tachyon与hdfs,以及spark整合

Tachyon 0.7.1伪分布式集群安装与测试: 
http://blog.csdn.net/stark_summer/article/details/48321605 
从官方文档得知，Spark 1.4.x和Tachyon 0.6.4版本兼容，而最新版的Tachyon 0.7.1和Spark 1.5.x兼容，目前所用的Spark为1.4.1，tachyon为 0.7.1

tachyon 与 hdfs整合

修改tachyon-env.sh

export TACHYON_UNDERFS_ADDRESS=hdfs://master:8020Dtachyon.data.folder=$TACHYON_UNDERFS_ADDRESS/tmp/tachyon/data12

上传文件到hdfs

 hadoop fs -put /home/cluster/data/test/bank/ /data/spark/

 hadoop fs -ls /data/spark/bank/Found 3 items-rw-r--r--   3 wangyue supergroup    4610348 2015-09-11 20:02 /data/spark/bank/bank-full.csv-rw-r--r--   3 wangyue supergroup       3864 2015-09-11 20:02 /data/spark/bank/bank-names.txt-rw-r--r--   3 wangyue supergroup     461474 2015-09-11 20:02 /data/spark/bank/bank.csv1234567

通过tachyon 读取/data/spark/bank/bank-full.csv文件

val bankFullFile = sc.textFile("tachyon://master:19998/data/spark/bank/bank-full.csv/bank-full.csv")2015-09-11 20:08:20,136 INFO  [main] storage.MemoryStore (Logging.scala:logInfo(59)) - ensureFreeSpace(177384) called with curMem=630803, maxMem=2579182382015-09-11 20:08:20,137 INFO  [main] storage.MemoryStore (Logging.scala:logInfo(59)) - Block broadcast_3 stored as values in memory (estimated size 173.2 KB, free 245.2 MB)2015-09-11 20:08:20,154 INFO  [main] storage.MemoryStore (Logging.scala:logInfo(59)) - ensureFreeSpace(17665) called with curMem=808187, maxMem=2579182382015-09-11 20:08:20,155 INFO  [main] storage.MemoryStore (Logging.scala:logInfo(59)) - Block broadcast_3_piece0 stored as bytes in memory (estimated size 17.3 KB, free 245.2 MB)2015-09-11 20:08:20,156 INFO  [sparkDriver-akka.actor.default-dispatcher-2] storage.BlockManagerInfo (Logging.scala:logInfo(59)) - Added broadcast_3_piece0 in memory on localhost:41040 (size: 17.3 KB, free: 245.9 MB)2015-09-11 20:08:20,157 INFO  [main] spark.SparkContext (Logging.scala:logInfo(59)) - Created broadcast 3 from textFile at <console>:21bankFullFile: org.apache.spark.rdd.RDD[String] = MapPartitionsRDD[7] at textFile at <console>:21123456789

count

bankFullFile.count()
但是发现报错如下:
2015-09-11 21:34:31,494 WARN  [Executor task launch worker-6]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,495 WARN  [Executor task launch worker-6]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,489 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,495 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,495 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,495 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,495 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,495 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,496 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,496 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,496 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,496 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,496 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,496 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing2015-09-11 21:34:31,496 WARN  [Executor task launch worker-7]  (RemoteBlockInStream.java:retrieveByteBufferFromRemoteMachine(320)) - Read nothing123456789101112131415161718

感觉错误很诡异,有人知道这是什么原因？tell me why?

但是 我在tachyon 文件系统中可以看到如下内容:

./bin/tachyon tfs ls /data/spark/bank/bank-full.csv/4502.29 KB09-11-2015 20:09:02:078  Not In Memory  /data/spark/bank/bank-full.csv/bank-full.csv123

而bank-full.csv在hdfs文件是

hadoop fs -ls /data/spark/bank/Found 3 items-rw-r--r--   3 wangyue supergroup    4610348 2015-09-11 20:02 /data/spark/bank/bank-full.csv-rw-r--r--   3 wangyue supergroup       3864 2015-09-11 20:02 /data/spark/bank/bank-names.txt-rw-r--r--   3 wangyue supergroup     461474 2015-09-11 20:02 /data/spark/bank/bank.csv123456

其实Tachyon本身将bank-full.csv文件加载到了内存，并存放到自身的文件系统里面：tachyon://master:19998/data/spark/bank/bank-full.csv/bank-full.csv” 
Tachyon的conf/tachyon-env.sh文件里面配置的，通过export TACHYON_UNDERFS_ADDRESS=hdfs://master:8020配置,这样tachyon://localhost:19998就可以获取hdfs文件指定路径文件

好吧，那我就先通过hdfs方式读取文件然后 保存到tachyon

scala> val bankfullfile =  sc.textFile("/data/spark/bank/bank-full.csv")
scala> bankfullfile.countres0: Long = 45212scala> bankfullfile.saveAsTextFile("tachyon://master:19998/data/spark/bank/newbankfullfile")12345

未完成,待续~

评论

1 楼 haorengoodman 2015-12-17  

Tachyon 能在做数据分类吗？
例如我有一坨hdfs文件，将这些文件加载到Tachyon ，但是之前的文件目录结构不符合现在的要求，需要重新划分文件目录结构，可以做到吗？？？