Я сталкиваюсь с проблемой, когда сортирует огромный набор данных ( 1.2 T ), основанный на 4 столбцах.Мне также нужно сразу же после сортировки разделить этот набор данных при записи окончательного набора данных в HDFS на основе одного из столбцов, используемых в функции сортировки.
Вот публикация в стеке, которую я опубликовал несколько дней назад и описывающаядругая проблема, с которой у меня был тот же код, но в отношении объединения двух наборов данных:
предыдущий выпуск
Я использовал ответ этого поста, чтобы улучшить свой код.Теперь соединение работает нормально.
Я протестировал код без сортировки, и он работает нормально.Чтобы выполнить сортировку, я подумал о разбиении данных на четыре столбца.
Размер одного раздела - 500 МБ .Затем у меня есть 2600=1.2T/500MB
разделов.
При выполнении задания на искру я получаю ошибку shuffle.RetryingBlockFetcher
( см. Журналы ошибок ниже ).
My вопросы :
- Каков наилучший способ сортировки данных в искре, чтобы избежать перемешивания?Или уменьшить его?
- Могу ли я исправить / добавить улучшения в мой код для выполнения сортировки?
- Действительно ли мне нужно сортировать таким образом?Разве я не могу использовать другие методы, такие как Group By?
Фрагмент моего кода:
EDITED
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").repartition(1300,col("MMED"),col("DEBCRED"),col("NMTGP"))//.repartition(1300,col("NO_NUM"))
val uh_flag_comment = new TransactionType().transform(uh)
val uh_repartitioned = uh_flag_comment.repartition(1300,col("NO_NUM"))
val uh_joined = uh_repartitioned.join(broadcast(smallDF), "NO_NUM")
.select(
uh.col("*"),
smallDF.col("PSP"),
smallDF.col("minrel"),
smallDF.col("Label"),
smallDF.col("StartDate"))
.withColumnRenamed("DVA_1", "DVA")
val uh_final = uh_joined.repartition(1300, col("PSP")).sortWithinPartitions(col("NO_NUM"), col("UHDIN"), col("HOURMV"))
return uh_final
TransactionType
- это класс, в котором ядобавьте новый столбец в мой uh
массив данных на основе значения 3 столбцов (MMED
, DEBCRED
, NMTGP
), используя регулярное выражение.
без сортировки и используя полную емкостькластер, код выполняется примерно в течение 1 часа.
План выполнения
== Physical Plan ==
Exchange hashpartitioning(PSP#82, 2600)
+- *Sort [PSP#82 ASC NULLS FIRST, NO_NUM#252 ASC NULLS FIRST, UHDIN#547 ASC NULLS FIRST, HOURMV#175 ASC NULLS FIRST], true, 0
+- Exchange rangepartitioning(PSP#82 ASC NULLS FIRST, NO_NUM#252 ASC NULLS FIRST, UHDIN#547 ASC NULLS FIRST, HOURMV#175 ASC NULLS FIRST, 200)
+- Exchange hashpartitioning(PSP#82, NO_NUM#252, UHDIN#547, HOURMV#175, 2600)
+- *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
:- Exchange hashpartitioning(NO_NUM#252, 1300)
: +- *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 (25003,25302,25114,20113,12017,20108,25046,12018,15379,15358,11011,20114,10118,12003,25097,20106,20133,10133,10142,15402,25026,25345,28023,15376,25019,28004,21701,25001,11008,15310,15003,2SOMEPORT,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(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[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, PSP#82, minrel#370, 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_NUM#13)) && isnotnull(NO_PSP#12))
: : +- *FileScan csv [NO_PSP#12,NO_NUM#13,C_SNUM#14,c_nrelpr#15] Batched: false, Format: CSV, Location: InMemoryFileIndex[SOMEHOST:SOMEPORT/SOMEPATH, PartitionFilters: [], PushedFilters: [In(c_nrelpr, [001,006]), In(C_SNUM, [030,033]), IsNotNull(NO_NUM), IsNotNull(NO_PSP)], 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[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[SOMEHOST:SOMEPORT/SOMEPATH, PartitionFilters: [], PushedFilters: [IsNotNull(PSP)], ReadSchema: struct<PSP:string,Label:string,StartDate:string>
Вот основные ошибки, которые я получил при запуске задания с сортировкой:
19/05/06 18:02:25 ERROR shuffle.RetryingBlockFetcher: Exception while beginning fetch of 214 outstanding blocks
java.io.IOException: Failed to connect to SOMEHOST/SOMEADDRESS:SOMEPORT
at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:232)
at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:182)
at org.apache.spark.network.netty.NettyBlockTransferService$$anon$1.createAndStart(NettyBlockTransferService.scala:98)
at org.apache.spark.network.shuffle.RetryingBlockFetcher.fetchAllOutstanding(RetryingBlockFetcher.java:141)
