Github user tigerquoll commented on a diff in the pull request:
https://github.com/apache/spark/pull/2516#discussion_r18128856
--- Diff:
core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala ---
@@ -406,22 +412,173 @@ private[spark] class SparkSubmitArguments(args:
Seq[String]) {
}
}
-object SparkSubmitArguments {
- /** Load properties present in the given file. */
- def getPropertiesFromFile(file: File): Seq[(String, String)] = {
- require(file.exists(), s"Properties file $file does not exist")
- require(file.isFile(), s"Properties file $file is not a normal file")
- val inputStream = new FileInputStream(file)
+private[spark] object SparkSubmitArguments {
+ /**
+ * Resolves Configuration sources in order of highest to lowest
+ * 1. Each map passed in as additionalConfig from first to last
+ * 2. Environment variables (including legacy variable mappings)
+ * 3. System config variables (eg by using -Dspark.var.name)
+ * 4 SPARK_DEFAULT_CONF/spark-defaults.conf or
SPARK_HOME/conf/spark-defaults.conf
+ * 5. hard coded defaults in class path at spark-submit-defaults.prop
+ *
+ * A property file specified by one of the means listed above gets read
in and the properties are
+ * considered to be at the priority of the method that specified the
files.
+ * A property specified in a property file will not override an existing
+ * config value at that same level
+ *
+ * @param additionalConfigs Seq of additional
Map[ConfigName->ConfigValue] in order of highest
+ * priority to lowest this will have priority
over internal sources
+ * @return Map[propName->propFile] containing values merged from all
sources in order of priority
+ */
+ def mergeSparkProperties(additionalConfigs: Seq [Map[String,String]]) = {
+ // Configuration read in from spark-submit-defaults.prop file found on
the classpath
+ var hardCodedDefaultConfig: Option[Map[String,String]] = None
+ var is: InputStream = null
+ var isr: Option[InputStreamReader] = None
try {
- val properties = new Properties()
- properties.load(inputStream)
- properties.stringPropertyNames().toSeq.map(k => (k,
properties(k).trim))
- } catch {
- case e: IOException =>
- val message = s"Failed when loading Spark properties file $file"
- throw new SparkException(message, e)
+ is =
Thread.currentThread().getContextClassLoader.getResourceAsStream(ClassPathSparkSubmitDefaults)
+
+ // only open InputStreamReader if InputStream was successfully opened
+ isr = Option(is).map{is: InputStream =>
+ new InputStreamReader(is, CharEncoding.UTF_8)
+ }
+
+ hardCodedDefaultConfig = isr.map( defaultValueStream =>
+
SparkSubmitArguments.getPropertyValuesFromStream(defaultValueStream))
} finally {
- inputStream.close()
+ Option(is).foreach(_.close)
+ isr.foreach(_.close)
}
+
+ if (hardCodedDefaultConfig.isEmpty || (hardCodedDefaultConfig.get.size
== 0)) {
+ throw new IllegalStateException(s"Default values not found at
classpath $ClassPathSparkSubmitDefaults")
+ }
+
+ // Configuration read in from defaults file if it exists
+ var sparkDefaultConfig = SparkSubmitArguments.getSparkDefaultFileConfig
+
+ if (sparkDefaultConfig.isDefinedAt(SparkPropertiesFile)) {
+ SparkSubmitArguments.getPropertyValuesFromFile(
+ sparkDefaultConfig.get(SparkPropertiesFile).get)
+ } else {
+ Map.empty
+ }
+
+ // Configuration from java system properties
+ val systemPropertyConfig =
SparkSubmitArguments.getPropertyMap(System.getProperties)
+
+ // Configuration variables from the environment
+ // support legacy variables
+ val environmentConfig = System.getenv().asScala
+
+ val legacyEnvVars = Seq("MASTER"->SparkMaster,
"DEPLOY_MODE"->SparkDeployMode,
+ "SPARK_DRIVER_MEMORY"->SparkDriverMemory,
"SPARK_EXECUTOR_MEMORY"->SparkExecutorMemory)
+
+ // legacy variables act at the priority of a system property
+ val propsWithEnvVars : mutable.Map[String,String] = new
mutable.HashMap() ++ systemPropertyConfig ++ legacyEnvVars
+ .map( {case(varName, propName) => (environmentConfig.get(varName),
propName) })
+ .filter( {case(varVariable, _) => varVariable.isDefined &&
!varVariable.get.isEmpty} )
+ .map{case(varVariable, propName) => (propName, varVariable.get)}
+
+ val ConfigSources = additionalConfigs ++ Seq (
+ environmentConfig,
+ propsWithEnvVars,
+ sparkDefaultConfig,
+ hardCodedDefaultConfig.get
+ )
+
+ // Load properties file at priority level of source that specified the
property file
--- End diff --
Hi @vanzin, I believe the existing behaviour is driven by environment
variables and files in the file system as well as command line options (though
command line options seem to have priority (the existing code base treats
priority on a variable-by-variable basis, so It can be easy to miss exceptions
to the rule).
if $SPARK_CONF_DIR/spark-defaults.conf iexists then it is loaded into a
hashmap. if SPARK_HOME/conf.spark-defaults.conf exists then it will replace the
above mentioned hashmap.
What ever is left is then overridden by what ever --conf properties have
been specified no the command line.
These values are then overridden by various command line specified options.
Finally legacy environment variables are checked and config values assigned
if they have yet to be assigned.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
