[
https://issues.apache.org/jira/browse/HIVE-13032?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15140196#comment-15140196
]
Prasanth Jayachandran commented on HIVE-13032:
----------------------------------------------
Verified that it's not included in other services. Beeline uses different
properties file and sets log4j.configurationFile again. Also double
initialization is not a problem as logger factory is initialized in a static
block which will be initialized automatically by the class loader. Using
Configurator to initialize again will return the same instance. Committing this
patch shortly.
> Hive services need HADOOP_CLIENT_OPTS for proper log4j2 initialization
> ----------------------------------------------------------------------
>
> Key: HIVE-13032
> URL: https://issues.apache.org/jira/browse/HIVE-13032
> Project: Hive
> Issue Type: Bug
> Affects Versions: 2.0.0, 2.1.0
> Reporter: Prasanth Jayachandran
> Assignee: Prasanth Jayachandran
> Priority: Blocker
> Attachments: HIVE-13032.1.patch, HIVE-13032.2.patch
>
>
> HIVE-12497 removed HADOOP_CLIENT_OPTS as it slowed down cli launch time. But
> it leads to log4j2 not being initialized when using services other than CLI.
> Other services like metastore, schematool etc. rely on log4j to initialize
> the logging based on the presence of log4j2.properties file in the classpath.
> If we use the standard name for log4j configuration file (log4j2.properties)
> then automatic initialization will happen. If not, we have to tell log4j to
> look for specific properties file. This is done via -Dlog4j.configurationFile
> system property. If we pass this system property via HADOOP_CLIENT_OPTS then
> all hive services will have logging initialized properly.
> In HIVE-12497, the problem was we had HADOOP_CLIENT_OPTS at the top of the
> script. As a result, hadoop and hbase commands tries to initialize logging
> which took long time slowing down the startup time.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
