can we start a new thread in foreachRDD in spark streaming? -
i want start child-thread in foreachrdd.
my situation is:
the job reading hdfs dir continuously, , every 100 batches, want launch model training task (i make snapshot of rdds @ time , start training task. training task takes long time(2 hours), , don't want training task influence reading new batch of data.
is starting new child thread solution? child thread use sparkcontext in main thread , use rdd in main thread?
Comments
Post a Comment