Type: New Feature
Affects Version/s: None
Fix Version/s: None
Currently, hive assumes a single metastore and the HADOOP_HOME is passed as a environment variable.
It would be desirable to support hive on top of multiple data centers (dfs + mr).
For eg. there could be 2 metastores: primary and secondary. They would have different dfs's , and there will be a
dfs->mr mapping maintained by the metastore.
Hive would be enhanced to support multiple metastores and all operations (ddl + query) would span multiple metastores.
Different consistency pluggable policies can be employed - for eg. if a table/partition can be present in both the metastores with different
last modification times, either the last one can be used or an error can be thrown.
It will be upto the application (outside hive) to copy the data from one metastore to another, and to maintain consistency inside.