Creating parquet Petastorm dataset through Spark fails with Overflow error (larger than 4GB)












0















I'm trying to implement Uber's Petastorm dataset creation which utilizes Spark to create a parquet file following the tutorial on their Github page.



The code:



spark = SparkSession.builder.config('spark.driver.memory', '10g').master('local[4]').getOrCreate()
sc = spark.sparkContext

with materialize_dataset(spark=spark, dataset_url='file:///opt/data/hello_world_dataset',
schema=MySchema, row_group_size_mb=256):

logging.info('Building RDD...')
rows_rdd = sc.parallelize(ids)
.map(row_generator) # Generator that yields lists of examples
.flatMap(lambda x: dict_to_spark_row(MySchema, x))

logging.info('Creating DataFrame...')
spark.createDataFrame(rows_rdd, MySchema.as_spark_schema())
.coalesce(10)
.write
.mode('overwrite')
.parquet('file:///opt/data/hello_world_dataset')


Now the RDD code executes successfully but fails only the .createDataFrame call with the following error:




_pickle.PicklingError: Could not serialize broadcast: OverflowError: cannot serialize a string larger than 4GiB




This is my first experience with Spark, so I can't really tell if this error originates in Spark or Petastorm.



Looking through other solutions to this error (in respect to Spark, not Petastorm) I saw that it might have to do with the pickling protocol, but I can't confirm that, neither did I find a way of altering the pickling protocol.



How could I avoid this error?










share|improve this question





























    0















    I'm trying to implement Uber's Petastorm dataset creation which utilizes Spark to create a parquet file following the tutorial on their Github page.



    The code:



    spark = SparkSession.builder.config('spark.driver.memory', '10g').master('local[4]').getOrCreate()
    sc = spark.sparkContext

    with materialize_dataset(spark=spark, dataset_url='file:///opt/data/hello_world_dataset',
    schema=MySchema, row_group_size_mb=256):

    logging.info('Building RDD...')
    rows_rdd = sc.parallelize(ids)
    .map(row_generator) # Generator that yields lists of examples
    .flatMap(lambda x: dict_to_spark_row(MySchema, x))

    logging.info('Creating DataFrame...')
    spark.createDataFrame(rows_rdd, MySchema.as_spark_schema())
    .coalesce(10)
    .write
    .mode('overwrite')
    .parquet('file:///opt/data/hello_world_dataset')


    Now the RDD code executes successfully but fails only the .createDataFrame call with the following error:




    _pickle.PicklingError: Could not serialize broadcast: OverflowError: cannot serialize a string larger than 4GiB




    This is my first experience with Spark, so I can't really tell if this error originates in Spark or Petastorm.



    Looking through other solutions to this error (in respect to Spark, not Petastorm) I saw that it might have to do with the pickling protocol, but I can't confirm that, neither did I find a way of altering the pickling protocol.



    How could I avoid this error?










    share|improve this question



























      0












      0








      0








      I'm trying to implement Uber's Petastorm dataset creation which utilizes Spark to create a parquet file following the tutorial on their Github page.



      The code:



      spark = SparkSession.builder.config('spark.driver.memory', '10g').master('local[4]').getOrCreate()
      sc = spark.sparkContext

      with materialize_dataset(spark=spark, dataset_url='file:///opt/data/hello_world_dataset',
      schema=MySchema, row_group_size_mb=256):

      logging.info('Building RDD...')
      rows_rdd = sc.parallelize(ids)
      .map(row_generator) # Generator that yields lists of examples
      .flatMap(lambda x: dict_to_spark_row(MySchema, x))

      logging.info('Creating DataFrame...')
      spark.createDataFrame(rows_rdd, MySchema.as_spark_schema())
      .coalesce(10)
      .write
      .mode('overwrite')
      .parquet('file:///opt/data/hello_world_dataset')


      Now the RDD code executes successfully but fails only the .createDataFrame call with the following error:




      _pickle.PicklingError: Could not serialize broadcast: OverflowError: cannot serialize a string larger than 4GiB




      This is my first experience with Spark, so I can't really tell if this error originates in Spark or Petastorm.



      Looking through other solutions to this error (in respect to Spark, not Petastorm) I saw that it might have to do with the pickling protocol, but I can't confirm that, neither did I find a way of altering the pickling protocol.



      How could I avoid this error?










      share|improve this question
















      I'm trying to implement Uber's Petastorm dataset creation which utilizes Spark to create a parquet file following the tutorial on their Github page.



      The code:



      spark = SparkSession.builder.config('spark.driver.memory', '10g').master('local[4]').getOrCreate()
      sc = spark.sparkContext

      with materialize_dataset(spark=spark, dataset_url='file:///opt/data/hello_world_dataset',
      schema=MySchema, row_group_size_mb=256):

      logging.info('Building RDD...')
      rows_rdd = sc.parallelize(ids)
      .map(row_generator) # Generator that yields lists of examples
      .flatMap(lambda x: dict_to_spark_row(MySchema, x))

      logging.info('Creating DataFrame...')
      spark.createDataFrame(rows_rdd, MySchema.as_spark_schema())
      .coalesce(10)
      .write
      .mode('overwrite')
      .parquet('file:///opt/data/hello_world_dataset')


      Now the RDD code executes successfully but fails only the .createDataFrame call with the following error:




      _pickle.PicklingError: Could not serialize broadcast: OverflowError: cannot serialize a string larger than 4GiB




      This is my first experience with Spark, so I can't really tell if this error originates in Spark or Petastorm.



      Looking through other solutions to this error (in respect to Spark, not Petastorm) I saw that it might have to do with the pickling protocol, but I can't confirm that, neither did I find a way of altering the pickling protocol.



