在尝试使用 Spark DataFrame API 联接(内部)一个较小的数据集(strong MB)(654 GB)和较小的数据集(535 MB)时,我遇到了问题。 / p>
我正在使用broadcast()函数将较小的数据集广播到工作节点。
我无法在这两个数据集之间进行联接。这是我遇到的错误的一个示例:
19/04/26 19:39:07 INFO executor.CoarseGrainedExecutorBackend: Got assigned task 1315
19/04/26 19:39:07 INFO executor.Executor: Running task 25.1 in stage 13.0 (TID 1315)
19/04/26 19:39:07 INFO output.FileOutputCommitter: File Output Committer Algorithm version is 1
19/04/26 19:39:07 INFO datasources.SQLHadoopMapReduceCommitProtocol: Using output committer class org.apache.hadoop.mapreduce.lib.output.FileOutputCommitter
19/04/26 19:39:07 INFO datasources.FileScanRDD: Reading File path: SOMEFILEPATH, range: 3087007744-3221225472, partition values: [empty row]
19/04/26 19:39:17 INFO datasources.FileScanRDD: Reading File path: SOMEFILEPATH, range: 15971909632-16106127360, partition values: [empty row]
19/04/26 19:39:24 WARN hdfs.DFSClient: DFSOutputStream ResponseProcessor exception for block isi_hdfs_pool:blk_4549851005_134218728
java.io.IOException: Connection reset by peer
at sun.nio.ch.FileDispatcherImpl.read0(Native Method)
at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:39)
at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223)
at sun.nio.ch.IOUtil.read(IOUtil.java:197)
at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:380)
at org.apache.hadoop.net.SocketInputStream$Reader.performIO(SocketInputStream.java:57)
at org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:142)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:161)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:131)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:118)
at java.io.FilterInputStream.read(FilterInputStream.java:83)
at java.io.FilterInputStream.read(FilterInputStream.java:83)
at org.apache.hadoop.hdfs.protocolPB.PBHelper.vintPrefixed(PBHelper.java:2280)
at org.apache.hadoop.hdfs.protocol.datatransfer.PipelineAck.readFields(PipelineAck.java:244)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:733)
19/04/26 19:39:27 ERROR util.Utils: Aborting task
com.univocity.parsers.common.TextWritingException: Error writing row.
Internal state when error was thrown: recordCount=458089, recordData=["SOMEDATA"]
at com.univocity.parsers.common.AbstractWriter.throwExceptionAndClose(AbstractWriter.java:916)
at com.univocity.parsers.common.AbstractWriter.writeRow(AbstractWriter.java:706)
at org.apache.spark.sql.execution.datasources.csv.UnivocityGenerator.write(UnivocityGenerator.scala:82)
at org.apache.spark.sql.execution.datasources.csv.CsvOutputWriter.write(CSVFileFormat.scala:139)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.execute(FileFormatWriter.scala:327)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:258)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:256)
at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1375)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:261)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$apply$mcV$sp$1.apply(FileFormatWriter.scala:191)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$apply$mcV$sp$1.apply(FileFormatWriter.scala:190)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:108)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:335)
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)
Caused by: java.lang.IllegalStateException: Error closing the output.
at com.univocity.parsers.common.AbstractWriter.close(AbstractWriter.java:861)
at com.univocity.parsers.common.AbstractWriter.throwExceptionAndClose(AbstractWriter.java:903)
at com.univocity.parsers.common.AbstractWriter.writeRow(AbstractWriter.java:811)
at com.univocity.parsers.common.AbstractWriter.writeRow(AbstractWriter.java:704)
... 15 more
Caused by: java.io.IOException: All datanodes DatanodeInfoWithStorage[10.241.209.34:585,null,DISK] are bad. Aborting...
