使用 zstd 壓縮編解碼器時 Spark 3.0.1 任務失敗

Question

我將Spark 3.0.1與用戶提供的Hadoop 3.2.0和Scala 2.12.10在Kubernetes 上運行一起使用。

讀取壓縮為活潑的拼花文件時，但是當我嘗試閱讀下面的錯誤下壓縮為zstd幾個任務拼花文件失敗，一切工作正常：

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，這些任務在重試后會成功，但並非總是如此，因此我的工作經常失敗。 如前所述，如果我使用壓縮為 snappy 的相同數據集，一切正常。

我還嘗試構建 Spark 和 Hadoop，更改 zstd-jni 版本，但仍然發生相同的行為。

有誰知道可能會發生什么？

謝謝！

Answer 1

正如評論的那樣，我使用以下屬性更新了 Spark (3.0.1) 配置以永久解決我的問題。 添加的文件路徑和配置如下：

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