Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[SPARK-4730][YARN] Warn against deprecated YARN settings #3590

Closed
wants to merge 2 commits into from
Closed

[SPARK-4730][YARN] Warn against deprecated YARN settings #3590

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
dcd1316
Warn against deprecated YARN settings
Dec 3, 2014
36e0753
Merge branch 'master' of github.com:apache/spark into yarn-settings
Dec 10, 2014
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
5 changes: 4 additions & 1 deletion core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -212,7 +212,10 @@ private[spark] class SparkSubmitArguments(args: Seq[String], env: Map[String, St
""".stripMargin
}

/** Fill in values by parsing user options. */
/**
* Fill in values by parsing user options.
* NOTE: Any changes here must be reflected in YarnClientSchedulerBackend.
*/
private def parseOpts(opts: Seq[String]): Unit = {
val EQ_SEPARATED_OPT="""(--[^=]+)=(.+)""".r

Expand Down
17 changes: 16 additions & 1 deletion yarn/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientSchedulerBackend.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,7 +65,8 @@ private[spark] class YarnClientSchedulerBackend(
*/
private def getExtraClientArguments: Seq[String] = {
val extraArgs = new ArrayBuffer[String]
val optionTuples = // List of (target Client argument, environment variable, Spark property)
// List of (target Client argument, environment variable, Spark property)
val optionTuples =
List(
("--driver-memory", "SPARK_MASTER_MEMORY", "spark.master.memory"),
("--driver-memory", "SPARK_DRIVER_MEMORY", "spark.driver.memory"),
Expand All @@ -78,11 +79,25 @@ private[spark] class YarnClientSchedulerBackend(
("--queue", "SPARK_YARN_QUEUE", "spark.yarn.queue"),
("--name", "SPARK_YARN_APP_NAME", "spark.app.name")
)
// Warn against the following deprecated environment variables: env var -> suggestion
val deprecatedEnvVars = Map(
"SPARK_MASTER_MEMORY" -> "SPARK_DRIVER_MEMORY or --driver-memory through spark-submit",
"SPARK_WORKER_INSTANCES" -> "SPARK_WORKER_INSTANCES or --num-executors through spark-submit",
"SPARK_WORKER_MEMORY" -> "SPARK_EXECUTOR_MEMORY or --executor-memory through spark-submit",
"SPARK_WORKER_CORES" -> "SPARK_EXECUTOR_CORES or --executor-cores through spark-submit")
// Do the same for deprecated properties: property -> suggestion
val deprecatedProps = Map("spark.master.memory" -> "--driver-memory through spark-submit")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

SPARK_MASTER_MEMORY and spark.master.memory are not applicable in yarn-client mode

should be removed, please refer SPARK-1953

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@zuxqoj the "master" here doesn't actually refer to the ApplicationMaster, but the driver, so SPARK-1953 is actually unrelated. It's really the result of bad naming that we inherited from a long time ago, and this is why we're deprecating it. I agree we should just remove it for 1.4.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@andrewor14 In SPARK-1953 properties related to driver memory were removed form yarn-client mode. SPARK_MASTER_MEMORY and spark.master.memory are related to driver memory, so they should also be removed. I have created jira for it https://issues.apache.org/jira/browse/SPARK-5951

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

SPARK-1953 did not remove any driver memory setting from yarn-client mode; it just introduced new ones for the AM, so it's unrelated to the discussion here. I do agree that we should remove these configs for 1.4, but the reason has nothing to do with the ApplicationMaster. Thanks for filing that JIRA.

optionTuples.foreach { case (optionName, envVar, sparkProp) =>
if (System.getenv(envVar) != null) {
extraArgs += (optionName, System.getenv(envVar))
if (deprecatedEnvVars.contains(envVar)) {
logWarning(s"NOTE: $envVar is deprecated. Use ${deprecatedEnvVars(envVar)} instead.")
}
} else if (sc.getConf.contains(sparkProp)) {
extraArgs += (optionName, sc.getConf.get(sparkProp))
if (deprecatedProps.contains(sparkProp)) {
logWarning(s"NOTE: $sparkProp is deprecated. Use ${deprecatedProps(sparkProp)} instead.")
}
}
}
extraArgs
Expand Down