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.setupPipelineForAppendOrRecovery(DFSOutputStream.java:1109)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.processDatanodeError(DFSOutputStream.java:871)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.run(DFSOutputStream.java:401)
19/04/26 19:39:27 WARN util.Utils: Suppressing exception in catch: Failed on local exception: java.io.IOException: Connection reset by peer; Host Details : local host is: "SOMENODEHOST"; destination host is: "SOMEDESTINATIONHOST":SOMEPORT;
java.io.IOException: Failed on local exception: java.io.IOException: Connection reset by peer; Host Details : local host is: "SOMENODEHOST"; destination host is: "SOMEDESTINATIONHOST":SOMEPORT;
at org.apache.hadoop.net.NetUtils.wrapException(NetUtils.java:776)
at org.apache.hadoop.ipc.Client.call(Client.java:1479)
at org.apache.hadoop.ipc.Client.call(Client.java:1412)
at org.apache.hadoop.ipc.ProtobufRpcEngine$Invoker.invoke(ProtobufRpcEngine.java:229)
at com.sun.proxy.$Proxy17.delete(Unknown Source)
at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB.delete(ClientNamenodeProtocolTranslatorPB.java:540)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at org.apache.hadoop.io.retry.RetryInvocationHandler.invokeMethod(RetryInvocationHandler.java:191)
at org.apache.hadoop.io.retry.RetryInvocationHandler.invoke(RetryInvocationHandler.java:102)
at com.sun.proxy.$Proxy18.delete(Unknown Source)
at org.apache.hadoop.hdfs.DFSClient.delete(DFSClient.java:2044)
at org.apache.hadoop.hdfs.DistributedFileSystem$14.doCall(DistributedFileSystem.java:707)
at org.apache.hadoop.hdfs.DistributedFileSystem$14.doCall(DistributedFileSystem.java:703)
at org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
at org.apache.hadoop.hdfs.DistributedFileSystem.delete(DistributedFileSystem.java:714)
at org.apache.hadoop.mapreduce.lib.output.FileOutputCommitter.abortTask(FileOutputCommitter.java:568)
at org.apache.hadoop.mapreduce.lib.output.FileOutputCommitter.abortTask(FileOutputCommitter.java:557)
at org.apache.spark.internal.io.HadoopMapReduceCommitProtocol.abortTask(HadoopMapReduceCommitProtocol.scala:159)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$1.apply$mcV$sp(FileFormatWriter.scala:266)
at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1384)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:261)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$apply$mcV$sp$1.apply(FileFormatWriter.scala:191)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$apply$mcV$sp$1.apply(FileFormatWriter.scala:190)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:108)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:335)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: java.io.IOException: Connection reset by peer
at sun.nio.ch.FileDispatcherImpl.read0(Native Method)
at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:39)
at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223)
at sun.nio.ch.IOUtil.read(IOUtil.java:197)
at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:380)
at org.apache.hadoop.net.SocketInputStream$Reader.performIO(SocketInputStream.java:57)
at org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:142)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:161)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:131)
at java.io.FilterInputStream.read(FilterInputStream.java:133)
at java.io.FilterInputStream.read(FilterInputStream.java:133)
at org.apache.hadoop.ipc.Client$Connection$PingInputStream.read(Client.java:520)
at java.io.BufferedInputStream.fill(BufferedInputStream.java:246)
at java.io.BufferedInputStream.read(BufferedInputStream.java:265)
at java.io.DataInputStream.readInt(DataInputStream.java:387)
at org.apache.hadoop.ipc.Client$Connection.receiveRpcResponse(Client.java:1084)
at org.apache.hadoop.ipc.Client$Connection.run(Client.java:979)
在将大型数据集与较小的数据集连接之前,我尝试将第一个记录的10000条记录与整个较小的数据集(535 MB)连接起来。我遇到了“期货超时[300 s]错误”。
然后我将spark.sql.broadcastTimeout变量增加到3600 s。工作正常。但是,当我尝试将其与整个数据集(654 GB)合并时,它给了我您可以在此处看到的错误( TextWriting Exception )。
我的问题是:
如何才能更有效地监视我的火花作业?而我应该如何进行?
您认为导致此错误发生的原因是什么?我该如何解决?
