[
https://issues.apache.org/jira/browse/DRILL-2767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14509550#comment-14509550
]
Alexander Zarei commented on DRILL-2767:
----------------------------------------
Update:
When I convert the data to optimized ORC format, I can query all tables back to
back without letting the cluster rest for a bit. That probably suggests there
is different between reading ORC and Text from Hive storage to Drill which we
can focus on to find the root cause.
> Fragment error on TPCH Scale Factor 30 on a query that completed successfully
> previously
> ----------------------------------------------------------------------------------------
>
> Key: DRILL-2767
> URL: https://issues.apache.org/jira/browse/DRILL-2767
> Project: Apache Drill
> Issue Type: Bug
> Components: Storage - Hive
> Affects Versions: 0.8.0
> Environment: AWS EMR cluster of three m1.xlarge nodes
> Reporter: Alexander Zarei
> Assignee: Venki Korukanti
> Attachments: drillbitcore1.log, drillbitcore1.out, drillbitcore2.log,
> drillbitcore2.out, drillbitmaster.out, lineitem table schema .png,
> second-set-core-1-drillbit.log, second-set-core-2-drillbit.log
>
>
> The following sequence led to the error:
> Executed the query
> bq. SELECT * FROM `realhive`.`tpch_text_30`.`lineitem`
> and it took about 43 minutes to execute successfully.
> After ward I ran the query
> bq. SELECT * FROM `realhive`.`tpch_text_2`.`lineitem`
> for 6 times to find an optimization value for the ODBC driver.
> Afterward, I submitted the first query again
> bq. SELECT * FROM `realhive`.`tpch_text_30`.`lineitem`
>
> and the Drill Cluster returned a fragment error.
> bq. ***[HY000]: [MapR][Drill] (1040) Drill failed to execute the query:
> SELECT * FROM `realhive`.`tpch_text_30`.`lineitem`[30024]Query execution
> error. Details:[RemoteRpcException: Failure while running fragment.[
> fb97e7be-d09e-46fe-8728-9577fd0d8795 on ip-10-12-62-65
> Log files with debug level for the Drillbits on the master node as well as
> the core nodes of the cluster are attached.
> Also the connection through the ODBC driver on Linux 32 bit was "Direct" to
> the drillbit on the master node of the Hadoop cluster.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
