[
https://issues.apache.org/jira/browse/HBASE-2378?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Todd Lipcon updated HBASE-2378:
-------------------------------
Attachment: hbase-2378.txt
Here's a fix as well as a test case. The core of the issue was that
ImmutableBytesWritable, when comparing Java objects in RAM (as opposed to in
byte form) counted shorter strings as less than longer strings, even if the
first characters were higher.
I ran Ruslan's code and was able to successfully use the resulting table. The
one modification I made was to explicitly set the sorting class to
ImmutableBytesWritable.Comparator, though I think it would work without that
change.
> Bulk insert with multiple reducers
> ----------------------------------
>
> Key: HBASE-2378
> URL: https://issues.apache.org/jira/browse/HBASE-2378
> Project: Hadoop HBase
> Issue Type: Bug
> Components: mapreduce
> Affects Versions: 0.20.3
> Reporter: Ruslan Salyakhov
> Assignee: Todd Lipcon
> Attachments: hbase-2378.txt, HFileOutputFormat.java,
> my_sample_log_1k.txt, MyHFilesWriter.java, MyKeyComparator.java,
> MyReadPerformance.java, MySampler.java,
> TestTotalOrderPartitionerForMyKeys.java, TotalOrderPartitioner.java
>
>
> If I run MR to prepare HFIles with more than one reducer then some values for
> keys are not appeared in the table after loadtable.rb script execution. With
> one reducer everything works fine.
> References:
> http://hadoop.apache.org/hbase/docs/r0.20.3/api/org/apache/hadoop/hbase/mapreduce/package-summary.html#bulk
> - the row id must be formatted as a ImmutableBytesWritable
> - MR job should ensure a total ordering among all keys
> MAPREDUCE-366 (patch-5668-3.txt)
> - TotalOrderPartitioner that uses the new API (attached)
> HBASE-2063
> - patched HFileOutputFormat (attached)
> Input data (attached):
> * my_sample_log_1k.txt - sample data, input for MyHFilesWriter
> Source (attached):
> * MyKeyComparator.java - comparator for my ImmutableBytesWritable keys
> * TestTotalOrderPartitionerForMyKeys.java - test case for my keys (note that
> I've set up MyKeyComparator to pass that test)
> * MyHFilesWriter.java - My MR job to prepare HFiles
> * HFileOutputFormat.java - from MAPREDUCE-366
> * TotalOrderPartitioner.java - from MAPREDUCE-366
> * MySampler.java - My RandomSampler based on Sampler from MAPREDUCE-366 BUT
> I've put the following string into getSample method (without that string it
> doesn't work):
> {code}
> reader.initialize(splits.get(i), new
> TaskAttemptContext(job.getConfiguration(), new TaskAttemptID()));
> {code}
> Test case:
> # comment the following string in MyHFilesWriter:
> //job.setSortComparatorClass(MyKeyComparator.class);
> # hadoop jar keyvalue-poc.jar MyHFilesWriter -in
> /test_hbase/my_sample_log_1k.txt -out /test_hbase/hfiles/01/ -r 1
> # hadoop jar keyvalue-poc.jar MyHFilesWriter -in
> /test_hbase/my_sample_log_1k.txt -out /test_hbase/hfiles/02/ -r 2
> # hbase> create 'tst_hfiles_01', {NAME => 'vals'}
> # hbase> create 'tst_hfiles_02', {NAME => 'vals'}
> # hbase org.jruby.Main /usr/lib/hbase-0.20/bin/loadtable.rb tst_hfiles_01
> /test_hbase/hfiles/01
> # hbase org.jruby.Main /usr/lib/hbase-0.20/bin/loadtable.rb tst_hfiles_02
> /test_hbase/hfiles/02
> # check values for keys
> # uncomment the following string in MyHFilesWriter:
> //job.setSortComparatorClass(MyKeyComparator.class);
> # hadoop jar keyvalue-poc.jar MyHFilesWriter -in
> /test_hbase/my_sample_log_1k.txt -out /test_hbase/hfiles/03/ -r 2
