidthPerSec 默认设置:10485761 M/S),参数含义:设置balance工具在运行中所能占用的带宽,设置的过大可能会造成mapred运行缓慢
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
/** <p>The balancer is a tool that balances disk space usage on an HDFS cluster 
* when some datanodes become full or when new empty nodes join the cluster.
* The tool is deployed as an application program that can be run by the
* cluster administrator on a live HDFS cluster while applications
* adding and deleting files.
*
* <p>SYNOPSIS
* <pre>
* To start:
* bin/start-balancer.sh [-threshold <threshold>]
* Example: bin/ start-balancer.sh
* start the balancer with a default threshold of 10%
* bin/ start-balancer.sh -threshold 5
* start the balancer with a threshold of 5%
* To stop:
* bin/ stop-balancer.sh
* </pre>
*
* <p>DESCRIPTION
* <p>The threshold parameter is a fraction in the range of (1%, 100%) with a
* default value of 10%. The threshold sets a target for whether the cluster
* is balanced. A cluster is balanced if for each datanode, the utilization
* of the node (ratio of used space at the node to total capacity of the node)
* differs from the utilization of the (ratio of used space in the cluster
* to total capacity of the cluster) by no more than the threshold value.
* The smaller the threshold, the more balanced a cluster will become.
* It takes more time to run the balancer for small threshold values.
* Also for a very small threshold the cluster may not be able to reach the
* balanced state when applications write and delete files concurrently.
*
* <p>The tool moves blocks from highly utilized datanodes to poorly
* utilized datanodes iteratively. In each iteration a datanode moves or
* receives no more than the lesser of 10G bytes or the threshold fraction
* of its capacity. Each iteration runs no more than 20 minutes.
* 每次移动不超过10G大小,每次移动不超过20分钟。
* At the end of each iteration, the balancer obtains updated datanodes
* information from the namenode.
*
* <p>A system property that limits the balancer's use of bandwidth is
* defined in the default configuration file:
* <pre>
* <property>
* <name>dfs.balance.bandwidthPerSec</name>
* <value>1048576</value>
* <description> Specifies the maximum bandwidth that each datanode
* can utilize for the balancing purpose in term of the number of bytes
* per second. </description>
* </property>
* </pre>
*
* <p>This property determines the maximum speed at which a block will be
* moved from one datanode to another. The default value is 1MB/s. The higher
* the bandwidth, the faster a cluster can reach the balanced state,
* but with greater competition with application processes. If an
* administrator changes the value of this property in the configuration
* file, the change is observed when HDFS is next restarted.
*
* <p>MONITERING BALANCER PROGRESS
* <p>After the balancer is started, an output file name where the balancer
* progress will be recorded is printed on the screen. The administrator
* can monitor the running of the balancer by reading the output file.
* The output shows the balancer's status iteration by iteration. In each
* iteration it prints the starting time, the iteration number, the total
* number of bytes that have been moved in the previous iterations,
* the total number of bytes that are left to move in order for the cluster
* to be balanced, and the number of bytes that are being moved in this
* iteration. Normally "Bytes Already Moved" is increasing while "Bytes Left
* To Move" is decreasing.

*
* <p>Running multiple instances of the balancer in an HDFS cluster is
* prohibited by the tool.
*
* <p>The balancer automatically exits when any of the following five
* conditions is satisfied:
* <ol>
* <li>The cluster is balanced;
* <li>No block can be moved;
* <li>No block has been moved for five consecutive(连续) iterations;
* <li>An IOException occurs while communicating with the namenode;
* <li>Another balancer is running.
* </ol>
* 下面5种情况会导致Balance操作的失败
* 1、整个集群已经达到平衡状态
* 2、经过计算发现没有可以被移动的block块
* 3、在连续5次的迭代中,没有block块被移动
* 4、当datanode节点与namenode节点通信的时候,发生IO异常
* 5、已经存在一个Balance操作
*
* <p>Upon exit, a balancer returns an exit code and prints one of the
* following messages to the output file in corresponding to the above exit
* reasons:
* <ol>
* <li>The cluster is balanced. Exiting
* <li>No block can be moved. Exiting...
* <li>No block has been moved for 5 iterations. Exiting...
* <li>Received an IO exception: failure reason. Exiting...
* <li>Another balancer is running. Exiting...
* </ol>
* 在下面的5种情况下,balancer操作会自动退出
* 1、整个集群已经达到平衡的状态
* 2、经过计算发现没有可以被移动block块
* 3、在5论的迭代没有block被移动
* 4、接收端发生了IO异常
* 5、已经存在一个balanr进程在工作
*
* <p>The administrator can interrupt the execution of the balancer at any
* time by running the command "stop-balancer.sh" on the machine where the
* balancer is running.
*/

