SparkSQL配置和使用初探

1.环境

  • OS:Red Hat Enterprise Linux Server release 6.4 (Santiago)
  • Hadoop:Hadoop 2.4.1
  • Hive：0.11.0
  • JDK:1.7.0_60
  • Spark:1.1.0（内置SparkSQL）
  • Scala:2.11.2

2.Spark集群规划

  • 账户:ebupt
  • master:eb174
  • slaves:eb174、eb175、eb176

3.SparkSQL发展历史

2014年9月11日，发布Spark1.1.0。Spark从1.0开始引入SparkSQL。Spark1.1.0变化较大是SparkSQL和MLlib。具体参见release note。

SparkSQL的前身是Shark。由于Shark自身的不完善，2014年6月1日Reynold Xin宣布：停止对Shark的开发。SparkSQL抛弃原有Shark的代码，汲取了Shark的一些优点，如内存列存储（In-Memory Columnar Storage）、Hive兼容性等，重新开发SparkSQL。

4.配置

  1. 安装配置同Spark-0.9.1（参见博文:Spark、Shark集群安装部署及遇到的问题解决）
  2. 将$HIVE_HOME/conf/hive-site.xml配置文件拷贝到$SPARK_HOME/conf目录下。
  3. 将$HADOOP_HOME/etc/hadoop/hdfs-site.xml配置文件拷贝到$SPARK_HOME/conf目录下。

