Suppose a spark job has two stages with independent dependencies (they do not depend on each other) and they are submitted concurrently/simultaneously (as Tasksets) by the DAG scheduler to the task scheduler. Can someone give more detailed insight on how the cores available on executors are distributed among the two ready stages/Tasksets? More precisely:
- Tasks from the second taskset/stage are not launched until the tasks of the previous taskset/stage complete? or, - Tasks from both tasksets can be launched (granted cores) simultaneously depending depending on the logic implemented by the taskscheduler e.g. FIFO/Fair? In general, suppose a new resource offer has triggered the taskscheduler to make decision to select some ready tasks (out of n ready taksets) for execution? what is the logic implemented by the taskscheduler in such case? Thanks. -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Spark-stage-concurrency-tp27529.html Sent from the Apache Spark User List mailing list archive at Nabble.com. --------------------------------------------------------------------- To unsubscribe e-mail: user-unsubscr...@spark.apache.org