Step2: balancer 源码

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
/**
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

package org.apache.hadoop.hdfs.server.balancer;

import static com.google.common.base.Preconditions.checkArgument;

import java.io.IOException;
import java.io.PrintStream;
import java.net.URI;
import java.text.DateFormat;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.Date;
import java.util.Iterator;
import java.util.LinkedList;
import java.util.List;
import java.util.Set;

import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.classification.InterfaceAudience;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.conf.Configured;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.fs.StorageType;
import org.apache.hadoop.hdfs.DFSConfigKeys;
import org.apache.hadoop.hdfs.DFSUtil;
import org.apache.hadoop.hdfs.HdfsConfiguration;
import org.apache.hadoop.hdfs.server.balancer.Dispatcher.DDatanode;
import org.apache.hadoop.hdfs.server.balancer.Dispatcher.DDatanode.StorageGroup;
import org.apache.hadoop.hdfs.server.balancer.Dispatcher.Source;
import org.apache.hadoop.hdfs.server.balancer.Dispatcher.Task;
import org.apache.hadoop.hdfs.server.balancer.Dispatcher.Util;
import org.apache.hadoop.hdfs.server.blockmanagement.BlockPlacementPolicy;
import org.apache.hadoop.hdfs.server.blockmanagement.BlockPlacementPolicyDefault;
import org.apache.hadoop.hdfs.server.namenode.UnsupportedActionException;
import org.apache.hadoop.hdfs.server.protocol.DatanodeStorageReport;
import org.apache.hadoop.hdfs.server.protocol.StorageReport;
import org.apache.hadoop.io.IOUtils;
import org.apache.hadoop.util.StringUtils;
import org.apache.hadoop.util.Time;
import org.apache.hadoop.util.Tool;
import org.apache.hadoop.util.ToolRunner;

import com.google.common.base.Preconditions;

/** <p>The balancer is a tool that balances disk space usage on an HDFS cluster
* when some datanodes become full or when new empty nodes join the cluster.
* The tool is deployed as an application program that can be run by the
* cluster administrator on a live HDFS cluster while applications
* adding and deleting files.
*
* <p>SYNOPSIS
* <pre>
* To start:
* bin/start-balancer.sh [-threshold <threshold>]
* Example: bin/ start-balancer.sh
* start the balancer with a default threshold of 10%
* bin/ start-balancer.sh -threshold 5
* start the balancer with a threshold of 5%
* bin/ start-balancer.sh -idleiterations 20
* start the balancer with maximum 20 consecutive idle iterations
* bin/ start-balancer.sh -idleiterations -1
* run the balancer with default threshold infinitely
* To stop:
* bin/ stop-balancer.sh
* </pre>
*
* <p>DESCRIPTION
* <p>The threshold parameter is a fraction in the range of (1%, 100%) with a
* default value of 10%. The threshold sets a target for whether the cluster
* is balanced. A cluster is balanced if for each datanode, the utilization
* of the node (ratio of used space at the node to total capacity of the node)
* differs from the utilization of the (ratio of used space in the cluster
* to total capacity of the cluster) by no more than the threshold value.
* The smaller the threshold, the more balanced a cluster will become.
* It takes more time to run the balancer for small threshold values.
* Also for a very small threshold the cluster may not be able to reach the
* balanced state when applications write and delete files concurrently.
*
* <p>The tool moves blocks from highly utilized datanodes to poorly
* utilized datanodes iteratively. In each iteration a datanode moves or
* receives no more than the lesser of 10G bytes or the threshold fraction
* of its capacity. Each iteration runs no more than 20 minutes.
* At the end of each iteration, the balancer obtains updated datanodes
* information from the namenode.
*
* <p>A system property that limits the balancer's use of bandwidth is
* defined in the default configuration file:
* <pre>
* <property>
* <name>dfs.balance.bandwidthPerSec</name>
* <value>1048576</value>
* <description> Specifies the maximum bandwidth that each datanode
* can utilize for the balancing purpose in term of the number of bytes
* per second. </description>
* </property>
* </pre>
*
* <p>This property determines the maximum speed at which a block will be
* moved from one datanode to another. The default value is 1MB/s. The higher
* the bandwidth, the faster a cluster can reach the balanced state,
* but with greater competition with application processes. If an
* administrator changes the value of this property in the configuration
* file, the change is observed when HDFS is next restarted.
*
* <p>MONITERING BALANCER PROGRESS
* <p>After the balancer is started, an output file name where the balancer
* progress will be recorded is printed on the screen. The administrator
* can monitor the running of the balancer by reading the output file.
* The output shows the balancer's status iteration by iteration. In each
* iteration it prints the starting time, the iteration number, the total
* number of bytes that have been moved in the previous iterations,
* the total number of bytes that are left to move in order for the cluster
* to be balanced, and the number of bytes that are being moved in this
* iteration. Normally "Bytes Already Moved" is increasing while "Bytes Left
* To Move" is decreasing.
*
* <p>Running multiple instances of the balancer in an HDFS cluster is
* prohibited by the tool.
*
* <p>The balancer automatically exits when any of the following five
* conditions is satisfied:
* <ol>
* <li>The cluster is balanced;
* <li>No block can be moved;
* <li>No block has been moved for specified consecutive iterations (5 by default);
* <li>An IOException occurs while communicating with the namenode;
* <li>Another balancer is running.
* </ol>
*
* <p>Upon exit, a balancer returns an exit code and prints one of the
* following messages to the output file in corresponding to the above exit
* reasons:
* <ol>
* <li>The cluster is balanced. Exiting
* <li>No block can be moved. Exiting...
* <li>No block has been moved for specified iterations (5 by default). Exiting...
* <li>Received an IO exception: failure reason. Exiting...
* <li>Another balancer is running. Exiting...
* </ol>
*
* <p>The administrator can interrupt the execution of the balancer at any
* time by running the command "stop-balancer.sh" on the machine where the
* balancer is running.
*/