5.运行

  1. 启动Spark集群
  2. 启动SparkSQL Client：./spark/bin/spark-sql --master spark://eb174:7077 --executor-memory 3g
  3. 运行SQL，访问hive的表：spark-sql> select count(*) from test.t1;
14/10/08 20:46:04 INFO ParseDriver: Parsing command: select count(*) from test.t1
14/10/08 20:46:05 INFO ParseDriver: Parse Completed
14/10/08 20:46:05 INFO metastore: Trying to connect to metastore with URI thrift://eb170:9083
14/10/08 20:46:05 INFO metastore: Waiting 1 seconds before next connection attempt.
14/10/08 20:46:06 INFO SparkDeploySchedulerBackend: Registered executor: Actor[akka.tcp://sparkExecutor@eb174:55408/user/Executor#1282322316] with ID 2
14/10/08 20:46:06 INFO SparkDeploySchedulerBackend: Registered executor: Actor[akka.tcp://sparkExecutor@eb176:56138/user/Executor#-264112470] with ID 0
14/10/08 20:46:06 INFO SparkDeploySchedulerBackend: Registered executor: Actor[akka.tcp://sparkExecutor@eb175:43791/user/Executor#-996481867] with ID 1
14/10/08 20:46:06 INFO BlockManagerMasterActor: Registering block manager eb174:54967 with 265.4 MB RAM
14/10/08 20:46:06 INFO BlockManagerMasterActor: Registering block manager eb176:60783 with 265.4 MB RAM
14/10/08 20:46:06 INFO BlockManagerMasterActor: Registering block manager eb175:35197 with 265.4 MB RAM
14/10/08 20:46:06 INFO metastore: Connected to metastore.
14/10/08 20:46:07 INFO deprecation: mapred.map.tasks is deprecated. Instead, use mapreduce.job.maps
14/10/08 20:46:07 INFO MemoryStore: ensureFreeSpace(406982) called with curMem=0, maxMem=278302556
14/10/08 20:46:07 INFO MemoryStore: Block broadcast_0 stored as values in memory (estimated size 397.4 KB, free 265.0 MB)
14/10/08 20:46:07 INFO MemoryStore: ensureFreeSpace(25198) called with curMem=406982, maxMem=278302556
14/10/08 20:46:07 INFO MemoryStore: Block broadcast_0_piece0 stored as bytes in memory (estimated size 24.6 KB, free 265.0 MB)
14/10/08 20:46:07 INFO BlockManagerInfo: Added broadcast_0_piece0 in memory on eb174:49971 (size: 24.6 KB, free: 265.4 MB)
14/10/08 20:46:07 INFO BlockManagerMaster: Updated info of block broadcast_0_piece0
14/10/08 20:46:07 INFO SparkContext: Starting job: collect at HiveContext.scala:415
14/10/08 20:46:08 INFO FileInputFormat: Total input paths to process : 1
14/10/08 20:46:08 INFO DAGScheduler: Registering RDD 5 (mapPartitions at Exchange.scala:86)
14/10/08 20:46:08 INFO DAGScheduler: Got job 0 (collect at HiveContext.scala:415) with 1 output partitions (allowLocal=false)
14/10/08 20:46:08 INFO DAGScheduler: Final stage: Stage 0(collect at HiveContext.scala:415)
14/10/08 20:46:08 INFO DAGScheduler: Parents of final stage: List(Stage 1)
14/10/08 20:46:08 INFO DAGScheduler: Missing parents: List(Stage 1)
14/10/08 20:46:08 INFO DAGScheduler: Submitting Stage 1 (MapPartitionsRDD[5] at mapPartitions at Exchange.scala:86), which has no missing parents
14/10/08 20:46:08 INFO MemoryStore: ensureFreeSpace(11000) called with curMem=432180, maxMem=278302556
14/10/08 20:46:08 INFO MemoryStore: Block broadcast_1 stored as values in memory (estimated size 10.7 KB, free 265.0 MB)
14/10/08 20:46:08 INFO MemoryStore: ensureFreeSpace(5567) called with curMem=443180, maxMem=278302556
14/10/08 20:46:08 INFO MemoryStore: Block broadcast_1_piece0 stored as bytes in memory (estimated size 5.4 KB, free 265.0 MB)
14/10/08 20:46:08 INFO BlockManagerInfo: Added broadcast_1_piece0 in memory on eb174:49971 (size: 5.4 KB, free: 265.4 MB)
14/10/08 20:46:08 INFO BlockManagerMaster: Updated info of block broadcast_1_piece0
14/10/08 20:46:08 INFO DAGScheduler: Submitting 2 missing tasks from Stage 1 (MapPartitionsRDD[5] at mapPartitions at Exchange.scala:86)
14/10/08 20:46:08 INFO TaskSchedulerImpl: Adding task set 1.0 with 2 tasks
14/10/08 20:46:08 INFO TaskSetManager: Starting task 0.0 in stage 1.0 (TID 0, eb174, NODE_LOCAL, 1199 bytes)
14/10/08 20:46:08 INFO TaskSetManager: Starting task 1.0 in stage 1.0 (TID 1, eb176, NODE_LOCAL, 1199 bytes)
14/10/08 20:46:08 INFO ConnectionManager: Accepted connection from [eb176/10.1.69.176:49289]
