Vsevolod Ostapenko edited comment on HIVE-13280 at 10/12/16 3:08 PM:

Hive-Hbase integration documentation 
(https://cwiki.apache.org/confluence/display/Hive/HBaseIntegration) states that 
hbase.mapred.output.outputtable property is optional, and needed only when one 
wants to insert into a table. The latter statement is obviously incorrect, as 
prior to Feb 26, 2016, this property wasn't even documented and inserts into 
HBase-backed tables were working just fine with MR engine.

If TEZ does require hbase.mapred.output.outputtable property to be explicitly 
set, documentation needs to be updated to indicate that fact.

One more thing, all the existing samples have hbase.mapred.output.outputtable 
and hbase.table.name set to the same value. If there is no use case when they 
are different, why the former even needed?

was (Author: seva_ostapenko):
Hive-Hbase integration documentation 
(https://cwiki.apache.org/confluence/display/Hive/HBaseIntegration) claims that 
hbase.mapred.output.outputtable property is optional, and provides no good 
explanation under what circumstances one would want or need to define it. In 
all the provided samples values of hbase.mapred.output.outputtable and 
hbase.table.name are the same, so samples are hot helpful and not 

If TEZ does require hbase.mapred.output.outputtable property to be explicitly 
set, documentation needs to be updated to indicate that fact.
Also, it would be helpful to provide some background why this property exists 
in the first place.

> Error when more than 1 mapper for HBase storage handler
> -------------------------------------------------------
>                 Key: HIVE-13280
>                 URL: https://issues.apache.org/jira/browse/HIVE-13280
>             Project: Hive
>          Issue Type: Bug
>          Components: HBase Handler
>    Affects Versions: 2.0.0
>            Reporter: Damien Carol
>            Assignee: Damien Carol
> With a simple query (select from orc table and insert into HBase external 
> table):
> {code:sql}
> insert into table register.register  select * from aa_temp
> {code}
> The aa_temp table have 45 orc files. It generate 45 mappers.
> Some mappers fail with this error:
> {noformat}
> Caused by: java.lang.IllegalArgumentException: Must specify table name
>         at 
> org.apache.hadoop.hbase.mapreduce.TableOutputFormat.setConf(TableOutputFormat.java:188)
>         at 
> org.apache.hive.common.util.ReflectionUtil.setConf(ReflectionUtil.java:101)
>         at 
> org.apache.hive.common.util.ReflectionUtil.newInstance(ReflectionUtil.java:87)
>         at 
> org.apache.hadoop.hive.ql.io.HiveFileFormatUtils.getHiveOutputFormat(HiveFileFormatUtils.java:300)
>         at 
> org.apache.hadoop.hive.ql.io.HiveFileFormatUtils.getHiveOutputFormat(HiveFileFormatUtils.java:290)
>         at 
> org.apache.hadoop.hive.ql.exec.FileSinkOperator.createHiveOutputFormat(FileSinkOperator.java:1126)
>         ... 25 more
> ]], Vertex did not succeed due to OWN_TASK_FAILURE, failedTasks:1 
> killedTasks:35, Vertex vertex_1457964631631_0015_3_00 [Map 1] killed/failed 
> due to:OWN_TASK_FAILURE]DAG did not succeed due to VERTEX_FAILURE. 
> failedVertices:1 killedVertices:0 (state=08S01,code=2)
> {noformat}
> If I do an ALTER CONCATENATE for aa_temp. And redo the query. Everything is 
> fine because there are only one mapper.

This message was sent by Atlassian JIRA

Reply via email to