@InterfaceAudience.Private
public class Balancer {
static final Log LOG = LogFactory.getLog(Balancer.class);

static final Path BALANCER_ID_PATH = new Path("/system/balancer.id");

private static final long GB = 1L << 30; //1GB
private static final long MAX_SIZE_TO_MOVE = 10*GB;

private static final String USAGE = "Usage: hdfs balancer"
+ "\n\t[-policy <policy>]\tthe balancing policy: "
+ BalancingPolicy.Node.INSTANCE.getName() + " or "
+ BalancingPolicy.Pool.INSTANCE.getName()
+ "\n\t[-threshold <threshold>]\tPercentage of disk capacity"
+ "\n\t[-exclude [-f <hosts-file> | <comma-separated list of hosts>]]"
+ "\tExcludes the specified datanodes."
+ "\n\t[-include [-f <hosts-file> | <comma-separated list of hosts>]]"
+ "\tIncludes only the specified datanodes."
+ "\n\t[-idleiterations <idleiterations>]"
+ "\tNumber of consecutive idle iterations (-1 for Infinite) before "
+ "exit."
+ "\n\t[-runDuringUpgrade]"
+ "\tWhether to run the balancer during an ongoing HDFS upgrade."
+ "This is usually not desired since it will not affect used space "
+ "on over-utilized machines.";

private final Dispatcher dispatcher;
private final NameNodeConnector nnc;
private final BalancingPolicy policy;
private final boolean runDuringUpgrade;
private final double threshold;

// all data node lists
private final Collection<Source> overUtilized = new LinkedList<Source>();
private final Collection<Source> aboveAvgUtilized = new LinkedList<Source>();
private final Collection<StorageGroup> belowAvgUtilized
= new LinkedList<StorageGroup>();
private final Collection<StorageGroup> underUtilized
= new LinkedList<StorageGroup>();

/* Check that this Balancer is compatible with the Block Placement Policy
* used by the Namenode.
*/

private static void checkReplicationPolicyCompatibility(Configuration conf
) throws UnsupportedActionException {
if (!(BlockPlacementPolicy.getInstance(conf, null, null, null) instanceof
BlockPlacementPolicyDefault)) {
throw new UnsupportedActionException(
"Balancer without BlockPlacementPolicyDefault");
}
}

/**
* Construct a balancer.
* Initialize balancer. It sets the value of the threshold, and
* builds the communication proxies to
* namenode as a client and a secondary namenode and retry proxies
* when connection fails.
*/

