Github user andrewor14 commented on a diff in the pull request:
https://github.com/apache/spark/pull/3082#discussion_r19780952
--- Diff:
core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala ---
@@ -124,6 +126,22 @@ private[spark] class ExecutorAllocationManager(sc:
SparkContext) extends Logging
throw new SparkException(s"spark.dynamicAllocation.minExecutors
($minNumExecutors) must " +
s"be less than or equal to spark.dynamicAllocation.maxExecutors
($maxNumExecutors)!")
}
+ // Verify that timeouts are positive
+ if (schedulerBacklogTimeout <= 0) {
+ throw new
SparkException(s"spark.dynamicAllocation.schedulerBacklogTimeout must be > 0!")
+ }
+ if (sustainedSchedulerBacklogTimeout <= 0) {
+ throw new SparkException(
+ s"spark.dynamicAllocation.sustainedSchedulerBacklogTimeout must be
> 0!")
+ }
+ if (executorIdleTimeout <= 0) {
+ throw new
SparkException(s"spark.dynamicAllocation.executorIdleTimeout must be > 0!")
+ }
+ // Verify that external shuffle service is enabled
+ if (!conf.getBoolean("spark.shuffle.service.enabled", false)) {
+ throw new SparkException(s"Dynamic allocation of executors requires
the external " +
+ s"shuffle service. You may enable this through
spark.shuffle.service.enabled.")
--- End diff --
This is tricky, because the user may not have set up an external shuffle
service when running Spark with dynamic allocation. If we automatically enable
this service for them, they will be confused if they see "connection refused"
messages when executors try to fetch files from a service that doesn't exist.
It's not intuitive to me why enabling a dynamic scaling feature will cause my
executors to die because of this.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