14/10/08 20:46:08 INFO ConnectionManager: Accepted connection from [eb174/10.1.69.174:33401]
14/10/08 20:46:08 INFO SendingConnection: Initiating connection to [eb176/10.1.69.176:60783]
14/10/08 20:46:08 INFO SendingConnection: Initiating connection to [eb174/10.1.69.174:54967]
14/10/08 20:46:08 INFO SendingConnection: Connected to [eb176/10.1.69.176:60783], 1 messages pending
14/10/08 20:46:08 INFO SendingConnection: Connected to [eb174/10.1.69.174:54967], 1 messages pending
14/10/08 20:46:08 INFO BlockManagerInfo: Added broadcast_1_piece0 in memory on eb176:60783 (size: 5.4 KB, free: 265.4 MB)
14/10/08 20:46:08 INFO BlockManagerInfo: Added broadcast_1_piece0 in memory on eb174:54967 (size: 5.4 KB, free: 265.4 MB)
14/10/08 20:46:08 INFO BlockManagerInfo: Added broadcast_0_piece0 in memory on eb174:54967 (size: 24.6 KB, free: 265.4 MB)
14/10/08 20:46:08 INFO BlockManagerInfo: Added broadcast_0_piece0 in memory on eb176:60783 (size: 24.6 KB, free: 265.4 MB)
14/10/08 20:46:10 INFO TaskSetManager: Finished task 1.0 in stage 1.0 (TID 1) in 2657 ms on eb176 (1/2)
14/10/08 20:46:10 INFO TaskSetManager: Finished task 0.0 in stage 1.0 (TID 0) in 2675 ms on eb174 (2/2)
14/10/08 20:46:10 INFO DAGScheduler: Stage 1 (mapPartitions at Exchange.scala:86) finished in 2.680 s
14/10/08 20:46:10 INFO TaskSchedulerImpl: Removed TaskSet 1.0, whose tasks have all completed, from pool 
14/10/08 20:46:10 INFO DAGScheduler: looking for newly runnable stages
14/10/08 20:46:10 INFO DAGScheduler: running: Set()
14/10/08 20:46:10 INFO DAGScheduler: waiting: Set(Stage 0)
14/10/08 20:46:10 INFO DAGScheduler: failed: Set()
14/10/08 20:46:10 INFO DAGScheduler: Missing parents for Stage 0: List()
14/10/08 20:46:10 INFO DAGScheduler: Submitting Stage 0 (MappedRDD[9] at map at HiveContext.scala:360), which is now runnable
14/10/08 20:46:10 INFO MemoryStore: ensureFreeSpace(9752) called with curMem=448747, maxMem=278302556
14/10/08 20:46:10 INFO MemoryStore: Block broadcast_2 stored as values in memory (estimated size 9.5 KB, free 265.0 MB)
14/10/08 20:46:10 INFO MemoryStore: ensureFreeSpace(4941) called with curMem=458499, maxMem=278302556
14/10/08 20:46:10 INFO MemoryStore: Block broadcast_2_piece0 stored as bytes in memory (estimated size 4.8 KB, free 265.0 MB)
14/10/08 20:46:10 INFO BlockManagerInfo: Added broadcast_2_piece0 in memory on eb174:49971 (size: 4.8 KB, free: 265.4 MB)
14/10/08 20:46:10 INFO BlockManagerMaster: Updated info of block broadcast_2_piece0
14/10/08 20:46:11 INFO DAGScheduler: Submitting 1 missing tasks from Stage 0 (MappedRDD[9] at map at HiveContext.scala:360)
14/10/08 20:46:11 INFO TaskSchedulerImpl: Adding task set 0.0 with 1 tasks
14/10/08 20:46:11 INFO TaskSetManager: Starting task 0.0 in stage 0.0 (TID 2, eb175, PROCESS_LOCAL, 948 bytes)
14/10/08 20:46:11 INFO StatsReportListener: Finished stage: org.apache.spark.scheduler.StageInfo@513f39c
14/10/08 20:46:11 INFO StatsReportListener: task runtime:(count: 2, mean: 2666.000000, stdev: 9.000000, max: 2675.000000, min: 2657.000000)
14/10/08 20:46:11 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:11 INFO StatsReportListener:     2.7 s   2.7 s   2.7 s   2.7 s   2.7 s   2.7 s   2.7 s   2.7 s   2.7 s
14/10/08 20:46:11 INFO StatsReportListener: shuffle bytes written:(count: 2, mean: 50.000000, stdev: 0.000000, max: 50.000000, min: 50.000000)
14/10/08 20:46:11 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:11 INFO StatsReportListener:     50.0 B  50.0 B  50.0 B  50.0 B  50.0 B  50.0 B  50.0 B  50.0 B  50.0 B
14/10/08 20:46:11 INFO StatsReportListener: task result size:(count: 2, mean: 1848.000000, stdev: 0.000000, max: 1848.000000, min: 1848.000000)
14/10/08 20:46:11 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:11 INFO StatsReportListener:     1848.0 B        1848.0 B        1848.0 B        1848.0 B        1848.0 B        1848.0 B    1848.0 B        1848.0 B        1848.0 B
14/10/08 20:46:11 INFO StatsReportListener: executor (non-fetch) time pct: (count: 2, mean: 86.309428, stdev: 0.103820, max: 86.413248, min: 86.205607)