at org.apache.spark.network.shuffle.RetryingBlockFetcher.start(RetryingBlockFetcher.java:121)
at org.apache.spark.network.netty.NettyBlockTransferService.fetchBlocks(NettyBlockTransferService.scala:108)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.sendRequest(ShuffleBlockFetcherIterator.scala:228)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.fetchUpToMaxBytes(ShuffleBlockFetcherIterator.scala:435)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.initialize(ShuffleBlockFetcherIterator.scala:323)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.<init>(ShuffleBlockFetcherIterator.scala:140)
at org.apache.spark.shuffle.BlockStoreShuffleReader.read(BlockStoreShuffleReader.scala:45)
at org.apache.spark.sql.execution.ShuffledRowRDD.compute(ShuffledRowRDD.scala:165)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53)
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: io.netty.channel.AbstractChannel$AnnotatedConnectException: Connection refused: SOMEHOST/SOMEADDRESS:SOMEPORT
at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:717)
at io.netty.channel.socket.nio.NioSocketChannel.doFinishConnect(NioSocketChannel.java:257)
at io.netty.channel.nio.AbstractNioChannel$AbstractNioUnsafe.finishConnect(AbstractNioChannel.java:291)
at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:631)
at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:566)
at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:480)
at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:442)
at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
... 1 more
19/05/06 18:02:25 INFO shuffle.RetryingBlockFetcher: Retrying fetch (1/3) for 214 outstanding blocks after 5000 ms
19/05/06 18:02:25 INFO storage.ShuffleBlockFetcherIterator: Started 6 remote fetches in 13 ms
19/05/06 18:02:28 INFO executor.Executor: Finished task 408.0 in stage 14.0 (TID 6696). 1733 bytes result sent to driver
19/05/06 18:02:28 INFO executor.CoarseGrainedExecutorBackend: Got assigned task 6816
19/05/06 18:02:28 INFO executor.Executor: Running task 466.1 in stage 14.0 (TID 6816)
19/05/06 18:02:28 INFO storage.ShuffleBlockFetcherIterator: Getting 5073 non-empty blocks out of 5089 blocks
19/05/06 18:02:28 INFO client.TransportClientFactory: Found inactive connection to SOMEHOST/SOMEADDRESS:SOMEPORT, creating a new one.
19/05/06 18:02:28 ERROR shuffle.RetryingBlockFetcher: Exception while beginning fetch of 82 outstanding blocks
java.io.IOException: Failed to connect to SOMEHOST/SOMEADDRESS:SOMEPORT
at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:232)
at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:182)
at org.apache.spark.network.netty.NettyBlockTransferService$$anon$1.createAndStart(NettyBlockTransferService.scala:98)
at org.apache.spark.network.shuffle.RetryingBlockFetcher.fetchAllOutstanding(RetryingBlockFetcher.java:141)
at org.apache.spark.network.shuffle.RetryingBlockFetcher.start(RetryingBlockFetcher.java:121)
at org.apache.spark.network.netty.NettyBlockTransferService.fetchBlocks(NettyBlockTransferService.scala:108)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.sendRequest(ShuffleBlockFetcherIterator.scala:228)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.fetchUpToMaxBytes(ShuffleBlockFetcherIterator.scala:435)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.initialize(ShuffleBlockFetcherIterator.scala:323)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.<init>(ShuffleBlockFetcherIterator.scala:140)
at org.apache.spark.shuffle.BlockStoreShuffleReader.read(BlockStoreShuffleReader.scala:45)
at org.apache.spark.sql.execution.ShuffledRowRDD.compute(ShuffledRowRDD.scala:165)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53)
at org.apache.spark.scheduler.Task.run(Task.scala:108)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:335)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: io.netty.channel.AbstractChannel$AnnotatedConnectException: Connection refused: SOMEHOST/SOMEADDRESS:SOMEPORT
at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:717)
at io.netty.channel.socket.nio.NioSocketChannel.doFinishConnect(NioSocketChannel.java:257)
at io.netty.channel.nio.AbstractNioChannel$AbstractNioUnsafe.finishConnect(AbstractNioChannel.java:291)
