hbase之宽表与窄表对split的影响
Hbase的hbase.hregion.max.filesize属性值用来指定region分割的阀值, 该值默认为268435456(256MB), 当一个列族文件大小超过该值时,将会分裂成两个region。
hbase的列可以有很多,设计时有两种方式可选择, 宽表(一行有很多列)和窄表
如有一个存储用户邮件的表
按宽表设计时,可以表示成(一个用户的所有邮件存成一行)
userid1 email1 emali2 email3 ... ... ... ... ... emailn
userid2 email1 emali2 email3 ... ... ... ... ... emailn
useridn
按窄表设计时,可以表示成(rowkey由用ID和emailID组成)
userid1_emialid1 email1
userid1_emialid2 email2
userid1_emialid3 email2
userid1_emialidn emailn
userid2_emialid1 email1
userid2_emialid2 email2
userid2_emialid3 email3
userid2_emialidn emailn
这两种设计方法会对region的分割造成影响, 今天在看HFileOutputFormat代码时发现它new出的RecordWriter对 region分割有一定的限制,
只有当rowkey不同是才会做分割, 而rowkey相同时即使region大小已经超过hbase.hregion.max.filesize值, 也不会分割
RecordWriter代码:
1. public void write(ImmutableBytesWritable row, KeyValue kv)
2. throws IOException {
3. long length = kv.getLength();
4. byte [] family = kv.getFamily();
5. WriterLength wl = this.writers.get(family);
6. if (wl == null || ((length + wl.written) >= maxsize) &&
7. Bytes.compareTo(this.previousRow, 0, this.previousRow.length,
8. kv.getBuffer(), kv.getRowOffset(), kv.getRowLength()) != 0) {
9. // Get a new writer.
10. Path basedir = new Path(outputdir, Bytes.toString(family));
11. if (wl == null) {
12. wl = new WriterLength();
13. this.writers.put(family, wl);
14. if (this.writers.size() > 1) throw new IOException("One family only");
15. // If wl == null, first file in family. Ensure family dir exits.
16. if (!fs.exists(basedir)) fs.mkdirs(basedir);
17. }
18. wl.writer = getNewWriter(wl.writer, basedir);
19. LOG.info("Writer=" + wl.writer.getPath() +
20. ((wl.written == 0)? "": ", wrote=" + wl.written));
21. wl.written = 0;
22. }
23. kv.updateLatestStamp(this.now);
24. wl.writer.append(kv);
25. wl.written += length;
26. // Copy the row so we know when a row transition.
27. this.previousRow = kv.getRow();
28. }
标红加粗部分说明当块大小大于hbase.hregion.max.filesize值, 并却当前行与上一次插入的行不同时才会分割region.
1. 宽表情况下, 单独一行大小超过hbase.hregion.max.filesize值, 不会做分割
2. 相同rowkey下插入很多不同版本的记录,即使大小超过hbase.hregion.max.filesize值, 也不会做分割
下面就来验证下:
为了尽早看到效果, 需要在hbase-site.xml中修改两个配置参数
1. <property>
2. <name>hbase.hregion.memstore.flush.size</name>
3. <value>5</value>
4. <description>
5. Memstore will be flushed to disk if size of the memstore
6. exceeds this number of bytes. Value is checked by a thread that runs
7. every hbase.server.thread.wakefrequency.
8. </description>
9. </property>
10. <property>
11. <name>hbase.hregion.max.filesize</name>
12. <value>10</value>
13. <description>
14. Maximum HStoreFile size. If any one of a column families' HStoreFiles has
15. grown to exceed this value, the hosting HRegion is split in two.
16. Default: 256M.
17. </description>
18. </property>
建测试表t1和t2
1. hbase(main):076:0* create 't1','f1'
2. 0 row(s) in 1.6460 seconds
3.
4. hbase(main):077:0> create 't2','f1'
5. 0 row(s) in 1.1790 seconds
查看系统表 .META.
1. hbase(main):081:0* scan '.META.'
2. ROW COLUMN+CELL
3. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:regioninfo,timestamp=1314720667384, value=REGION => {NAME =>'t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad.', STARTKEY => '', ENDK
4. . EY => '', ENCODED =>d8acd6bc659ac8326b88850d645a90ad, TABLE => {{NAME => 't1', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SCOPE
5. => '0', COMPRESSION => 'NONE',VERSIONS => '3', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY => 'false',BLOCKCACHE => 'true'}]}}
6. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:server,timestamp=1314720667941,value=yinjie:60020
7. .
8. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:serverstartcode,timestamp=1314720667941,value=1314716290123
9. .
10. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:regioninfo,timestamp=1314720672241, value=REGION => {NAME =>'t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71.', STARTKEY => '', ENDK
11. . EY => '', ENCODED =>16bb3d2563eab3b4e25477c64e007e71, TABLE => {{NAME => 't2', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SCOPE
12. => '0', COMPRESSION => 'NONE',VERSIONS => '3', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY => 'false',BLOCKCACHE => 'true'}]}}
13. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:server,timestamp=1314720672346,value=yinjie:60020
14. .
15. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:serverstartcode,timestamp=1314720672346,value=1314716290123
16. .
17. 2 row(s) in 0.0230 seconds
可以看到此时,t1,t2都已有一个region
先往t1表插入10条记录,rowkwy相同
1. hbase(main):086:0* for i in 0..9 do\
2. hbase(main):087:1* put 't1','row1',"f1:c#{i}","swallow#{i}"\
3. hbase(main):088:1* end
4. 0 row(s) in 0.0180 seconds
5.
6. 0 row(s) in 0.0070 seconds
7.
8. 0 row(s) in 0.0420 seconds
9.
10. 0 row(s) in 0.0620 seconds
11.
12. 0 row(s) in 0.0120 seconds
13.
14. 0 row(s) in 0.0770 seconds
15.
16. 0 row(s) in 0.0150 seconds
17.
18. 0 row(s) in 0.1290 seconds
19.
20. 0 row(s) in 10.0740 seconds
21.
22. 0 row(s) in 0.1230 seconds
23.
24. => 0..9
25. hbase(main):089:0>
查看t1记录
1. hbase(main):089:0>
2. ROW COLUMN+CELL
3. row1 column=f1:c0,timestamp=1314720946495,value=swallow0
4. row1 column=f1:c1,timestamp=1314720946507,value=swallow1
5. row1 column=f1:c2,timestamp=1314720946903,value=swallow2
6. row1 column=f1:c3,timestamp=1314720946939,value=swallow3
7. row1 column=f1:c4,timestamp=1314720946976,value=swallow4
8. row1 column=f1:c5,timestamp=1314720947055,value=swallow5
9. row1 column=f1:c6,timestamp=1314720947070,value=swallow6
10. row1 column=f1:c7,timestamp=1314720947198,value=swallow7
11. row1 column=f1:c8,timestamp=1314720957272,value=swallow8
12. row1 column=f1:c9,timestamp=1314720957392,value=swallow9
13. 1 row(s) in 0.0300 seconds
查看 .META.
1. hbase(main):090:0>
2. ROW COLUMN+CELL
3. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:regioninfo,timestamp=1314720667384, value=REGION => {NAME =>'t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad.', STARTKEY => '', ENDK
4. . EY => '', ENCODED =>d8acd6bc659ac8326b88850d645a90ad, TABLE => {{NAME => 't1', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SCOPE
5. => '0', COMPRESSION => 'NONE',VERSIONS => '3', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY => 'false',BLOCKCACHE => 'true'}]}}
6. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:server,timestamp=1314720667941,value=yinjie:60020
7. .
8. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:serverstartcode,timestamp=1314720667941,value=1314716290123
9. .
10. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:regioninfo,timestamp=1314720672241, value=REGION => {NAME =>'t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71.', STARTKEY => '', ENDK
11. . EY => '', ENCODED =>16bb3d2563eab3b4e25477c64e007e71, TABLE => {{NAME => 't2', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SCOPE
12. => '0', COMPRESSION => 'NONE',VERSIONS => '3', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY => 'false',BLOCKCACHE => 'true'}]}}
13. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:server,timestamp=1314720672346,value=yinjie:60020
14. .
15. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:serverstartcode,timestamp=1314720672346,value=1314716290123
16. .
17. 2 row(s) in 0.0210 seconds
可以看到t1仍旧只有一个region
接下去往往t2表插入10条相同记录,但rowkwy不同
1. hbase(main):091:0>
2. hbase(main):092:1* put 't2',"row#{i}","f1:c#{i}","swallow#{i}"\
3. hbase(main):093:1* end
4. 0 row(s) in 0.1140 seconds
5.
6. 0 row(s) in 0.0080 seconds
7.
8. 0 row(s) in 0.0410 seconds
9.
10. 0 row(s) in 0.0820 seconds
11.
12. 0 row(s) in 0.0210 seconds
13.
14. 0 row(s) in 0.0410 seconds
15.
16. 0 row(s) in 0.0200 seconds
17.
18. 0 row(s) in 0.1210 seconds
19.
20. 0 row(s) in 0.0140 seconds
21.
22. 0 row(s) in 0.0360 seconds
23.
24. => 0..9
查看t2记录
1. hbase(main):097:0* scan 't2'
2. ROW COLUMN+CELL
3. row0 column=f1:c0,timestamp=1314721110769,value=swallow0
4. row1 column=f1:c1,timestamp=1314721110787,value=swallow1
5. row2 column=f1:c2,timestamp=1314721110830,value=swallow2
6. row3 column=f1:c3,timestamp=1314721110916,value=swallow3
7. row4 column=f1:c4,timestamp=1314721110932,value=swallow4
8. row5 column=f1:c5,timestamp=1314721110971,value=swallow5
9. row6 column=f1:c6,timestamp=1314721110989,value=swallow6
10. row7 column=f1:c7,timestamp=1314721111121,value=swallow7
11. row8 column=f1:c8,timestamp=1314721111130,value=swallow8
12. row9 column=f1:c9,timestamp=1314721111172,value=swallow9
13. 10 row(s) in 1.0450 seconds
查看 .META.