      How could I avoid this error?







      python pyspark petastorm






      share|improve this question















      share|improve this question













      share|improve this question




      share|improve this question








      edited Nov 21 '18 at 10:04







      bluesummers

















      asked Nov 19 '18 at 8:51









      bluesummersbluesummers

      2,23612043




      2,23612043
























          1 Answer
          1






          active

          oldest

          votes


















          0














          The problem lies in the pickling that is done to pass the data between the different processes, the default pickling protocol is 2, and we need to use 4 in order to pass objects larger than 4GB.



          To change the pickling protocol, before creation a Spark session, use the following code



          from pyspark import broadcast
          import pickle


          def broadcast_dump(self, value, f):
          pickle.dump(value, f, 4) # was 2, 4 is first protocol supporting >4GB
          f.close()

          return f.name


          broadcast.Broadcast.dump = broadcast_dump





          share|improve this answer























            Your Answer






            StackExchange.ifUsing("editor", function () {
            StackExchange.using("externalEditor", function () {
            StackExchange.using("snippets", function () {
            StackExchange.snippets.init();
            });
            });
            }, "code-snippets");

            StackExchange.ready(function() {
            var channelOptions = {
            tags: "".split(" "),
            id: "1"
            };
            initTagRenderer("".split(" "), "".split(" "), channelOptions);

            StackExchange.using("externalEditor", function() {
            // Have to fire editor after snippets, if snippets enabled
            if (StackExchange.settings.snippets.snippetsEnabled) {
            StackExchange.using("snippets", function() {
            createEditor();
            });
            }
            else {
            createEditor();
            }
            });

            function createEditor() {
            StackExchange.prepareEditor({
            heartbeatType: 'answer',
            autoActivateHeartbeat: false,
            convertImagesToLinks: true,
            noModals: true,
            showLowRepImageUploadWarning: true,
            reputationToPostImages: 10,
            bindNavPrevention: true,
            postfix: "",
            imageUploader: {
            brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
            contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
            allowUrls: true
            },
            onDemand: true,
            discardSelector: ".discard-answer"
            ,immediatelyShowMarkdownHelp:true
            });


            }
            });














            draft saved

            draft discarded


















            StackExchange.ready(
            function () {
            StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f53371112%2fcreating-parquet-petastorm-dataset-through-spark-fails-with-overflow-error-larg%23new-answer', 'question_page');
            }
            );

            Post as a guest















            Required, but never shown

























            1 Answer
            1






            active

            oldest

            votes








            1 Answer
            1






            active

            oldest

            votes









            active

            oldest

            votes






            active

            oldest

            votes









            0














            The problem lies in the pickling that is done to pass the data between the different processes, the default pickling protocol is 2, and we need to use 4 in order to pass objects larger than 4GB.



            To change the pickling protocol, before creation a Spark session, use the following code



            from pyspark import broadcast
            import pickle


            def broadcast_dump(self, value, f):
            pickle.dump(value, f, 4) # was 2, 4 is first protocol supporting >4GB
            f.close()

            return f.name


            broadcast.Broadcast.dump = broadcast_dump





            share|improve this answer




























              0














              The problem lies in the pickling that is done to pass the data between the different processes, the default pickling protocol is 2, and we need to use 4 in order to pass objects larger than 4GB.



              To change the pickling protocol, before creation a Spark session, use the following code



              from pyspark import broadcast
              import pickle


              def broadcast_dump(self, value, f):
              pickle.dump(value, f, 4) # was 2, 4 is first protocol supporting >4GB
              f.close()

              return f.name


              broadcast.Broadcast.dump = broadcast_dump





              share|improve this answer


























                0












                0








                0







                The problem lies in the pickling that is done to pass the data between the different processes, the default pickling protocol is 2, and we need to use 4 in order to pass objects larger than 4GB.



                To change the pickling protocol, before creation a Spark session, use the following code



                from pyspark import broadcast
                import pickle


                def broadcast_dump(self, value, f):
                pickle.dump(value, f, 4) # was 2, 4 is first protocol supporting >4GB
                f.close()

                return f.name


                broadcast.Broadcast.dump = broadcast_dump





                share|improve this answer













                The problem lies in the pickling that is done to pass the data between the different processes, the default pickling protocol is 2, and we need to use 4 in order to pass objects larger than 4GB.



                To change the pickling protocol, before creation a Spark session, use the following code



                from pyspark import broadcast
                import pickle


                def broadcast_dump(self, value, f):
                pickle.dump(value, f, 4) # was 2, 4 is first protocol supporting >4GB
                f.close()

                return f.name


                broadcast.Broadcast.dump = broadcast_dump






                share|improve this answer












                share|improve this answer



                share|improve this answer










                answered Nov 21 '18 at 10:05









                bluesummersbluesummers

                2,23612043




                2,23612043






























                    draft saved

                    draft discarded




















































                    Thanks for contributing an answer to Stack Overflow!


                    • Please be sure to answer the question. Provide details and share your research!

                    But avoid



                    • Asking for help, clarification, or responding to other answers.

                    • Making statements based on opinion; back them up with references or personal experience.


                    To learn more, see our tips on writing great answers.




                    draft saved


                    draft discarded














                    StackExchange.ready(
                    function () {
                    StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f53371112%2fcreating-parquet-petastorm-dataset-through-spark-fails-with-overflow-error-larg%23new-answer', 'question_page');
                    }
                    );

                    Post as a guest















                    Required, but never shown





















































                    Required, but never shown














                    Required, but never shown












                    Required, but never shown







                    Required, but never shown

































                    Required, but never shown














                    Required, but never shown












                    Required, but never shown







                    Required, but never shown







                    Popular posts from this blog

                    Guess what letter conforming each word

                    Port of Spain

                    Run scheduled task as local user group (not BUILTIN)