Hi Mani!
This time I'll give my personal opinion, I really prefer to choose the "best fit" for each table. For example, find the best #mappers for each kinda table, if they need compression or a special where clause, transform a datatype from DB to another in Hadoop. About the performance issues, guess it's hard to have more performance using import-all than import each table separately, especially in a large number of tables like your case and even more if you're planning to run ETL over them.
Like I said above, it's just my humble opinion. Also, you can take a look at the documentation, there are some rules to use import-all:
http://sqoop.apache.org/docs/1.4.2/SqoopUserGuide.html#_literal_sqoop_import_all_tables_literal
- Each table must have a single-column primary key.
- You must intend to import all columns of each table.
- You must not intend to use non-default splitting column, nor impose any conditions via a
WHERE clause.
Another good reason to take the import table as an option is the fact that you may have some troubles during the import process of all tables at once. To debug the problem it will take longer, and if you attempt to fix a possible issue on the last tables you can get bored to wait for the whole job to finish to see if the fix worked 😄
Hope this helps!
