Re: Spark stuck at removing broadcast variable
This might be obvious but just checking anyways, did you confirm whether or not all of the messages have already been consumed by Spark? If that's the case then I wouldn't expect much to happen unless new data comes into your Kafka topic. If you're a hundred percent sure that there's still plenty more data to be consumed by Spark and that didn't happen, then I would suggest generating Java thread dumps (use Java's jstack command) from your driver's process. On Sat, Apr 18, 2020 at 2:43 PM Sean Owen wrote: > I don't think that means it's stuck on removing something; it was > removed. Not sure what it is waiting on - more data perhaps? > > On Sat, Apr 18, 2020 at 2:22 PM Alchemist > wrote: > > > > I am running a simple Spark structured streaming application that is > pulling data from a Kafka Topic. I have a Kafka Topic with nearly 1000 > partitions. I am running this app on 6 node EMR cluster with 4 cores and > 16GB RAM. I observed that Spark is trying to pull data from all 1024 Kafka > partition and after running successful for few iteration it is stuck with > following exception: > > > > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 101 > > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 66 > > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 77 > > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 78 > > > > 20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on > in memory (size: 4.5 KB, free: 2.7 GB) > > 20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on > ip- in memory (size: 4.5 KB, free: 2.7 GB) > > 20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on > ip- in memory (size: 4.5 KB, free: 2.7 GB) > > Then Sparks show RUNNING but it is NOT Processing any data. > > > > - > To unsubscribe e-mail: user-unsubscr...@spark.apache.org > >
Re: Spark stuck at removing broadcast variable
I don't think that means it's stuck on removing something; it was removed. Not sure what it is waiting on - more data perhaps? On Sat, Apr 18, 2020 at 2:22 PM Alchemist wrote: > > I am running a simple Spark structured streaming application that is pulling > data from a Kafka Topic. I have a Kafka Topic with nearly 1000 partitions. I > am running this app on 6 node EMR cluster with 4 cores and 16GB RAM. I > observed that Spark is trying to pull data from all 1024 Kafka partition and > after running successful for few iteration it is stuck with following > exception: > > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 101 > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 66 > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 77 > 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 78 > > 20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on in > memory (size: 4.5 KB, free: 2.7 GB) > 20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on ip- in > memory (size: 4.5 KB, free: 2.7 GB) > 20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on ip- in > memory (size: 4.5 KB, free: 2.7 GB) > Then Sparks show RUNNING but it is NOT Processing any data. > - To unsubscribe e-mail: user-unsubscr...@spark.apache.org
Spark stuck at removing broadcast variable
I am running a simple Spark structured streaming application that is pulling data from a Kafka Topic. I have a Kafka Topic with nearly 1000 partitions. I am running this app on 6 node EMR cluster with 4 cores and 16GB RAM. I observed that Spark is trying to pull data from all 1024 Kafka partition and after running successful for few iteration it is stuck with following exception: 20/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 10120/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 6620/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 7720/04/18 00:51:41 INFO ContextCleaner: Cleaned accumulator 78 20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on in memory (size: 4.5 KB, free: 2.7 GB)20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on ip- in memory (size: 4.5 KB, free: 2.7 GB)20/04/18 00:51:41 INFO BlockManagerInfo: Removed broadcast_2_piece0 on ip- in memory (size: 4.5 KB, free: 2.7 GB)Then Sparks show RUNNING but it is NOT Processing any data.
Re: Spark structured streaming - performance tuning
Just to clarify - I didn't write this explicitly in my answer. When you're working with Kafka, every partition in Kafka is mapped into Spark partition. And in Spark, every partition is mapped into task. But you can use `coalesce` to decrease the number of Spark partitions, so you'll have less tasks... Srinivas V at "Sat, 18 Apr 2020 10:32:33 +0530" wrote: SV> Thank you Alex. I will check it out and let you know if I have any questions SV> On Fri, Apr 17, 2020 at 11:36 PM Alex Ott wrote: SV> http://shop.oreilly.com/product/0636920047568.do has quite good information SV> on it. For Kafka, you need to start with approximation that processing of SV> each partition is a separate task that need to be executed, so you need to SV> plan number of cores correspondingly. SV> SV> Srinivas V at "Thu, 16 Apr 2020 22:49:15 +0530" wrote: SV> SV> Hello, SV> SV> Can someone point me to a good video or document which takes about performance tuning for structured streaming app? SV> SV> I am looking especially for listening to Kafka topics say 5 topics each with 100 portions . SV> SV> Trying to figure out best cluster size and number of executors and cores required. -- With best wishes,Alex Ott http://alexott.net/ Twitter: alexott_en (English), alexott (Russian) - To unsubscribe e-mail: user-unsubscr...@spark.apache.org