apache-spark - 使用 zstd 压缩编解码器时 Spark 3.0.1 任务失败

Question

我正在使用 Spark 3.0.1 用户提供 Hadoop 3.2.0 和 斯卡拉 2.12.10 运行于 Kubernetes .
读取压缩为 的 Parquet 文件时一切正常活泼 ，但是当我尝试读取压缩为 的 Parquet 文件时zstd 几个任务在以下错误下失败:

java.io.IOException: Decompression error: Version not supported
at com.github.luben.zstd.ZstdInputStream.readInternal(ZstdInputStream.java:164)
at com.github.luben.zstd.ZstdInputStream.read(ZstdInputStream.java:120)
at java.io.BufferedInputStream.fill(BufferedInputStream.java:246)
at java.io.BufferedInputStream.read1(BufferedInputStream.java:286)
at java.io.BufferedInputStream.read(BufferedInputStream.java:345)
at java.io.ObjectInputStream$PeekInputStream.read(ObjectInputStream.java:2781)
at java.io.ObjectInputStream$PeekInputStream.readFully(ObjectInputStream.java:2797)
at java.io.ObjectInputStream$BlockDataInputStream.readShort(ObjectInputStream.java:3274)
at java.io.ObjectInputStream.readStreamHeader(ObjectInputStream.java:934)
at java.io.ObjectInputStream.(ObjectInputStream.java:396)
at org.apache.spark.MapOutputTracker$.deserializeObject$1(MapOutputTracker.scala:954)
at org.apache.spark.MapOutputTracker$.deserializeMapStatuses(MapOutputTracker.scala:964)
at org.apache.spark.MapOutputTrackerWorker.$anonfun$getStatuses$2(MapOutputTracker.scala:856)
at org.apache.spark.util.KeyLock.withLock(KeyLock.scala:64)
at org.apache.spark.MapOutputTrackerWorker.getStatuses(MapOutputTracker.scala:851)
at org.apache.spark.MapOutputTrackerWorker.getMapSizesByExecutorId(MapOutputTracker.scala:808)
at org.apache.spark.shuffle.sort.SortShuffleManager.getReader(SortShuffleManager.scala:128)
at org.apache.spark.sql.execution.ShuffledRowRDD.compute(ShuffledRowRDD.scala:185)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:313)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:313)
at org.apache.spark.rdd.ZippedPartitionsRDD2.compute(ZippedPartitionsRDD.scala:89)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:313)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:313)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:313)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:90)
at org.apache.spark.scheduler.Task.run(Task.scala:127)
at org.apache.spark.executor.Executor$TaskRunner.$anonfun$run$3(Executor.scala:446)
at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1377)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:449)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)

我不明白的是，这些任务在重试后会成功，但并非总是如此，因此我的工作经常失败。如前所述，如果我使用压缩为 snappy 的相同数据集，一切正常。
我还尝试构建 Spark 和 Hadoop，更改 zstd-jni 版本，但仍然发生相同的行为。
有谁知道可能会发生什么？
谢谢!

Answer 1

正如评论的那样，我使用以下属性更新了 Spark (3.0.1) 配置以永久解决我的问题。添加的文件路径和配置如下:

$SPARK_HOME/conf/spark-defaults.conf
spark.shuffle.mapStatus.compression.codec lz4