14/10/08 20:46:11 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:11 INFO StatsReportListener:     86 %    86 %    86 %    86 %    86 %    86 %    86 %    86 %    86 %
14/10/08 20:46:11 INFO StatsReportListener: other time pct: (count: 2, mean: 13.690572, stdev: 0.103820, max: 13.794393, min: 13.586752)
14/10/08 20:46:11 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:11 INFO StatsReportListener:     14 %    14 %    14 %    14 %    14 %    14 %    14 %    14 %    14 %
14/10/08 20:46:11 INFO ConnectionManager: Accepted connection from [eb175/10.1.69.175:36187]
14/10/08 20:46:11 INFO SendingConnection: Initiating connection to [eb175/10.1.69.175:35197]
14/10/08 20:46:11 INFO SendingConnection: Connected to [eb175/10.1.69.175:35197], 1 messages pending
14/10/08 20:46:11 INFO BlockManagerInfo: Added broadcast_2_piece0 in memory on eb175:35197 (size: 4.8 KB, free: 265.4 MB)
14/10/08 20:46:12 INFO MapOutputTrackerMasterActor: Asked to send map output locations for shuffle 0 to sparkExecutor@eb175:58085
14/10/08 20:46:12 INFO MapOutputTrackerMaster: Size of output statuses for shuffle 0 is 140 bytes
14/10/08 20:46:12 INFO TaskSetManager: Finished task 0.0 in stage 0.0 (TID 2) in 1428 ms on eb175 (1/1)
14/10/08 20:46:12 INFO DAGScheduler: Stage 0 (collect at HiveContext.scala:415) finished in 1.432 s
14/10/08 20:46:12 INFO TaskSchedulerImpl: Removed TaskSet 0.0, whose tasks have all completed, from pool 
14/10/08 20:46:12 INFO StatsReportListener: Finished stage: org.apache.spark.scheduler.StageInfo@6e8030b0
14/10/08 20:46:12 INFO StatsReportListener: task runtime:(count: 1, mean: 1428.000000, stdev: 0.000000, max: 1428.000000, min: 1428.000000)
14/10/08 20:46:12 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:12 INFO StatsReportListener:     1.4 s   1.4 s   1.4 s   1.4 s   1.4 s   1.4 s   1.4 s   1.4 s   1.4 s
14/10/08 20:46:12 INFO StatsReportListener: fetch wait time:(count: 1, mean: 0.000000, stdev: 0.000000, max: 0.000000, min: 0.000000)
14/10/08 20:46:12 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:12 INFO StatsReportListener:     0.0 ms  0.0 ms  0.0 ms  0.0 ms  0.0 ms  0.0 ms  0.0 ms  0.0 ms  0.0 ms
14/10/08 20:46:12 INFO StatsReportListener: remote bytes read:(count: 1, mean: 100.000000, stdev: 0.000000, max: 100.000000, min: 100.000000)
14/10/08 20:46:12 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:12 INFO StatsReportListener:     100.0 B 100.0 B 100.0 B 100.0 B 100.0 B 100.0 B 100.0 B 100.0 B 100.0 B
14/10/08 20:46:12 INFO SparkContext: Job finished: collect at HiveContext.scala:415, took 4.787407158 s
14/10/08 20:46:12 INFO StatsReportListener: task result size:(count: 1, mean: 1072.000000, stdev: 0.000000, max: 1072.000000, min: 1072.000000)
14/10/08 20:46:12 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:12 INFO StatsReportListener:     1072.0 B        1072.0 B        1072.0 B        1072.0 B        1072.0 B        1072.0 B    1072.0 B        1072.0 B        1072.0 B
14/10/08 20:46:12 INFO StatsReportListener: executor (non-fetch) time pct: (count: 1, mean: 80.252101, stdev: 0.000000, max: 80.252101, min: 80.252101)
14/10/08 20:46:12 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:12 INFO StatsReportListener:     80 %    80 %    80 %    80 %    80 %    80 %    80 %    80 %    80 %
14/10/08 20:46:12 INFO StatsReportListener: fetch wait time pct: (count: 1, mean: 0.000000, stdev: 0.000000, max: 0.000000, min: 0.000000)
14/10/08 20:46:12 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:12 INFO StatsReportListener:      0 %     0 %     0 %     0 %     0 %     0 %     0 %     0 %     0 %
14/10/08 20:46:12 INFO StatsReportListener: other time pct: (count: 1, mean: 19.747899, stdev: 0.000000, max: 19.747899, min: 19.747899)
14/10/08 20:46:12 INFO StatsReportListener:     0%      5%      10%     25%     50%     75%     90%     95%     100%
14/10/08 20:46:12 INFO StatsReportListener:     20 %    20 %    20 %    20 %    20 %    20 %    20 %    20 %    20 %
5078
Time taken: 7.581 seconds
View Code

