[
https://issues.apache.org/jira/browse/FLINK-2954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15028613#comment-15028613
]
ASF GitHub Bot commented on FLINK-2954:
---------------------------------------
Github user mxm commented on a diff in the pull request:
https://github.com/apache/flink/pull/1409#discussion_r45970338
--- Diff:
flink-core/src/main/java/org/apache/flink/configuration/ConfigConstants.java ---
@@ -243,6 +243,20 @@
*/
public static final String YARN_PROPERTIES_FILE_LOCATION =
"yarn.properties-file.location";
+ /**
+ * Prefix for passing custom environment variables to Flink's
ApplicationMaster (JobManager).
+ * For example for passing LD_LIBRARY_PATH as an env variable to the
AppMaster, set:
+ * yarn.application-master.env.LD_LIBRARY_PATH: "/usr/lib/native"
+ * in the flink-conf.yaml.
+ */
+ public static final String YARN_APPLICATION_MASTER_ENV_PREFIX =
"yarn.application-master.env.";
+
+ /**
+ * Similar to the {@see YARN_APPLICATION_MASTER_ENV_PREFIX}, this
configuration prefix allows
+ * setting custom environment variables.
+ */
+ public static final String YARN_TASK_MANAGER_ENV_PREFIX =
"yarn.taskmanager.env.";
--- End diff --
I wonder about the naming here. Maybe this should be
`YARN_RESOURCE_MANAGER_ENV_PREFIX`? Or change
`YARN_APPLICATION_MASTER_ENV_PREFIX` to `YARN_JOB_MANAGER_ENV_PREFIX`?
> Not able to pass custom environment variables in cluster to processes that
> spawning TaskManager
> -----------------------------------------------------------------------------------------------
>
> Key: FLINK-2954
> URL: https://issues.apache.org/jira/browse/FLINK-2954
> Project: Flink
> Issue Type: Bug
> Components: Command-line client, Distributed Runtime
> Affects Versions: 0.10.0
> Reporter: Jian Jiang
> Assignee: Robert Metzger
> Priority: Critical
> Fix For: 1.0.0
>
>
> There are programs that rely on custom environment variables. In hadoop
> mapreduce job we can use -Dmapreduce.map.env and - Dmapreduce.reduce.env to
> do pass them. Similarly in Spark
> we can use --conf 'spark.executor.XXX=value for XXX'. There is no such
> feature yet in Flink.
> This has given Flink a serious disadvantage when customers need such feature.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