您将在下面找到有关群集,Spark作业的执行和配置的一些信息。
我正在生产环境中工作(请参阅下面的集群配置)。我无法升级我的Spark版本。我没有Spark UI或yarn UI来监视我的工作。我所能找到的只是纱线记录。
def readCsv(spark: SparkSession, path: String): DataFrame = {
spark.read
.option("header", true)
.option("escape", "\"")
.option("mode", "FAILFAST")
.csv(path)
}
val uh_months = readCsv(spark, input_dir_terro + "HDFS_PATH_OF_ALL_THE_CSV_FILES")
.withColumnRenamed("NUM", "NO_NUM")
.where(col("BEWC").isin(
LIST OF VALUES))
.withColumn("january", lit("1960-01-01"))
val uh = uh_months
.withColumn("UHDIN", datediff(to_date(unix_timestamp(col("UHDIN_YYYYMMDD"), "yyyyMMdd").cast(TimestampType)),
to_date(unix_timestamp(col("january"), "yyyy-MM-dd").cast(TimestampType))))
// .withColumn("DVA_1", to_date((unix_timestamp(col("DVA"), "ddMMMyyyy")).cast(TimestampType)))
.withColumn("DVA_1", date_format(col("DVA"), "dd/MM/yyyy"))
.drop("UHDIN_YYYYMMDD")
.drop("january")
.drop("DVA")
val uh_joined = uh.join(broadcast(smallDF), "KEY")
.select(
uh.col("*"),
smallDF.col("PSP"),
smallDF.col("minrel"),
smallDF.col("Label"),
smallDF.col("StartDate"))
.where(smallDF.col("PSP").isNotNull)
.withColumnRenamed("DVA_1", "DVA")
.where(col("BKA").isNotNull)
smallDF是经过一些聚合和转换后获得的535 MB数据帧。
== Physical Plan ==
*Project [NO_NUM#252, DEV#153, DEBCRED#154, BDGRORI#155, BDGREUR#156, BEWC#157, MSG30_NL#158, SCAPMV#159, USERID#160, MMED#161, TNUM#162, NMTGP#163, BKA#164, CATEXT#165, SEQETAT#166, ACCTYPE#167, BRAND#168, FAMILY#169, SUBFAMILY#170, FORCED_DVA#172, BYBANK#173, CPTE_PROTEGE#174, HOURMV#175, RDFB#176, ... 30 more fields]
+- *BroadcastHashJoin [NO_NUM#252], [NO_NUM#13], Inner, BuildRight
:- *Project [NUM#152 AS NO_NUM#252, DEV#153, DEBCRED#154, BDGRORI#155, BDGREUR#156, BEWC#157, MSG30_NL#158, SCAPMV#159, USERID#160, MMED#161, TNUM#162, NMTGP#163, BKA#164, CATEXT#165, SEQETAT#166, ACCTYPE#167, BRAND#168, FAMILY#169, SUBFAMILY#170, FORCED_DVA#172, BYBANK#173, CPTE_PROTEGE#174, HOURMV#175, RDFB#176, ... 26 more fields]
: +- *Filter ((BEWC#157 INSET (25003,25302,25114,20113,12017,20108,25046,12018,15379,15358,11011,20114,10118,12003,25097,20106,20133,10133,10142,15402,25026,25345,28023,15376,25019,28004,21701,25001,11008,15310,15003,28020,22048,15470,25300,25514,25381,25339,15099,25301,28005,28026,25098,25018,15323,25376,15804,15414,25344,25102,15458,15313,28002,25385,22051,25214,15031,12005,15425,20145,22011,15304,25027,14020,11007,25901,15343,22049,20112,12031,20127,15339,25421,15432,28025,25340,25325,20150,28011,25368,25304,22501,25369,28022,15098,12032,15375,25002,25008,10116,10101,22502,25090,15004,20105,12030,22503,15095,22007,15809,15342,15311,25216,10103,20122,11019,20142,15097,20147,20149,25005,25205,25380,15380,10120,25015,15384,11003,10110,25016,15090,25307,15001,25390,15312,10115,25219,15806,15459,12016,15359,15395,15302,12021,11701,10111,10148,25379,15807,10102,25352,25355,12010,25095,25394,20101,25413,15385,25322,28027,11026,15533,25201,25371,10128,11028,12020,15819,10143,28028,10123,10125,11020,25029,10122,25343,15015,12033,25014,12012,25024,25375,11023,25501,25402,22001,15317,12014,16114,20501,15046,12001,12022,10104,10117,12002,25499,10145,10153,12011,15350,15300,10119,25305,15345,25374,11027,25430,28021,25202,10121,28024,25101,28001,15321,11025,25358,15333,15501,25533,15372,12008,11015,10114,10113,10112,15303,15320,28006,22002,25359,10132,15497,25353,11029,25425,15374,12019,25437,11022,15357,20148,20111,26114,25099,25354,10124,25303,11010,20120,20135,15820,15331,28029) && isnotnull(BKA#164)) && isnotnull(NUM#152))
: +- *FileScan csv [UHDIN_YYYYMMDD#151,NUM#152,DEV#153,DEBCRED#154,BDGRORI#155,BDGREUR#156,BEWC#157,MSG30_NL#158,SCAPMV#159,USERID#160,MMED#161,TNUM#162,NMTGP#163,BKA#164,CATEXT#165,SEQETAT#166,ACCTYPE#167,BRAND#168,FAMILY#169,SUBFAMILY#170,DVA#171,FORCED_DVA#172,BYBANK#173,CPTE_PROTEGE#174,... 26 more fields] Batched: false, Format: CSV, Location: InMemoryFileIndex[hdfs://SOMEHOST:SOMEPORT/SOMEPATH..., PartitionFilters: [], PushedFilters: [In(BEWC, [25003,25302,25114,20113,12017,20108,25046,12018,15379,15358,11011,20114,10118,12003,25..., ReadSchema: struct<UHDIN_YYYYMMDD:string,NUM:string,DEV:string,DEBCRED:string,BDGRORI:string,BDGREUR:string,B...
