Block源代码


Block是HDFS中的最基本单位是对数据块的抽象,它通过一个final long(blkid)来进行标识。一个Block拥有三个longs,它们分别是block-id 、block length和generation stamp。Block名的格式是”blk”+blkid。通过set和get等方法能够得到关于Block的id、name、len等信息
 
一个Block对应着两个文件,其中一个存数据,一个存放元数据信息。它的元数据信息文件的格式如下:”blk_”+blkid+”_”+version.meta
和Block相关的类如下所示:

 
  private long blockId;
 
例如${hadoop.tmp.dir}/dfs/data/current /blk_826540629399449945,这一串数字就是blockId
 blockId 是block编号
 
  private long numBytes;
 blockId 是文件实际大小
 
  private long generationStamp;
 类似个时间戳,每次修改加1
 
在equals中用到 generationStamp
在hashcode没有用到generationStamp
 
 
 
 
文件名称和block的转换
  static long filename2id(String name) {
    return Long.parseLong(name.substring("blk_".length()));
 
  }
 
 
 
  public boolean   public boolean equals(Object o) {    if (!(o instanceof Block)) {      return false;    }    final Block that = (Block)o;    //Wildcard generationStamp is ALLOWED here    return this.blockId == that.blockId      && GenerationStamp.equalsWithWildcard(          this.generationStamp, that.generationStamp);  }
  /** {@inheritDoc} */  public int hashCode() {    //GenerationStamp is IRRELEVANT and should not be used here    return 37 * 17 + (int) (blockId^(blockId>>>32));  }(Object o) {    if (!(o instanceof Block)) {      return false;    }    final Block that = (Block)o;    //Wildcard generationStamp is ALLOWED here    return this.blockId == that.blockId      && GenerationStamp.equalsWithWildcard(          this.generationStamp, that.generationStamp);  }
  /** {@inheritDoc} */  public int hashCode() {    //GenerationStamp is IRRELEVANT and should not be used here    return 37 * 17 + (int) (blockId^(blockId>>>32));  }
 
 

猜你喜欢

转载自zhaomengsen.iteye.com/blog/2059644