Hi Bryan, I think the ContextCleaner will take care of the broadcasted variables, see i.e.,
https://jaceklaskowski.gitbooks.io/mastering-apache-spark/content/spark-service-contextcleaner.html If it is easy to spot when to cleanup the broadcast variables in your case, a "xBroadcasted.destroy()" will make things more explicit. Best, Anastasios On Mon, Jan 2, 2017 at 4:26 PM, <bryan.jeff...@gmail.com> wrote: > All, > > > > Anyone have a thought? > > > > Thank you, > > > > Bryan Jeffrey > > > > *From: *bryan.jeff...@gmail.com > *Sent: *Friday, December 30, 2016 1:20 PM > *To: *user <user@spark.apache.org> > *Subject: *Broadcast destroy > > > > All, > > > > If we are updating broadcast variables do we need to manually destroy the > replaced broadcast, or will they be automatically pruned? > > > > Thank you, > > > > Bryan Jeffrey > > > > Sent from my Windows 10 phone > > > > > -- -- Anastasios Zouzias <a...@zurich.ibm.com>