Sqoop and Oracle tablespace issue

[Brenden Cobb <Br...@humedica.com>]

Hello- We're using Sqoop 1.4.5 with Oracle 12c and find that Sqoop --direct exports to Oracle are consuming drastically more space compared to Sqlldr.





Specifics:



Example data set: 24,087,140 records

The destination Oracle table is  partitioned on one column and the table uses basic compression.



Results for Sqoop export (command below):

2,818,572,288 bytes and  86,016 blocks



Results for Sqlldr:

872,415,232 bytes 26,624 blocks



Obviously a huge difference for Oracle storage resources.



Appreciate any advice/insight regarding these findings.





Sqoop command:



$ sqoop-1.4.5/bin/sqoop export \

-D mapred.child.java.opts='-Xmx4g' \

-D sqoop.export.records.per.statement=5000 \

-D sqoop.export.statements.per.transaction=1000 \

-D mapred.task.timeout=0 \

--connect jdbc:oracle:thin:@test:1521/TEST \

--username test \

--password test \

--direct \

--table TEST \

--input-null-string '\\N' \

--input-null-non-string '\\N' \

--export-dir test_test \

--fields-terminated-by '|' \

-m 18





Thanks