+- BroadcastExchange HashedRelationBroadcastMode(List(input[0, string, true]))
+- *Project [NO_NUM#13, minrel#370, PSP#82, Label#105, StartDate#106]
+- *SortMergeJoin [PSP#381], [PSP#82], Inner
:- *Sort [PSP#381 ASC NULLS FIRST], false, 0
: +- Exchange hashpartitioning(PSP#381, 200)
: +- *Project [PSP#381, NO_NUM#13, minrel#370]
: +- SortMergeJoin [PSP#381, C_SNUM#14, minrel#370, NO_NUM#13], [NO_PSP#47, C_SNUM_1#387, C_NRELPR#50, NO_NUM_1#400], LeftOuter
: :- *Sort [PSP#381 ASC NULLS FIRST, C_SNUM#14 ASC NULLS FIRST, minrel#370 ASC NULLS FIRST, NO_NUM#13 ASC NULLS FIRST], false, 0
: : +- Exchange hashpartitioning(PSP#381, C_SNUM#14, minrel#370, NO_NUM#13, 200)
: : +- SortAggregate(key=[NO_PSP#12, C_SNUM#14, NO_NUM#13], functions=[min(C_NRELPR#15)])
: : +- *Sort [NO_PSP#12 ASC NULLS FIRST, C_SNUM#14 ASC NULLS FIRST, NO_NUM#13 ASC NULLS FIRST], false, 0
: : +- Exchange hashpartitioning(NO_PSP#12, C_SNUM#14, NO_NUM#13, 200)
: : +- SortAggregate(key=[NO_PSP#12, C_SNUM#14, NO_NUM#13], functions=[partial_min(C_NRELPR#15)])
: : +- *Sort [NO_PSP#12 ASC NULLS FIRST, C_SNUM#14 ASC NULLS FIRST, NO_NUM#13 ASC NULLS FIRST], false, 0
: : +- *Project [NO_PSP#12, C_SNUM#14, NO_NUM#13, C_NRELPR#15]
: : +- *Filter (((C_NRELPR#15 IN (001,006) && C_SNUM#14 IN (030,033)) && isnotnull(NO_PSP#12)) && isnotnull(NO_NUM#13))
: : +- *FileScan csv [NO_PSP#12,NO_NUM#13,C_SNUM#14,c_nrelpr#15] Batched: false, Format: CSV, Location: InMemoryFileIndex[hdfs://SOMEHOST:SOMEPORT/SOMEPATH..., PartitionFilters: [], PushedFilters: [In(c_nrelpr, [001,006]), In(C_SNUM, [030,033]), IsNotNull(NO_PSP), IsNotNull(NO_NUM)], ReadSchema: struct<NO_PSP:string,NO_NUM:string,C_SNUM:string,c_nrelpr:string>
: +- *Sort [NO_PSP#47 ASC NULLS FIRST, C_SNUM_1#387 ASC NULLS FIRST, C_NRELPR#50 ASC NULLS FIRST, NO_NUM_1#400 ASC NULLS FIRST], false, 0
: +- Exchange hashpartitioning(NO_PSP#47, C_SNUM_1#387, C_NRELPR#50, NO_NUM_1#400, 200)
: +- *Project [NO_PSP#47, NO_NUM#48 AS NO_NUM_1#400, C_SNUM#49 AS C_SNUM_1#387, c_nrelpr#50]
: +- *FileScan csv [NO_PSP#47,NO_NUM#48,C_SNUM#49,c_nrelpr#50] Batched: false, Format: CSV, Location: InMemoryFileIndex[hdfs://SOMEHOST:SOMEPORT/SOMEPATH..., PartitionFilters: [], PushedFilters: [], ReadSchema: struct<NO_PSP:string,NO_NUM:string,C_SNUM:string,c_nrelpr:string>
+- *Sort [PSP#82 ASC NULLS FIRST], false, 0
+- Exchange hashpartitioning(PSP#82, 200)
