0
我有两个数据集已经使用相同的分区程序进行分区并存储在HDFS中。这些数据集是两个不同的Spark作业的输出,我们没有控制权。现在,我想加入这两个数据集来产生不同的信息。用Spark加入HDFS中的两个数据文件?
Example:
Data Set 1:
ORDER_ID CUSTOMER_ID ITEMS
OD1 C1 1,2,3 -> partition 0
OD2 C2 3,4,5 -> partition 0
OD3 C4 1,2,3 -> partition 1
OD4 C3 1,3 -> partition 1
Data Set 1:
ORDER_ID CUSTOMER_ID REFUND_ITEMS
OD1 C1 1 -> partition 0
OD2 C2 5 -> partition 0
OD3 C4 2,3 -> partition 1
OD4 C3 3 -> partition 1
Options are:
1) Create two RDDs from the datasets and join them.
2) Create one RDD using one of the dataset.
-> For each partition in the RDD get the actual partition id i.e OD1 -> 0, OD3 -> 1 (using some custom logic)
-> Load data from HDFS for that partition for dataset 2
-> Iterate over both the dataset and produce combined result.
For option 2 I don't know how to read a specific file form HDFS in the Spark executor. (I have the full URI for location of the file)