From cd739bd756875bd52e9bd8ae801e0ae10a1f6937 Mon Sep 17 00:00:00 2001
From: GuoQiang Li <witgo@qq.com>
Date: Wed, 29 Oct 2014 23:02:58 -0700
Subject: [SPARK-1720][SPARK-1719] use LD_LIBRARY_PATH instead of
 -Djava.library.path

- [X] Standalone
- [X] YARN
- [X] Mesos
- [X]  Mac OS X
- [X] Linux
- [ ]  Windows

This is another implementation about #1031

Author: GuoQiang Li <witgo@qq.com>

Closes #2711 from witgo/SPARK-1719 and squashes the following commits:

c7b26f6 [GuoQiang Li] review commits
4488e41 [GuoQiang Li] Refactoring CommandUtils
a444094 [GuoQiang Li] review commits
40c0b4a [GuoQiang Li] Add buildLocalCommand method
c1a0ddd [GuoQiang Li] fix comments
156ce88 [GuoQiang Li] review commit
38aa377 [GuoQiang Li] Refactor CommandUtils.scala
4269e00 [GuoQiang Li] Refactor SparkSubmitDriverBootstrapper.scala
7a1d634 [GuoQiang Li] use LD_LIBRARY_PATH instead of -Djava.library.path
---
 .../scala/org/apache/spark/deploy/yarn/ClientBase.scala    | 14 +++++++++++---
 .../apache/spark/deploy/yarn/ExecutorRunnableUtil.scala    | 11 ++++++++++-
 2 files changed, 21 insertions(+), 4 deletions(-)

(limited to 'yarn/common')

diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala
index 8ea0e7cf40..f95d723791 100644
--- a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala
@@ -39,6 +39,7 @@ import org.apache.hadoop.yarn.conf.YarnConfiguration
 import org.apache.hadoop.yarn.util.Records
 
 import org.apache.spark.{Logging, SecurityManager, SparkConf, SparkContext, SparkException}
+import org.apache.spark.util.Utils
 
 /**
  * The entry point (starting in Client#main() and Client#run()) for launching Spark on YARN.
@@ -312,6 +313,10 @@ private[spark] trait ClientBase extends Logging {
 
     val javaOpts = ListBuffer[String]()
 
+    // Set the environment variable through a command prefix
+    // to append to the existing value of the variable
+    var prefixEnv: Option[String] = None
+
     // Add Xmx for AM memory
     javaOpts += "-Xmx" + args.amMemory + "m"
 
@@ -348,8 +353,11 @@ private[spark] trait ClientBase extends Logging {
       sparkConf.getOption("spark.driver.extraJavaOptions")
         .orElse(sys.env.get("SPARK_JAVA_OPTS"))
         .foreach(opts => javaOpts += opts)
-      sparkConf.getOption("spark.driver.libraryPath")
-        .foreach(p => javaOpts += s"-Djava.library.path=$p")
+      val libraryPaths = Seq(sys.props.get("spark.driver.extraLibraryPath"),
+        sys.props.get("spark.driver.libraryPath")).flatten
+      if (libraryPaths.nonEmpty) {
+        prefixEnv = Some(Utils.libraryPathEnvPrefix(libraryPaths))
+      }
     }
 
     // For log4j configuration to reference
@@ -384,7 +392,7 @@ private[spark] trait ClientBase extends Logging {
         "--num-executors ", args.numExecutors.toString)
 
     // Command for the ApplicationMaster
-    val commands = Seq(Environment.JAVA_HOME.$() + "/bin/java", "-server") ++
+    val commands = prefixEnv ++ Seq(Environment.JAVA_HOME.$() + "/bin/java", "-server") ++
       javaOpts ++ amArgs ++
       Seq(
         "1>", ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stdout",
diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ExecutorRunnableUtil.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ExecutorRunnableUtil.scala
index 5cb4753de2..88dad0febd 100644
--- a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ExecutorRunnableUtil.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ExecutorRunnableUtil.scala
@@ -30,6 +30,7 @@ import org.apache.hadoop.yarn.conf.YarnConfiguration
 import org.apache.hadoop.yarn.util.{ConverterUtils, Records}
 
 import org.apache.spark.{Logging, SparkConf}
+import org.apache.spark.util.Utils
 
 trait ExecutorRunnableUtil extends Logging {
 
@@ -47,6 +48,11 @@ trait ExecutorRunnableUtil extends Logging {
       localResources: HashMap[String, LocalResource]): List[String] = {
     // Extra options for the JVM
     val javaOpts = ListBuffer[String]()
+
+    // Set the environment variable through a command prefix
+    // to append to the existing value of the variable
+    var prefixEnv: Option[String] = None
+
     // Set the JVM memory
     val executorMemoryString = executorMemory + "m"
     javaOpts += "-Xms" + executorMemoryString + " -Xmx" + executorMemoryString + " "
@@ -58,6 +64,9 @@ trait ExecutorRunnableUtil extends Logging {
     sys.env.get("SPARK_JAVA_OPTS").foreach { opts =>
       javaOpts += opts
     }
+    sys.props.get("spark.executor.extraLibraryPath").foreach { p =>
+      prefixEnv = Some(Utils.libraryPathEnvPrefix(Seq(p)))
+    }
 
     javaOpts += "-Djava.io.tmpdir=" +
       new Path(Environment.PWD.$(), YarnConfiguration.DEFAULT_CONTAINER_TEMP_DIR)
@@ -101,7 +110,7 @@ trait ExecutorRunnableUtil extends Logging {
     // For log4j configuration to reference
     javaOpts += ("-Dspark.yarn.app.container.log.dir=" + ApplicationConstants.LOG_DIR_EXPANSION_VAR)
 
-    val commands = Seq(Environment.JAVA_HOME.$() + "/bin/java",
+    val commands = prefixEnv ++ Seq(Environment.JAVA_HOME.$() + "/bin/java",
       "-server",
       // Kill if OOM is raised - leverage yarn's failure handling to cause rescheduling.
       // Not killing the task leaves various aspects of the executor and (to some extent) the jvm in
-- 
cgit v1.2.3