Hive issue using yarn - hadoop

Question

I am running hive sql on yarn,
it's throwing error with join condition , I am able to create External as well as internal table but failed to create table when use command
create table as AS SELECT name from student.
when running same query through hive cli it's working fine but with spring jog it throws error
2016-03-28 04:26:50,692 [Thread-17] WARN
org.apache.hadoop.hive.shims.HadoopShimsSecure - Can't fetch tasklog:
TaskLogServlet is not supported in MR2 mode.
Task with the most failures(4):
-----
Task ID:
task_1458863269455_90083_m_000638
-----
Diagnostic Messages for this Task:
AttemptID:attempt_1458863269455_90083_m_000638_3 Timed out after 1 secs
2016-03-28 04:26:50,842 [main] INFO
org.apache.hadoop.yarn.client.api.impl.YarnClientImpl - Killed application
application_1458863269455_90083
2016-03-28 04:26:50,849 [main] ERROR com.mapr.fs.MapRFileSystem - Failed to
delete path maprfs:/home/pro/amit/warehouse/scratdir/hive_2016-03-28_04-
24-32_038_8553676376881087939-1/_task_tmp.-mr-10003, error: No such file or
directory (2)
2016-03-28 04:26:50,852 [main] ERROR org.apache.hadoop.hive.ql.Driver -
FAILED: Execution Error, return code 2 from
As per my findings I think there is some issue with scratdir.
Kindly suggest if any one face same issue.

Answer 1

This issue occurs if the recursive directory doesnot exist. Hive doesnt automatically create directories recursively.
Please check existence of directories to child\table level from root

Answer 2

I faced a similar issue while running the below Hive query
select * from <db_name>.<internal_tbl_name> where <field_name_of_double_type> in (<list_of_double_values>) order by <list_of_order_fields> limit 10;
I performed an explain on the above statement and below was the result.
fs.FileUtil: Failed to delete file or dir [/hdfs/Hadoop_Misc_Logs/Edge01/local_scratch/<hive_username>/41289638-cd53-4d4b-88c9-3359e9ec99e2/hive_2017-05-08_04-26-36_658_6626096693992380903-1/.nfs0000000057b93e2d00001590]: it still exists.
2017-05-08 04:26:37,969 WARN [41289638-cd53-4d4b-88c9-3359e9ec99e2 main] fs.FileUtil: Failed to delete file or dir [/hdfs/Hadoop_Misc_Logs/Edge01/local_scratch/<hive_username>/41289638-cd53-4d4b-88c9-3359e9ec99e2/hive_2017-05-08_04-26-36_658_6626096693992380903-1/.nfs0000000057b93e2700001591]: it still exists.
Time taken: 0.886 seconds, Fetched: 24 row(s)
And checked the logs through
yarn logs -applicationID application_1458863269455_90083
The error happened after a MapR upgrade from the admin team. It is probably due to some upgrade or installation issue and Tez configurations (as suggested by the line 873 in log below). Or probably, the Hive query is syntactically not supporting the Tez optimization. Saying so, because another Hive query on an external table is running fine in my case. Have to check a bit deeper though.
Though not sure but the error line in the logs that looks to be most relevant is as follows:
2017-05-08 00:01:47,873 [ERROR] [main] |web.WebUIService|: Tez UI History URL is not set
Solution:
It is probably happening due to some open files or applications that are using some resources. Pls check https://unix.stackexchange.com/questions/11238/how-to-get-over-device-or-resource-busy
You can run the explain <your_Hive_statement>
In the result execution plan, you can come across the filenames/dirs that Hive execution engine fails to delete e.g.
2017-05-08 04:26:37,969 WARN [41289638-cd53-4d4b-88c9-3359e9ec99e2 main] fs.FileUtil: Failed to delete file or dir [/hdfs/Hadoop_Misc_Logs/Edge01/local_scratch/<hive_username>/41289638-cd53-4d4b-88c9-3359e9ec99e2/hive_2017-05-08_04-26-36_658_6626096693992380903-1/.nfs0000000057b93e2d00001590]: it still exists.
Go to the path given in the step 2 e.g. /hdfs/Hadoop_Misc_Logs/Edge01/local_scratch/<hive_username>/41289638-cd53-4d4b-88c9-3359e9ec99e2/hive_2017-05-08_04-26-36_658_6626096693992380903-1/
In path 3, doing ls -a or lsof +D /path will show the open process_ids blocking the files from delete.
If you run ps -ef | grep <pid>, you get
hive_username <pid> 19463 1 05:19 pts/8 00:00:35 /opt/mapr/tools/jdk1.7.0_51/jre/bin/java -Xmx256m -Dhiveserver2.auth=PAM -Dhiveserver2.authentication.pam.services=login -Dmapr_sec_enabled=true -Dhadoop.login=maprsasl -Djava.net.preferIPv4Stack=true -Dhadoop.log.dir=/opt/mapr/hadoop/hadoop-2.7.0/logs -Dhadoop.log.file=hadoop.log -Dhadoop.home.dir=/opt/mapr/hadoop/hadoop-2.7.0 -Dhadoop.id.str=hive_username -Dhadoop.root.logger=INFO,console -Djava.library.path=/opt/mapr/hadoop/hadoop-2.7.0/lib/native -Dhadoop.policy.file=hadoop-policy.xml -Djava.net.preferIPv4Stack=true -Xmx512m -Dlog4j.configurationFile=hive-log4j2.properties -Dlog4j.configurationFile=hive-log4j2.properties -Djava.util.logging.config.file=/opt/mapr/hive/hive-2.1/bin/../conf/parquet-logging.properties -Dhadoop.security.logger=INFO,NullAppender -Djava.security.auth.login.config=/opt/mapr/conf/mapr.login.conf -Dzookeeper.saslprovider=com.mapr.security.maprsasl.MaprSaslProvider -Djavax.net.ssl.trustStore=/opt/mapr/conf/ssl_truststore org.apache.hadoop.util.RunJar /opt/mapr/hive/hive-2.1//lib/hive-cli-2.1.1-mapr-1703.jar org.apache.hadoop.hive.cli.CliDriver
CONCLUSION:
The HiveCLiDriver clearly shows that running "Hive on Spark" (or managed) tables through Hive CLI is not supported any more from Hive 2.0 onwards and it is going to be deprecated going forward. You have to use HiveContext in Spark for running Hive queries. But you can still run queries on Hive external tables through Hive CLI.