Balancer(NameNodeConnector theblockpool, Parameters p, Configuration conf) {
final long movedWinWidth = conf.getLong(
DFSConfigKeys.DFS_BALANCER_MOVEDWINWIDTH_KEY,
DFSConfigKeys.DFS_BALANCER_MOVEDWINWIDTH_DEFAULT);
final int moverThreads = conf.getInt(
DFSConfigKeys.DFS_BALANCER_MOVERTHREADS_KEY,
DFSConfigKeys.DFS_BALANCER_MOVERTHREADS_DEFAULT);
final int dispatcherThreads = conf.getInt(
DFSConfigKeys.DFS_BALANCER_DISPATCHERTHREADS_KEY,
DFSConfigKeys.DFS_BALANCER_DISPATCHERTHREADS_DEFAULT);
final int maxConcurrentMovesPerNode = conf.getInt(
DFSConfigKeys.DFS_DATANODE_BALANCE_MAX_NUM_CONCURRENT_MOVES_KEY,
DFSConfigKeys.DFS_DATANODE_BALANCE_MAX_NUM_CONCURRENT_MOVES_DEFAULT);

this.nnc = theblockpool;
this.dispatcher = new Dispatcher(theblockpool, p.nodesToBeIncluded,
p.nodesToBeExcluded, movedWinWidth, moverThreads, dispatcherThreads,
maxConcurrentMovesPerNode, conf);
this.threshold = p.threshold;
this.policy = p.policy;
this.runDuringUpgrade = p.runDuringUpgrade;
}

private static long getCapacity(DatanodeStorageReport report, StorageType t) {
long capacity = 0L;
for(StorageReport r : report.getStorageReports()) {
if (r.getStorage().getStorageType() == t) {
capacity += r.getCapacity();
}
}
return capacity;
}

private static long getRemaining(DatanodeStorageReport report, StorageType t) {
long remaining = 0L;
for(StorageReport r : report.getStorageReports()) {
if (r.getStorage().getStorageType() == t) {
remaining += r.getRemaining();
}
}
return remaining;
}

/**
* Given a datanode storage set, build a network topology and decide
* over-utilized storages, above average utilized storages,
* below average utilized storages, and underutilized storages.
* The input datanode storage set is shuffled in order to randomize
* to the storage matching later on.
*
* @return the number of bytes needed to move in order to balance the cluster.
*/

private long init(List<DatanodeStorageReport> reports) {
// compute average utilization
for (DatanodeStorageReport r : reports) {
policy.accumulateSpaces(r);
}
policy.initAvgUtilization();

// create network topology and classify utilization collections:
// over-utilized, above-average, below-average and under-utilized.
long overLoadedBytes = 0L, underLoadedBytes = 0L;
for(DatanodeStorageReport r : reports) {
final DDatanode dn = dispatcher.newDatanode(r.getDatanodeInfo());
for(StorageType t : StorageType.getMovableTypes()) {
final Double utilization = policy.getUtilization(r, t);
if (utilization == null) { // datanode does not have such storage type
continue;
}

final long capacity = getCapacity(r, t);
final double utilizationDiff = utilization - policy.getAvgUtilization(t);
final double thresholdDiff = Math.abs(utilizationDiff) - threshold;
final long maxSize2Move = computeMaxSize2Move(capacity,
getRemaining(r, t), utilizationDiff, threshold);

final StorageGroup g;
if (utilizationDiff > 0) {
final Source s = dn.addSource(t, maxSize2Move, dispatcher);
if (thresholdDiff <= 0) { // within threshold
aboveAvgUtilized.add(s);
} else {
overLoadedBytes += percentage2bytes(thresholdDiff, capacity);
overUtilized.add(s);
}
g = s;
} else {
g = dn.addTarget(t, maxSize2Move);
if (thresholdDiff <= 0) { // within threshold
belowAvgUtilized.add(g);
} else {
underLoadedBytes += percentage2bytes(thresholdDiff, capacity);
underUtilized.add(g);
}
}
dispatcher.getStorageGroupMap().put(g);
}
}

logUtilizationCollections();

Preconditions.checkState(dispatcher.getStorageGroupMap().size()
== overUtilized.size() + underUtilized.size() + aboveAvgUtilized.size()
+ belowAvgUtilized.size(),
"Mismatched number of storage groups");

// return number of bytes to be moved in order to make the cluster balanced
return Math.max(overLoadedBytes, underLoadedBytes);
}

private static long computeMaxSize2Move(final long capacity, final long remaining,
final double utilizationDiff, final double threshold) {
final double diff = Math.min(threshold, Math.abs(utilizationDiff));
long maxSizeToMove = percentage2bytes(diff, capacity);
if (utilizationDiff < 0) {
maxSizeToMove = Math.min(remaining, maxSizeToMove);
}
return Math.min(MAX_SIZE_TO_MOVE, maxSizeToMove);
}

private static long percentage2bytes(double percentage, long capacity) {
Preconditions.checkArgument(percentage >= 0, "percentage = %s < 0",
percentage);
return (long)(percentage * capacity / 100.0);
}