注意：

  1. 在启动spark-sql时，如果不指定master，则以local的方式运行，master既可以指定standalone的地址，也可以指定yarn；
  2. 当设定master为yarn时(spark-sql --master yarn)时，可以通过http://$master:8088页面监控到整个job的执行过程；
  3. 如果在$SPARK_HOME/conf/spark-defaults.conf中配置了spark.master spark://eb174:7077，那么在启动spark-sql时不指定master也是运行在standalone集群之上。

6.遇到的问题及解决方案

① 在spark-sql客户端命令行界面运行SQL语句出现无法解析UnknownHostException:ebcloud（这是hadoop的dfs.nameservices）

14/10/08 20:42:44 ERROR CliDriver: org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in stage 1.0 failed 4 times, most recent failure: Lost task 0.3 in stage 1.0 (TID 4, eb174): java.lang.IllegalArgumentException: java.net.UnknownHostException: ebcloud
        org.apache.hadoop.security.SecurityUtil.buildTokenService(SecurityUtil.java:377)
        org.apache.hadoop.hdfs.NameNodeProxies.createNonHAProxy(NameNodeProxies.java:240)
        org.apache.hadoop.hdfs.NameNodeProxies.createProxy(NameNodeProxies.java:144)
        org.apache.hadoop.hdfs.DFSClient.<init>(DFSClient.java:579)
        org.apache.hadoop.hdfs.DFSClient.<init>(DFSClient.java:524)
        org.apache.hadoop.hdfs.DistributedFileSystem.initialize(DistributedFileSystem.java:146)
        org.apache.hadoop.fs.FileSystem.createFileSystem(FileSystem.java:2397)
        org.apache.hadoop.fs.FileSystem.access$200(FileSystem.java:89)
        org.apache.hadoop.fs.FileSystem$Cache.getInternal(FileSystem.java:2431)

原因：Spark无法正确获取HDFS的地址。因此，将hadoop的HDFS配置文件hdfs-site.xml拷贝到$SPARK_HOME/conf目录下。

 

②

14/10/08 20:26:46 INFO BlockManagerMaster: Updated info of block broadcast_0_piece0 
14/10/08 20:26:46 INFO SparkContext: Starting job: collect at HiveContext.scala:415 
14/10/08 20:29:19 WARN RetryInvocationHandler: Exception while invoking class org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB.getFileInfo over eb171/10.1.69.171:8020. Not retrying because failovers (15) exceeded maximum allowed (15) 
java.net.ConnectException: Call From eb174/10.1.69.174 to eb171:8020 failed on connection exception: java.net.ConnectException: Connection refused; For more details see: http://wiki.apache.org/hadoop/ConnectionRefused 
at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method) 
at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:57) 
at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45) 
at java.lang.reflect.Constructor.newInstance(Constructor.java:526) 
at org.apache.hadoop.net.NetUtils.wrapWithMessage(NetUtils.java:783) 
at org.apache.hadoop.net.NetUtils.wrapException(NetUtils.java:730) 
at org.apache.hadoop.ipc.Client.call(Client.java:1414) 
at org.apache.hadoop.ipc.Client.call(Client.java:1363)

原因：hdfs连接失败，原因是hdfs-site.xml未全部同步到slaves的节点上。

7.参考资料

  1. sparkSQL1.1入门之六：sparkSQL之基础应用
  2. Spark SQL CLI描述
  3. local class incompatible serialVersionUID
  4. Cannot Submit Tez Application

posted on 2014-10-08 21:26  BYRHuangQiang  阅读(8101)  评论(1)  编辑  收藏  举报

导航