+- *Project [PSP#82, Label#105, StartDate#106]
+- *Filter isnotnull(PSP#82)
+- *FileScan csv [PSP#82,Label#105,StartDate#106] Batched: false, Format: CSV, Location: InMemoryFileIndex[hdfs://SOMEHOST:SOMEPORT/SOMEPATH..., PartitionFilters: [], PushedFilters: [IsNotNull(PSP)], ReadSchema: struct<PSP:string,Label:string,StartDate:string>
-主:纱
-执行者内存:42G
-executor-cores:5
-驱动程序内存:42G
-num-executors:28
-spark.sql.broadcastTimeout = 3600
-spark.kryoserializer.buffer.max = 512
-spark.yarn.executor.memoryOverhead = 2400
-spark.driver.maxResultSize = 500m
-spark.memory.storageFraction = 0.9
-spark.memory.fraction = 0.9
-spark.hadoop.fs.permissions.umask-mode = 007
我们使用IntelliJ构建工件(jar),然后将其发送到服务器。然后执行bash脚本。该脚本:
导出一些环境变量(SPARK_HOME,HADOOP_CONF_DIR,PATH和SPARK_LOCAL_DIRS)
使用上面的spark配置中定义的所有参数启动spark-submit命令
检索该应用程序的纱线日志
答案 0 :(得分:0)
以下是有关您代码的一些改进:
repartition
连接的KEY列添加uh
,分区数应约为650GB / 500MB ~ 1300
。cache
小型数据集spark.broadcast.blockSize
的值,可能会增加它。以下是更改后的代码外观:
val uh_months = readCsv(spark, input_dir_terro + "HDFS_PATH_OF_ALL_THE_CSV_FILES")
.withColumnRenamed("OLD_KEY", "KEY")
.where(col("code").isin(LIST OF VALUES))
.withColumn("january", lit("1960-01-01"))
val uh = uh_months
.withColumn("UHDIN", datediff(to_date(unix_timestamp(col("UHDIN_YYYYMMDD"), "yyyyMMdd").cast(TimestampType)),
to_date(unix_timestamp(col("january"), "yyyy-MM-dd").cast(TimestampType))))
// .withColumn("field_1", to_date((unix_timestamp(col("field"), "ddMMMyyyy")).cast(TimestampType)))
.withColumn("field_1", date_format(col("field"), "dd/MM/yyyy"))
.drop("UHDIN_YYYYMMDD")
.drop("january")
.drop("field")
.repartition(1300, $"KEY") //change 1: repartition based on KEY with 1300 (650GB/500MB~1300)
//change 2: always prune as much information as possible before joining!
val smallerDF = smallDF
.where(smallDF.col("ID").isNotNull && col("field_6").isNotNull)
.select("KEY", "ID", "field_3", "field_4", "field_5")
//change 3: you can optionally cache the small dataset
smallerDF.cache()
//change 4: adjust spark.broadcast.blockSize i.e spark.conf.set("spark.broadcast.blockSize","16m"
val uh_joined = uh.join(broadcast(smallerDF), "KEY")
.select(
uh.col("*"),
smallerDF.col("ID"),
smallerDF.col("field_3"),
smallerDF.col("field_4"),
smallerDF.col("field_5"))
.withColumnRenamed("field_1", "field")
最后一句话与您的集群配置有关,我会尝试将num-executors
至少增加32,因为在如此大的集群中并行化级别应该更高。