/* log the over utilized & under utilized nodes */
private void logUtilizationCollections() {
logUtilizationCollection("over-utilized", overUtilized);
if (LOG.isTraceEnabled()) {
logUtilizationCollection("above-average", aboveAvgUtilized);
logUtilizationCollection("below-average", belowAvgUtilized);
}
logUtilizationCollection("underutilized", underUtilized);
}

private static <T extends StorageGroup>
void logUtilizationCollection(String name, Collection<T> items) {
LOG.info(items.size() + " " + name + ": " + items);
}

/**
* Decide all <source, target> pairs and
* the number of bytes to move from a source to a target
* Maximum bytes to be moved per storage group is
* min(1 Band worth of bytes, MAX_SIZE_TO_MOVE).
* @return total number of bytes to move in this iteration
*/

private long chooseStorageGroups() {
// First, match nodes on the same node group if cluster is node group aware
if (dispatcher.getCluster().isNodeGroupAware()) {
chooseStorageGroups(Matcher.SAME_NODE_GROUP);
}

// Then, match nodes on the same rack
chooseStorageGroups(Matcher.SAME_RACK);
// At last, match all remaining nodes
chooseStorageGroups(Matcher.ANY_OTHER);

return dispatcher.bytesToMove();
}

/** Decide all <source, target> pairs according to the matcher. */
private void chooseStorageGroups(final Matcher matcher) {
/* first step: match each overUtilized datanode (source) to
* one or more underUtilized datanodes (targets).
*/

chooseStorageGroups(overUtilized, underUtilized, matcher);

/* match each remaining overutilized datanode (source) to
* below average utilized datanodes (targets).
* Note only overutilized datanodes that haven't had that max bytes to move
* satisfied in step 1 are selected
*/

chooseStorageGroups(overUtilized, belowAvgUtilized, matcher);

/* match each remaining underutilized datanode (target) to
* above average utilized datanodes (source).
* Note only underutilized datanodes that have not had that max bytes to
* move satisfied in step 1 are selected.
*/

chooseStorageGroups(underUtilized, aboveAvgUtilized, matcher);
}

/**
* For each datanode, choose matching nodes from the candidates. Either the
* datanodes or the candidates are source nodes with (utilization > Avg), and
* the others are target nodes with (utilization < Avg).
*/

private <G extends StorageGroup, C extends StorageGroup>
void chooseStorageGroups(Collection<G> groups, Collection<C> candidates,
Matcher matcher) {
for(final Iterator<G> i = groups.iterator(); i.hasNext();) {
final G g = i.next();
for(; choose4One(g, candidates, matcher); );
if (!g.hasSpaceForScheduling()) {
i.remove();
}
}
}

/**
* For the given datanode, choose a candidate and then schedule it.
* @return true if a candidate is chosen; false if no candidates is chosen.
*/

private <C extends StorageGroup> boolean choose4One(StorageGroup g,
Collection<C> candidates, Matcher matcher) {
final Iterator<C> i = candidates.iterator();
final C chosen = chooseCandidate(g, i, matcher);

if (chosen == null) {
return false;
}
if (g instanceof Source) {
matchSourceWithTargetToMove((Source)g, chosen);
} else {
matchSourceWithTargetToMove((Source)chosen, g);
}
if (!chosen.hasSpaceForScheduling()) {
i.remove();
}
return true;
}

private void matchSourceWithTargetToMove(Source source, StorageGroup target) {
long size = Math.min(source.availableSizeToMove(), target.availableSizeToMove());
final Task task = new Task(target, size);
source.addTask(task);
target.incScheduledSize(task.getSize());
dispatcher.add(source, target);
LOG.info("Decided to move "+StringUtils.byteDesc(size)+" bytes from "
+ source.getDisplayName() + " to " + target.getDisplayName());
}

/** Choose a candidate for the given datanode. */
private <G extends StorageGroup, C extends StorageGroup>
C chooseCandidate(G g, Iterator<C> candidates, Matcher matcher) {
if (g.hasSpaceForScheduling()) {
for(; candidates.hasNext(); ) {
final C c = candidates.next();
if (!c.hasSpaceForScheduling()) {
candidates.remove();
} else if (matcher.match(dispatcher.getCluster(),
g.getDatanodeInfo(), c.getDatanodeInfo())) {
return c;
}
}
}
return null;
}