1. hbase(main):102:0>
2. ROW COLUMN+CELL
3. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:regioninfo,timestamp=1314720667384, value=REGION => {NAME =>'t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad.', STARTKEY => '', ENDK
4. . EY => '', ENCODED =>d8acd6bc659ac8326b88850d645a90ad, TABLE => {{NAME => 't1', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SCOPE
5. => '0', COMPRESSION => 'NONE',VERSIONS => '3', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY => 'false',BLOCKCACHE => 'true'}]}}
6. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:server,timestamp=1314720667941,value=yinjie:60020
7. .
8. t1,,1314720667274.d8acd6bc659ac8326b88850d645a90ad column=info:serverstartcode,timestamp=1314720667941,value=1314716290123
9. .
10. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:regioninfo,timestamp=1314721112130, value=REGION => {NAME =>'t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71.', STARTKEY => '', ENDK
11. . EY => '', ENCODED =>16bb3d2563eab3b4e25477c64e007e71, OFFLINE => true, SPLIT => true, TABLE => {{NAME=> 't2', FAMILIES => [{NAME => 'f1', BLOOMFILT
12. ER => 'NONE', REPLICATION_SCOPE=> '0', VERSIONS => '3', COMPRESSION => 'NONE', TTL => '2147483647', BLOCKSIZE =>'65536', IN_MEMORY => 'false', BLOC
13. KCACHE =>'true'}]}}
14. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:server,timestamp=1314720672346,value=yinjie:60020
15. .
16. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:serverstartcode,timestamp=1314720672346,value=1314716290123
17. .
18. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:splitA,timestamp=1314721112130, value=REGION => {NAME =>'t2,,1314721111490.71df02214242923574b71fe5e2a19360.', STARTKEY => '', ENDKEY =
19. . > 'row0', ENCODED =>71df02214242923574b71fe5e2a19360, TABLE => {{NAME => 't2', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SCOPE
20. => '0', VERSIONS => '3',COMPRESSION => 'NONE', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY =>'false', BLOCKCACHE => 'true'}]}}
21. t2,,1314720672168.16bb3d2563eab3b4e25477c64e007e71 column=info:splitB,timestamp=1314721112130, value=REGION => {NAME =>'t2,row0,1314721111490.915ee8d4a32c59a4ec3960e335b061ca.', STARTKEY => 'row0',
22. . ENDKEY => '', ENCODED =>915ee8d4a32c59a4ec3960e335b061ca, TABLE => {{NAME => 't2', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SC
23. OPE => '0', VERSIONS => '3',COMPRESSION => 'NONE', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY =>'false', BLOCKCACHE => 'true'}]}}
24. t2,,1314721111490.71df02214242923574b71fe5e2a19360 column=info:regioninfo,timestamp=1314721112267, value=REGION => {NAME =>'t2,,1314721111490.71df02214242923574b71fe5e2a19360.', STARTKEY => '', ENDK
25. . EY => 'row0', ENCODED =>71df02214242923574b71fe5e2a19360, TABLE => {{NAME => 't2', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATION_SC
26. OPE => '0', VERSIONS => '3',COMPRESSION => 'NONE', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY =>'false', BLOCKCACHE => 'true'}]}}
27. t2,,1314721111490.71df02214242923574b71fe5e2a19360 column=info:server,timestamp=1314721112267,value=yinjie:60020
28. .
29. t2,,1314721111490.71df02214242923574b71fe5e2a19360 column=info:serverstartcode,timestamp=1314721112267,value=1314716290123
30. .
31. t2,row0,1314721111490.915ee8d4a32c59a4ec3960e335b0 column=info:regioninfo,timestamp=1314721112627, value=REGION => {NAME =>'t2,row0,1314721111490.915ee8d4a32c59a4ec3960e335b061ca.', STARTKEY => 'row
32. 61ca. 0', ENDKEY => '', ENCODED =>915ee8d4a32c59a4ec3960e335b061ca, TABLE => {{NAME => 't2', FAMILIES => [{NAME =>'f1', BLOOMFILTER => 'NONE', REPLICATIO
33. N_SCOPE => '0', VERSIONS =>'3', COMPRESSION => 'NONE', TTL => '2147483647', BLOCKSIZE => '65536', IN_MEMORY=> 'false', BLOCKCACHE => 'true'}]}}
34. t2,row0,1314721111490.915ee8d4a32c59a4ec3960e335b0 column=info:server,timestamp=1314721112627,value=yinjie:60020
35. 61ca.
36. t2,row0,1314721111490.915ee8d4a32c59a4ec3960e335b0 column=info:serverstartcode,timestamp=1314721112627,value=1314716290123
37. 61ca.
38. 4 row(s) in 0.0380 seconds
可以看到t2的region已经分裂.