> for example, results:
> {code}
> hbase(main):006:0* count 'tst_hfiles_01', 100
> Current count: 100, row: 0.14.USA.IL.602.ELMHURST.1.1.0.0
>
> Current count: 200, row: 0.245.USA.ME.500.PORTLAND.1.1.0.0
>
> Current count: 300, row: 0.34.USA.FL.Rollup.Rollup.1.1.0.0
>
> Current count: 400, row: 0.443.USA.CA.803.LOS.ANGELES.1.1.0
>
> Current count: 500, row: 0.8.USA.CO.751.CASTLE.ROCK.1.1.0
>
> Current count: 600, row: 1.14.DZA.Rollup.Rollup.Rollup.1.1.0.1
>
> Current count: 700, row: 1.159.SWE.AB.Rollup.Rollup.1.1.0.1
>
> Current count: 800, row: 1.17.USA.TN.659.CLARKSVILLE.1.1.0.1
>
> Current count: 900, row: 1.220.USA.MI.505.SOUTHFIELD.1.1.0.1
>
> 999 row(s) in 0.0930 seconds
> hbase(main):007:0> count 'tst_hfiles_02', 100
> Current count: 100, row: 0.231.USA.GA.524.BUFORD.1.1.0.1
>
> Current count: 200, row: 0.4.USA.VA.573.Rollup.1.1.0.0
>
> Current count: 300, row: 0.9.ROU.B.-1.BUCHAREST.1.1.0.0
>
> Current count: 400, row: 1.16.USA.IA.679.Rollup.1.1.1.0
>
> Current count: 500, row: 1.245.NOR.03.-1.OSLO.1.1.0.0
>
> Current count: 600, row: 0.245.GBR.ENG.826005.BEXLEY.1.1.0.1
>
> Current count: 700, row: 0.48.GBR.ENG.826027.Rollup.1.1.0.1
>
> Current count: 800, row: 1.14.SWE.Rollup.Rollup.Rollup.1.1.0.1
>
> Current count: 900, row: 1.201.GBR.ENG.826005.LONDON.1.1.0.1
>
> 999 row(s) in 0.1630 seconds
> hbase(main):008:0> get 'tst_hfiles_01', '0.14.USA.IL.602.ELMHURST.1.1.0.0'
> COLUMN CELL
>
> vals:key0 timestamp=1269542753914, value=0
>
> vals:key1 timestamp=1269542753914, value=14
>
> vals:key2 timestamp=1269542753914, value=USA
>
> vals:key3 timestamp=1269542753914, value=IL
>
> vals:key4 timestamp=1269542753914, value=602
>
> vals:key5 timestamp=1269542753914, value=ELMHURST
>
> vals:key6 timestamp=1269542753914, value=1
>
> vals:key7 timestamp=1269542753914, value=1
>
> vals:key8 timestamp=1269542753914, value=0
>
> vals:key9 timestamp=1269542753914, value=0
>
> vals:val0 timestamp=1269542753914, value=2
>
> 11 row(s) in 0.0160 seconds
> hbase(main):009:0> get 'tst_hfiles_02', '0.14.USA.IL.602.ELMHURST.1.1.0.0'
> COLUMN CELL
>
> 0 row(s) in 0.0220 seconds
> {code}
> with MyKeyComparator
> {code}
> java.io.IOException: Added a key not lexically larger than previous
> key=.103.FRA.V.-1.LYON.1.1.0.0valskey0'XXX,
> lastkey=1.20.USA.AOL.0.AOL.1.1.0.0valsval0'XXX
> at
> org.apache.hadoop.hbase.io.hfile.HFile$Writer.checkKey(HFile.java:551)
> at org.apache.hadoop.hbase.io.hfile.HFile$Writer.append(HFile.java:513)
> at org.apache.hadoop.hbase.io.hfile.HFile$Writer.append(HFile.java:481)
> at
> com.contextweb.hadoop.hbase.mapred.HFileOutputFormat$1.write(HFileOutputFormat.java:77)
> at
> com.contextweb.hadoop.hbase.mapred.HFileOutputFormat$1.write(HFileOutputFormat.java:49)
> at
> org.apache.hadoop.mapred.ReduceTask$NewTrackingRecordWriter.write(ReduceTask.java:508)
> at
> org.apache.hadoop.mapreduce.TaskInputOutputContext.write(TaskInputOutputContext.java:80)
> at
> org.apache.hadoop.hbase.mapreduce.KeyValueSortReducer.reduce(KeyValueSortReducer.java:46)
> at
> org.apache.hadoop.hbase.mapreduce.KeyValueSortReducer.reduce(KeyValueSortReducer.java:35)
> at org.apache.hadoop.mapreduce.Reducer.run(Reducer.java:176)
> at
> org.apache.hadoop.mapred.ReduceTask.runNewReducer(ReduceTask.java:566)
> at org.apache.hadoop.mapred.ReduceTask.run(ReduceTask.java:408)
> at org.apache.hadoop.mapred.Child.main(Child.java:170)
> {code}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