/* reset all fields in a balancer preparing for the next iteration */
void resetData(Configuration conf) {
this.overUtilized.clear();
this.aboveAvgUtilized.clear();
this.belowAvgUtilized.clear();
this.underUtilized.clear();
this.policy.reset();
dispatcher.reset(conf);;
}

static class Result {
final ExitStatus exitStatus;
final long bytesLeftToMove;
final long bytesBeingMoved;
final long bytesAlreadyMoved;

Result(ExitStatus exitStatus, long bytesLeftToMove, long bytesBeingMoved,
long bytesAlreadyMoved) {
this.exitStatus = exitStatus;
this.bytesLeftToMove = bytesLeftToMove;
this.bytesBeingMoved = bytesBeingMoved;
this.bytesAlreadyMoved = bytesAlreadyMoved;
}

void print(int iteration, PrintStream out) {
out.printf("%-24s %10d %19s %18s %17s%n",
DateFormat.getDateTimeInstance().format(new Date()), iteration,
StringUtils.byteDesc(bytesAlreadyMoved),
StringUtils.byteDesc(bytesLeftToMove),
StringUtils.byteDesc(bytesBeingMoved));
}
}

Result newResult(ExitStatus exitStatus, long bytesLeftToMove, long bytesBeingMoved) {
return new Result(exitStatus, bytesLeftToMove, bytesBeingMoved,
dispatcher.getBytesMoved());
}

Result newResult(ExitStatus exitStatus) {
return new Result(exitStatus, -1, -1, dispatcher.getBytesMoved());
}

/** Run an iteration for all datanodes. */
Result runOneIteration() {
try {
final List<DatanodeStorageReport> reports = dispatcher.init();
final long bytesLeftToMove = init(reports);
if (bytesLeftToMove == 0) {
System.out.println("The cluster is balanced. Exiting...");
return newResult(ExitStatus.SUCCESS, bytesLeftToMove, -1);
} else {
LOG.info( "Need to move "+ StringUtils.byteDesc(bytesLeftToMove)
+ " to make the cluster balanced." );
}

// Should not run the balancer during an unfinalized upgrade, since moved
// blocks are not deleted on the source datanode.
if (!runDuringUpgrade && nnc.isUpgrading()) {
return newResult(ExitStatus.UNFINALIZED_UPGRADE, bytesLeftToMove, -1);
}

/* Decide all the nodes that will participate in the block move and
* the number of bytes that need to be moved from one node to another
* in this iteration. Maximum bytes to be moved per node is
* Min(1 Band worth of bytes, MAX_SIZE_TO_MOVE).
*/

final long bytesBeingMoved = chooseStorageGroups();
if (bytesBeingMoved == 0) {
System.out.println("No block can be moved. Exiting...");
return newResult(ExitStatus.NO_MOVE_BLOCK, bytesLeftToMove, bytesBeingMoved);
} else {
LOG.info( "Will move " + StringUtils.byteDesc(bytesBeingMoved) +
" in this iteration");
}

/* For each pair of <source, target>, start a thread that repeatedly
* decide a block to be moved and its proxy source,
* then initiates the move until all bytes are moved or no more block
* available to move.
* Exit no byte has been moved for 5 consecutive iterations.
*/

if (!dispatcher.dispatchAndCheckContinue()) {
return newResult(ExitStatus.NO_MOVE_PROGRESS, bytesLeftToMove, bytesBeingMoved);
}

return newResult(ExitStatus.IN_PROGRESS, bytesLeftToMove, bytesBeingMoved);
} catch (IllegalArgumentException e) {
System.out.println(e + ". Exiting ...");
return newResult(ExitStatus.ILLEGAL_ARGUMENTS);
} catch (IOException e) {
System.out.println(e + ". Exiting ...");
return newResult(ExitStatus.IO_EXCEPTION);
} catch (InterruptedException e) {
System.out.println(e + ". Exiting ...");
return newResult(ExitStatus.INTERRUPTED);
} finally {
dispatcher.shutdownNow();
}
}

/**
* Balance all namenodes.
* For each iteration,
* for each namenode,
* execute a {@link Balancer} to work through all datanodes once.
*/

static int run(Collection<URI> namenodes, final Parameters p,
Configuration conf) throws IOException, InterruptedException {
final long sleeptime =
conf.getLong(DFSConfigKeys.DFS_HEARTBEAT_INTERVAL_KEY,
DFSConfigKeys.DFS_HEARTBEAT_INTERVAL_DEFAULT) * 2000 +
conf.getLong(DFSConfigKeys.DFS_NAMENODE_REPLICATION_INTERVAL_KEY,
DFSConfigKeys.DFS_NAMENODE_REPLICATION_INTERVAL_DEFAULT) * 1000;
LOG.info("namenodes = " + namenodes);
LOG.info("parameters = " + p);

