scala - 如何在外部 spark shuffle 服务上修复 "Error opening block StreamChunkId"

Question

我正在尝试从 kubernetes 集群中的 zeppelin 部署中运行 spark 作业。我还有一个运行在不同命名空间上的 spark shuffle 服务(daemonset - v2.2.0-k8s)。这是我的 spark 配置(在 zeppelin pod 上设置)

--conf spark.kubernetes.executor.docker.image=<spark-executor> 
--conf spark.executor.cores=5
--conf spark.driver.memory=5g
--conf spark.executor.memory=5g
--conf spark.kubernetes.authenticate.driver.serviceAccountName=<svc-account> 
--conf spark.local.dir=/tmp/spark-local 
--conf spark.executor.instances=5 
--conf spark.dynamicAllocation.enabled=true 
--conf spark.shuffle.service.enabled=true 
--conf spark.kubernetes.shuffle.labels="app=spark-shuffle,spark-version=2.2.0" 
--conf spark.dynamicAllocation.maxExecutors=5   
--conf spark.dynamicAllocation.minExecutors=1 
--conf spark.kubernetes.shuffle.namespace=<namespace> 
--conf spark.kubernetes.docker.image.pullPolicy=IfNotPresent 
--conf spark.kubernetes.initcontainer.docker.image=kubespark/spark-init:v2.2.0-kubernetes-0.5.0 
--conf spark.kubernetes.resourceStagingServer.uri=<ip:port>

但是我从 zeppelin 产生的外部 spark-shuffle 和 spark 执行器中得到以下日志:

+ /sbin/tini -s -- /opt/spark/bin/spark-class org.apache.spark.deploy.k8s.KubernetesExternalShuffleService 1
SLF4J: Class path contains multiple SLF4J bindings.
SLF4J: Found binding in [jar:file:/opt/spark/jars/slf4j-log4j12-1.7.16.jar!/org/sl
f4j/impl/StaticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/opt/spark/jars/kubernetes-client-3.0.1.jar!/org
/slf4j/impl/StaticLoggerBinder.class]
SLF4J: See http://www.slf4j.org/codes.html#multiple_bindings for an explanation.
SLF4J: Actual binding is of type [org.slf4j.impl.Log4jLoggerFactory]
2020-01-14 03:37:31 INFO  ExternalShuffleService:2574 - Started daemon with proces
s name: 10@unawa2-shuffle-unawa2-spark-shuffle-d5cfg
2020-01-14 03:37:31 INFO  SignalUtils:54 - Registered signal handler for TERM
2020-01-14 03:37:31 INFO  SignalUtils:54 - Registered signal handler for HUP
2020-01-14 03:37:31 INFO  SignalUtils:54 - Registered signal handler for INT
2020-01-14 03:37:31 WARN  NativeCodeLoader:62 - Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
2020-01-14 03:37:31 INFO  SecurityManager:54 - Changing view acls to: root
2020-01-14 03:37:31 INFO  SecurityManager:54 - Changing modify acls to: root
2020-01-14 03:37:31 INFO  SecurityManager:54 - Changing view acls groups to:
2020-01-14 03:37:31 INFO  SecurityManager:54 - Changing modify acls groups to:
2020-01-14 03:37:31 INFO  SecurityManager:54 - SecurityManager: authentication disabled; ui acls disabled; users  with view permissions: Set(root); groups with view permissions: Set(); users  with modify permissions: Set(root); groups with modify permissions: Set()
2020-01-14 03:37:32 INFO  KubernetesExternalShuffleService:54 - Starting shuffle service on port 7337 (auth enabled = false)
2020-01-14 03:38:35 INFO  KubernetesShuffleBlockHandler:54 - Received registration request from app spark-application-1578973110574 (remote address /192.168.2.37:40318).
2020-01-14 03:38:36 INFO  ExternalShuffleBlockResolver:135 - Registered executor AppExecId{appId=spark-application-1578973110574, execId=5} with ExecutorShuffleInfo{localDirs=[/tmp/spark-local/blockmgr-8a26a714-3ecb-46dd-8499-ff796fa97744], subDirsPerLocalDir=64, shuffleManager=org.apache.spark.shuffle.sort.SortShuffleManager}
2020-01-14 03:39:15 ERROR TransportRequestHandler:127 - Error opening block StreamChunkId{streamId=527834012000, chunkIndex=0} for request from /192.168.3.130:50896
java.lang.RuntimeException: Failed to open file: /tmp/spark-local/blockmgr-8a26a714-3ecb-46dd-8499-ff796fa97744/0f/shuffle_1_0_0.index
        at org.apache.spark.network.shuffle.ExternalShuffleBlockResolver.getSortBasedShuffleBlockData(ExternalShuffleBlockResolver.java:249)
        at org.apache.spark.network.shuffle.ExternalShuffleBlockResolver.getBlockData(ExternalShuffleBlockResolver.java:174)
        at org.apache.spark.network.shuffle.ExternalShuffleBlockHandler$1.next(ExternalShuffleBlockHandler.java:105)
        at org.apache.spark.network.shuffle.ExternalShuffleBlockHandler$1.next(ExternalShuffleBlockHandler.java:95)
        at org.apache.spark.network.server.OneForOneStreamManager.getChunk(OneForOneStreamManager.java:89)
        at org.apache.spark.network.server.TransportRequestHandler.processFetchRequest(TransportRequestHandler.java:125)
        at org.apache.spark.network.server.TransportRequestHandler.handle(TransportRequestHandler.java:103)
        at org.apache.spark.network.server.TransportChannelHandler.channelRead(TransportChannelHandler.java:118)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:357)
        at io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:343)
        at io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:336)
        at io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:287)
.
.
.
Caused by: java.util.concurrent.ExecutionException: java.io.FileNotFoundException: /tmp/spark-local/blockmgr-8a26a714-3ecb-46dd-8499-ff796fa97744/0f/shuffle_1_0_0.index (No such file or directory)
        at org.spark_project.guava.util.concurrent.AbstractFuture$Sync.getValue(AbstractFuture.java:306)
        at org.spark_project.guava.util.concurrent.AbstractFuture$Sync.get(AbstractFuture.java:293)
        at org.spark_project.guava.util.concurrent.AbstractFuture.get(AbstractFuture.java:116)
        at org.spark_project.guava.util.concurrent.Uninterruptibles.getUninterruptibly(Uninterruptibles.java:135)

知道如何解决这个问题吗？

[编辑]

我将本地目录/tmp/spark-local 安装到我的 pod 中。当我 ssh进入每个节点，我确认 block 管理器存在于一个工作节点中(我猜这是预期的行为)。当来自另一个工作节点的其中一个 shuffle pod 尝试访问同一个 block 管理器时，会发生该错误。

Answer 1

评论线程中的摘要。

为了在启用动态分配的 Kubernetes 上运行 Spark，您可以:

关注 apache-spark-on-k8s guide

重要笔记:

你的图像应该基于 kubespark使用 fork 的 Apache Spark 2.2.0 构建的图像

该功能是实验性的，不支持

等到 Spark 3.0.0 与 SPARK-24432 一起发布和 SPARK-25299