BUG-16608: Oozie table import job fails with error where user hive wants to write to table dir owned by the table owner.
Problem: The job fails with the following permission error:
Copying data from hdfs://arpit-falcon-2.cs1cloud.internal:8020/projects/ivory/staging/FALCON_FEED_REPLICATION_raaw-logs16-a6acf050-a038-48d5-9867-de63707291a8_corp-cdd34e35-86b6-45ae-a6cf-d6e879b7b7fb/default/HCatReplication_oneSourceOneTarget_hyphen/dt=2010-01-01-20/2010-01-01-20-00/data/dt=2010-01-01-20 Copying file: hdfs://arpit-falcon-2.cs1cloud.internal:8020/projects/ivory/staging/FALCON_FEED_REPLICATION_raaw-logs16-a6acf050-a038-48d5-9867-de63707291a8_corp-cdd34e35-86b6-45ae-a6cf-d6e879b7b7fb/default/HCatReplication_oneSourceOneTarget_hyphen/dt=2010-01-01-20/2010-01-01-20-00/data/dt=2010-01-01-20/data.txt FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.DDLTask. MetaException(message:Got exception: org.apache.hadoop.security.AccessControlException Permission denied: user=hive, access=WRITE, inode="/tmp/falcon-regression/HCatReplication/HCatReplication_oneSourceOneTarget_hyphen":arpit:hdfs:drwxr-xr-x at org.apache.hadoop.hdfs.server.namenode.FSPermissionChecker.checkFsPermission(FSPermissionChecker.java:265)
BUG-16476: Oozie-Hive tests run as hadoopqa creates/accesses the /tmp/hive-hadoop folder.
Problem: Oozie-Hive tests were run as "hadoopqa" user, concurrently with hcatalog tests. When the tests failed, the HFDFS permissions were as shown below. It is unclear why /tmp/hive-hadoop folder was ever created.
D:\hdp\hadoop-2.4.0.2.1.1.0-1533\bin>hadoop.cmd dfs -ls /tmp drwxr-xr-x - hadoop hdfs 0 2014-04-09 19:01 /tmp/hive-hadoop drwxr-xr-x - hadoopqa hdfs 0 2014-04-09 18:50 /tmp/hive-hadoopqa
BUG-13551: Oozie does not understand _HOST in the Kerberos principal name.
Problem: Oozie currently expects the actual hostname in the kerberos principal. This is unlike other services in the stack, where we can just send _HOST and at run time the service replaces _HOST with machine hostname. This is important so that in a HA setup we can push the same configs to all Oozie servers.
BUG-10177: Oozie workflows that contain Hive queries which run mapreduce jobs fail on secure clusters.
Problem: There is a bug in Hive (HIVE-5618) where delegation tokens are requested for a user who does not have the ability to do so (such as when it is launched from Oozie).
Workaround: Set the configuration parameter before any query statements in the script file are launched as part of the Hive action.
hive.server2.enable.doAs = false
This parameter instructs Hive not to request delegation tokens, which should not be done when running under Oozie.
BUG-9671: Oozie reports the job as failed when the app and job completed successfully when RM is restarted multiple times
Problem: From the Oozie log:
2013-10-05 23:04:58,952 DEBUG HadoopAccessorService:545 - USER[hrt_qa] GROUP[-] TOKEN[] APP[wordcount-wf] JOB[0000003-131005052220011-oozie-oozi-W] ACTION[0000003-131005052220011-oozie-oozi-W@wc] Checking if filesystem hdfs is supported 2013-10-05 23:04:58,954 WARN MapReduceActionExecutor:542 - USER[hrt_qa] GROUP[-] TOKEN[] APP[wordcount-wf] JOB[0000003-131005052220011-oozie-oozi-W] ACTION[0000003-131005052220011-oozie-oozi-W@wc] Launch erMapper died, check Hadoop log for job [hor12n01.gq1.ygridcore.net:8032:job_1381013595258_0001]
But this job and the application complete successfully.