System.out.println("Time Stamp Iteration# Bytes Already Moved Bytes Left To Move Bytes Being Moved");

List<NameNodeConnector> connectors = Collections.emptyList();
try {
connectors = NameNodeConnector.newNameNodeConnectors(namenodes,
Balancer.class.getSimpleName(), BALANCER_ID_PATH, conf, p.maxIdleIteration);

boolean done = false;
for(int iteration = 0; !done; iteration++) {
done = true;
Collections.shuffle(connectors);
for(NameNodeConnector nnc : connectors) {
final Balancer b = new Balancer(nnc, p, conf);
final Result r = b.runOneIteration();
r.print(iteration, System.out);

// clean all lists
b.resetData(conf);
if (r.exitStatus == ExitStatus.IN_PROGRESS) {
done = false;
} else if (r.exitStatus != ExitStatus.SUCCESS) {
//must be an error statue, return.
return r.exitStatus.getExitCode();
}
}

if (!done) {
Thread.sleep(sleeptime);
}
}
} finally {
for(NameNodeConnector nnc : connectors) {
IOUtils.cleanup(LOG, nnc);
}
}
return ExitStatus.SUCCESS.getExitCode();
}

/* Given elaspedTime in ms, return a printable string */
private static String time2Str(long elapsedTime) {
String unit;
double time = elapsedTime;
if (elapsedTime < 1000) {
unit = "milliseconds";
} else if (elapsedTime < 60*1000) {
unit = "seconds";
time = time/1000;
} else if (elapsedTime < 3600*1000) {
unit = "minutes";
time = time/(60*1000);
} else {
unit = "hours";
time = time/(3600*1000);
}

return time+" "+unit;
}

static class Parameters {
static final Parameters DEFAULT = new Parameters(
BalancingPolicy.Node.INSTANCE, 10.0,
NameNodeConnector.DEFAULT_MAX_IDLE_ITERATIONS,
Collections.<String> emptySet(), Collections.<String> emptySet(),
false);

final BalancingPolicy policy;
final double threshold;
final int maxIdleIteration;
// exclude the nodes in this set from balancing operations
Set<String> nodesToBeExcluded;
//include only these nodes in balancing operations
Set<String> nodesToBeIncluded;
/**
* Whether to run the balancer during upgrade.
*/

final boolean runDuringUpgrade;

Parameters(BalancingPolicy policy, double threshold, int maxIdleIteration,
Set<String> nodesToBeExcluded, Set<String> nodesToBeIncluded,
boolean runDuringUpgrade) {
this.policy = policy;
this.threshold = threshold;
this.maxIdleIteration = maxIdleIteration;
this.nodesToBeExcluded = nodesToBeExcluded;
this.nodesToBeIncluded = nodesToBeIncluded;
this.runDuringUpgrade = runDuringUpgrade;
}

@Override
public String toString() {
return String.format("%s.%s [%s,"
+ " threshold = %s,"
+ " max idle iteration = %s, "
+ "number of nodes to be excluded = %s,"
+ " number of nodes to be included = %s,"
+ " run during upgrade = %s]",
Balancer.class.getSimpleName(), getClass().getSimpleName(),
policy, threshold, maxIdleIteration,
nodesToBeExcluded.size(), nodesToBeIncluded.size(),
runDuringUpgrade);
}
}

static class Cli extends Configured implements Tool {
/**
* Parse arguments and then run Balancer.
*
* @param args command specific arguments.
* @return exit code. 0 indicates success, non-zero indicates failure.
*/

@Override
public int run(String[] args) {
final long startTime = Time.monotonicNow();
final Configuration conf = getConf();

try {
checkReplicationPolicyCompatibility(conf);

final Collection<URI> namenodes = DFSUtil.getNsServiceRpcUris(conf);
return Balancer.run(namenodes, parse(args), conf);
} catch (IOException e) {
System.out.println(e + ". Exiting ...");
return ExitStatus.IO_EXCEPTION.getExitCode();
} catch (InterruptedException e) {
System.out.println(e + ". Exiting ...");
return ExitStatus.INTERRUPTED.getExitCode();
} finally {
System.out.format("%-24s ",
DateFormat.getDateTimeInstance().format(new Date()));
System.out.println("Balancing took "
+ time2Str(Time.monotonicNow() - startTime));
}
}

