如何根据hadoop中的特定条件在一个datanode上存储数据？(How to store data on one datanode based on specific criteria in hadoop?)

 是否有可能将数据保存在hadoop中的一个数据节点上，这意味着它不是（或非常有限，就像它用完空间时）分布在整个系统中（复制很好）。 就像我保留每分钟的记录日志并将它们存储在hadoop上的文件夹结构中，如下所示： 
 
 
 
 /年月日 
 
 
 现在我想强制hadoop将例如每个月的文件夹存储在一个（或者更多，如果超出空间）datanode，所以当我执行读取请求时，namenode应该在最好的情况下只返回一个datanode作为数据块的位置。 
 
 这可能吗？ 如何在Java中实现（比如在datanode上将所有文件基于父文件夹的哈希分组）？ 或者这是一个hadoop配置？ 

Is it possible to keep data on one datanode in hadoop, meaning it is not (or very limited like when it runs out of space) distributed across the system (replication is fine). Like I keep record logs for every minute and store them in a folder structure on hadoop that looks like this:
 
 
 
/year/month/day
 
 
Now I want to force hadoop to store e.g. every month folder on one (or more if space is exceeded) datanode only, so when I perform a read request the namenode should in the best case only return one datanode as the location of the data blocks.
 
Is this possible? How would an implementation in Java look like (like grouping all files based on the hash of the parent folders together on a datanode)? Or this this a hadoop configuration? 

原文：https://stackoverflow.com/questions/37603415