5

我有 pyspark 数据框，其维度为 (28002528,21) 并尝试使用以下代码行将其转换为 pandas 数据框：

pd_df=spark_df.toPandas()

我收到了这个错误：

第一部分

Py4JJavaError: An error occurred while calling o170.collectToPython.
: org.apache.spark.SparkException: Job aborted due to stage failure: Task 3 in stage 39.0 failed 1 times, most recent failure: Lost task 3.0 in stage 39.0 (TID 89, localhost, executor driver): java.lang.OutOfMemoryError: Java heap space
    at java.util.Arrays.copyOf(Arrays.java:3236)
    at java.io.ByteArrayOutputStream.grow(ByteArrayOutputStream.java:118)
    at java.io.ByteArrayOutputStream.ensureCapacity(ByteArrayOutputStream.java:93)
    at java.io.ByteArrayOutputStream.write(ByteArrayOutputStream.java:153)
    at net.jpountz.lz4.LZ4BlockOutputStream.flushBufferedData(LZ4BlockOutputStream.java:220)
    at net.jpountz.lz4.LZ4BlockOutputStream.write(LZ4BlockOutputStream.java:173)
    at java.io.DataOutputStream.write(DataOutputStream.java:107)
    at org.apache.spark.sql.catalyst.expressions.UnsafeRow.writeToStream(UnsafeRow.java:552)
    at org.apache.spark.sql.execution.SparkPlan$$anonfun$2.apply(SparkPlan.scala:256)
    at org.apache.spark.sql.execution.SparkPlan$$anonfun$2.apply(SparkPlan.scala:247)
    at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$25.apply(RDD.scala:830)
    at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$25.apply(RDD.scala:830)
    at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
    at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
    at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
    at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
    at org.apache.spark.scheduler.Task.run(Task.scala:109)
    at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:345)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:748)


Driver stacktrace:
        at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$failJobAndIndependentStages(DAGScheduler.scala:1599)
        ...
        ...

Caused by: java.lang.OutOfMemoryError: Java heap space
        ...
        ...

第二部分 

Exception happened during processing of request from ('127.0.0.1', 56842)
ERROR:py4j.java_gateway:An error occurred while trying to connect to the Java server (127.0.0.1:56657)
Traceback (most recent call last):
        ...
        ...    
ConnectionResetError: [WinError 10054] An existing connection was forcibly closed by the remote host

During handling of the above exception, another exception occurred:
        ...
        ...

我还尝试对原始 pyspark 数据框进行采样

smaple_pd_df=spark_df.sample(0.05).toPandas()

我收到一个错误，看起来只是上一个错误的第一部分

4

2 回答 2

1

您得到 java.lang.OutOfMemoryError这可能意味着您正在尝试将所有数据加载到没有足够 RAM 来处理整个 DataFrame 的单个节点中。如果您使用 Databricks 等云解决方案提供商，请尝试增加集群 RAM 的大小。

于 2019-02-25T13:15:24.817 回答
1

什么toPandas()是将整个数据帧收集到一个节点中（如@ulmefors 的回答中所述）。

更具体地说，它将它收集给驱动程序。您应该微调的具体选项是spark.driver.memory，相应地增加它。

否则，如果您打算对这个（相当大的）pandas 数据帧进行进一步的转换，您可以考虑先在 pyspark 中进行转换，然后将（较小的）结果收集到驱动程序中，希望这将适合内存。

Spark 配置文档中提供了更多详细信息，请点击此处。

于 2019-04-05T17:17:34.803 回答