/** parse command line arguments */
static Parameters parse(String[] args) {
BalancingPolicy policy = Parameters.DEFAULT.policy;
double threshold = Parameters.DEFAULT.threshold;
int maxIdleIteration = Parameters.DEFAULT.maxIdleIteration;
Set<String> nodesTobeExcluded = Parameters.DEFAULT.nodesToBeExcluded;
Set<String> nodesTobeIncluded = Parameters.DEFAULT.nodesToBeIncluded;
boolean runDuringUpgrade = Parameters.DEFAULT.runDuringUpgrade;

if (args != null) {
try {
for(int i = 0; i < args.length; i++) {
if ("-threshold".equalsIgnoreCase(args[i])) {
checkArgument(++i < args.length,
"Threshold value is missing: args = " + Arrays.toString(args));
try {
threshold = Double.parseDouble(args[i]);
if (threshold < 1 || threshold > 100) {
throw new IllegalArgumentException(
"Number out of range: threshold = " + threshold);
}
LOG.info( "Using a threshold of " + threshold );
} catch(IllegalArgumentException e) {
System.err.println(
"Expecting a number in the range of [1.0, 100.0]: "
+ args[i]);
throw e;
}
} else if ("-policy".equalsIgnoreCase(args[i])) {
checkArgument(++i < args.length,
"Policy value is missing: args = " + Arrays.toString(args));
try {
policy = BalancingPolicy.parse(args[i]);
} catch(IllegalArgumentException e) {
System.err.println("Illegal policy name: " + args[i]);
throw e;
}
} else if ("-exclude".equalsIgnoreCase(args[i])) {
checkArgument(++i < args.length,
"List of nodes to exclude | -f <filename> is missing: args = "
+ Arrays.toString(args));
if ("-f".equalsIgnoreCase(args[i])) {
checkArgument(++i < args.length,
"File containing nodes to exclude is not specified: args = "
+ Arrays.toString(args));
nodesTobeExcluded = Util.getHostListFromFile(args[i], "exclude");
} else {
nodesTobeExcluded = Util.parseHostList(args[i]);
}
} else if ("-include".equalsIgnoreCase(args[i])) {
checkArgument(++i < args.length,
"List of nodes to include | -f <filename> is missing: args = "
+ Arrays.toString(args));
if ("-f".equalsIgnoreCase(args[i])) {
checkArgument(++i < args.length,
"File containing nodes to include is not specified: args = "
+ Arrays.toString(args));
nodesTobeIncluded = Util.getHostListFromFile(args[i], "include");
} else {
nodesTobeIncluded = Util.parseHostList(args[i]);
}
} else if ("-idleiterations".equalsIgnoreCase(args[i])) {
checkArgument(++i < args.length,
"idleiterations value is missing: args = " + Arrays
.toString(args));
maxIdleIteration = Integer.parseInt(args[i]);
LOG.info("Using a idleiterations of " + maxIdleIteration);
} else if ("-runDuringUpgrade".equalsIgnoreCase(args[i])) {
runDuringUpgrade = true;
LOG.info("Will run the balancer even during an ongoing HDFS "
+ "upgrade. Most users will not want to run the balancer "
+ "during an upgrade since it will not affect used space "
+ "on over-utilized machines.");
} else {
throw new IllegalArgumentException("args = "
+ Arrays.toString(args));
}
}
checkArgument(nodesTobeExcluded.isEmpty() || nodesTobeIncluded.isEmpty(),
"-exclude and -include options cannot be specified together.");
} catch(RuntimeException e) {
printUsage(System.err);
throw e;
}
}

return new Parameters(policy, threshold, maxIdleIteration,
nodesTobeExcluded, nodesTobeIncluded, runDuringUpgrade);
}

private static void printUsage(PrintStream out) {
out.println(USAGE + "\n");
}
}

/**
* Run a balancer
* @param args Command line arguments
*/

public static void main(String[] args) {
if (DFSUtil.parseHelpArgument(args, USAGE, System.out, true)) {
System.exit(0);
}

try {
System.exit(ToolRunner.run(new HdfsConfiguration(), new Cli(), args));
} catch (Throwable e) {
LOG.error("Exiting balancer due an exception", e);
System.exit(-1);
}
}
}

原创文章,转载请注明: 转载自sparkjvm的博客
本博客的文章集合: http://www.itweet.cn/archives/

参考:http://hadoop.apache.org/docs/r2.6.0/hadoop-project-dist/hadoop-hdfs/HDFSCommands.html#balancer