at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:631)
at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:566)
at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:480)
at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:442)
at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
... 1 more
EDITED
Другой тип ошибки:
19/05/06 18:06:16 ERROR executor.Executor: Exception in task 309.1 in stage 13.1 (TID 7592)
java.io.FileNotFoundException: /applis/hadoop/yarn/local/usercache/MYUSER/appcache/application_1555263602441_0123/blockmgr-aa586b76-ff58-4f88-b168-288c3e1b9f61/3c/temp_shuffle_ea967624-f633-4481-9a05-249b561e3c38 (No such file or directory)
at java.io.FileInputStream.open0(Native Method)
at java.io.FileInputStream.open(FileInputStream.java:195)
at java.io.FileInputStream.<init>(FileInputStream.java:138)
at org.spark_project.guava.io.Files$FileByteSource.openStream(Files.java:124)
at org.spark_project.guava.io.Files$FileByteSource.openStream(Files.java:114)
at org.spark_project.guava.io.ByteSource.copyTo(ByteSource.java:202)
at org.spark_project.guava.io.Files.copy(Files.java:436)
at org.spark_project.guava.io.Files.move(Files.java:651)
at org.apache.spark.shuffle.sort.UnsafeShuffleWriter.mergeSpills(UnsafeShuffleWriter.java:277)
at org.apache.spark.shuffle.sort.UnsafeShuffleWriter.closeAndWriteOutput(UnsafeShuffleWriter.java:216)
at org.apache.spark.shuffle.sort.UnsafeShuffleWriter.write(UnsafeShuffleWriter.java:169)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53)
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)
19/05/06 18:06:16 ERROR executor.Executor: Exception in task 502.1 in stage 13.1 (TID 7599)
java.io.FileNotFoundException: /applis/hadoop/yarn/local/usercache/MYUSER/appcache/application_1555263602441_0123/blockmgr-aa586b76-ff58-4f88-b168-288c3e1b9f61/34/temp_shuffle_dd202cd1-ad8f-41c4-b4d1-d79621cd169e (No such file or directory)
at java.io.FileOutputStream.open0(Native Method)
at java.io.FileOutputStream.open(FileOutputStream.java:270)
at java.io.FileOutputStream.<init>(FileOutputStream.java:213)
at org.apache.spark.storage.DiskBlockObjectWriter.initialize(DiskBlockObjectWriter.scala:102)
at org.apache.spark.storage.DiskBlockObjectWriter.open(DiskBlockObjectWriter.scala:115)
at org.apache.spark.storage.DiskBlockObjectWriter.write(DiskBlockObjectWriter.scala:247)
at org.apache.spark.shuffle.sort.ShuffleExternalSorter.writeSortedFile(ShuffleExternalSorter.java:201)
at org.apache.spark.shuffle.sort.ShuffleExternalSorter.closeAndGetSpills(ShuffleExternalSorter.java:405)
at org.apache.spark.shuffle.sort.UnsafeShuffleWriter.closeAndWriteOutput(UnsafeShuffleWriter.java:209)
at org.apache.spark.shuffle.sort.UnsafeShuffleWriter.write(UnsafeShuffleWriter.java:169)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53)
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)
19/05/06 18:06:16 INFO executor.Executor: Finished task 200.2 in stage 13.1 (TID 7568). 2826 bytes result sent to driver
19/05/06 18:06:16 ERROR executor.CoarseGrainedExecutorBackend: RECEIVED SIGNAL TERM
19/05/06 18:06:16 INFO util.ShutdownHookManager: Shutdown hook called
Некоторая информация / контекст:
Я работаю в производственной среде (см. Конфигурацию кластераниже).Я не могу обновить свою версию свечи.У меня нет искрового или пряжевого интерфейса для мониторинга моей работы.Все, что я могу получить, это журналы пряжи.
Версия Spark: 2.2
Конфигурация кластера:
- 21 вычислительный узел (рабочие)
- 8каждый ядро
- 64 ГБ ОЗУ на узел
Текущая конфигурация Spark:
-мастер: пряжа
-программа-память исполнителя: 42G
-executor-core: 5
-драйвер памяти: 42G
-num-исполнителей: 32
-spark.sql.broadcastTimeout = 3600
-spark.kryoserializer.buffer.max = 512
-spark.yarn.executor.memoryOverhead = 2400
-spark.driver.maxResultSize = 500 м
-spark.memory.storageFraction = 0,3
-spark.memory.fraction = 0,9
-spark.hadoop.fs.permissions.umask-mode = 007
Как этозадание выполнено:
Мы создаем артефакт (jar) с помощью IntelliJ и затем отправляем его на сервер.Затем выполняется скрипт bash.Этот скрипт:
экспортирует некоторые переменные среды (SPARK_HOME, HADOOP_CONF_DIR, PATH и SPARK_LOCAL_DIRS)
запускает команду spark-submit со всеми параметрамиопределенный в конфигурации искры выше
извлекает журналы пряжи приложения