Oracle Data Pump

Data Pump takes the old export and import utilities one step further, you can have total control over the job running (stop it, pause it, check it, restart it). Data pump is a server side technology and it can transfer large amounts of data very quickly using parallel streams to achieve maximum throughput, they can be 15-45% faster than the older import/export utilities. Advantages using data pump are

ability to estimate jobs times
ability to restart failed jobs
perform fine-grained object selection
monitor running jobs
directly load a database from a remote instance via the network
remapping capabilities
improved performance using parallel executions

A couple of notes is that you cannot export to a tape device only to disk, and the import will only work with version of oracle 10.1 or greater. Also remember that the expdp and impdp are command line tools and run from within the Operating System.

Data Pump Uses

You can use data pump for the following

migrating databases
copying databases
transferring oracle databases between different operating systems
backing up important tables before you change them
moving database objects from one tablespace to another
transporting tablespace's between databases
reorganizing fragmented table data
extracting the DDL for tables and other objects such as stored procedures and packages

Data Pump components

Data pump technology consists of three major components

dbms_datapump - the main engine for driving data dictionary metadata loading and unloading
dbms_metadata - used to extract the appropriate metadata
command-line - expdp and impdp are the import/export equivalents

Data Access methods

Data pump has two methods for loading data, direct path or external table path you as a dba have no control with what data pump uses, normally simple structures such as heap tables without triggers will use direct path more complex tables will use the external path, oracle will always try and use the direct-path method.

Direct Path	bypasses the database buffer cache and writes beyond the high water mark when finished adjusts the high water mark, No undo is generated and can switch off redo as well, minimal impact to users as does not use SGA. Must disable triggers on tables before use.
External Path	Uses the database buffer cache acts as a SELECT statement into a dump file, during import reconstructs statements into INSERT statements, so whole process is like a normal SELECT/INSERT job. Both undo and redo are generated and uses a normal COMMIT just like a DML statement would.

In the following cases oracle will use the external path if any of the below are in use

clustered tables
active triggers in the table
a single partition in a table with a global index
referential integrity constraints
domain indexes on LOB columns
tables with fine-grained access control enabled in the insert mode
tables with BFILE or opaque type columns

Data Pump files

You will use three types's of files when using data pump, all files will be created on the server.

dump files - holds the data and metadata
log files - the resulting output from the data pump command
sql files - contain the DDL statements describing the objects included in the job but can contain data
Master data pump tables - when using datapump it will create tables within the schema, this is used for controlling the datapump job, the table is removed when finished.

Data Pump privileges

In order to advance features of data pump you need exp_full_database and imp_full_database privileges.

How Data Pump works

The Master Control Process (MCP), has the process name DMnn, only one master job runs per job which controls the whole Data Pump job, it performs the following

create jobs and controls them
creates and manages the worker processes
monitors the jobs and logs the process
maintains the job state and restart information in the master table (create in the users schema running the job)
manages the necessary files including the dump file set

The master process creates a master table which contains job details (state, restart info), this table is created in the users schema who is running the Data Pump job. Once the job has finished it dumps the table contents into the data pump file and deletes the table. When you import the data pump file it re-creates the table and reads it to verify the correct sequence in which the it should import the various database objects.

The worker process is named DWnn and is the process that actually performs the work, you can have a number of worker process running on the same job (parallelism). The work process updates the master table with the various job status.

The shadow process is created when the client logs in to the oracle server it services data pump API requests, it creates the job consisting of the master table and the master process.

The client processes are the expdp and impdp commands.

Running Data Pump

You can either run via a command line specifying options or use a parameter file, there are many options to Data Pump so it would be best to check out the Oracle documentation, I have given a few examples below

Exporting
database	expdp vallep/password directory=datapump full=y dumpfile=data.dmp filesize=2G parallel=2 logfile=full.log Note: increase the parallel option based on the number of CPU's you have
schema	expdp sys/password schemas=testuser dumpfile=data.dmp logfile=schema.log
table	expdp vallep/password tables=accounts,employees dumpfile=data.dmp content=metadata_only
tablespace	expdp vallep/password tablespaces=users dumpfile=data.dmp logfile=tablespace.log
Importing
database	impdp system/password full=y dumpfile=data.dmp nologfile=y
schema change	impdp system/password schemas=’HR’ remap_schema=’HR:HR_TEST’ content=data_only impdp system/passwd remap_schema=’TEST:TEST3’ tables=test log=… dumpfile=… directory=…
Other Options
directory	specifies a oracle directory object
filesize	split the dump file into specific sizes (could be used if filesystem has 2GB limit)
parfile	specify the parameter file
content	contents option can be ALL, METADATA_ONLY or DATA_ONLY
compression	compression is used by default but you can stop it
exclude/include	metadata filtering
query	selectively export table data using a SQL statement
estimate	Calculate job estimates where the vaild keywords are blocks and statistics
estimate_only	Calculate job estimates without performing the export
network link	you can perform a export across a network
encryption	you can encrypt data within the data pump file
parallel	increase worker processes to increase throughput, base it on number of CPU's
remap_schema	move objects from one schema to another
remap_datafile	change the name of the datafile when moving across different systems
remap_tablespace	move from one tablespace to another
Useful Views
DBA_DATAPUMP_JOBS	summary information of all currently running data pump jobs
DBA_DATAPUMP_SESSIONS	displays the user currently running data pump jobs
V$SESSION_LONGOPS	display information like totalwork, sofar, units and opname
Privileges
IMP_FULL_DATABASE	required if using advanced features
EXP_FULL_DATABASE	required if using advanced features

DBMS_DATAPUMP package

The package dbms_datapump can be used for the following

starting/stopping/restarting a job
monitoring a job
detaching from a job

exporting	declare d1 number; begin d1 := dbms_datapump.open('export','schema',null, 'test1', 'latest'); dbms_datapump.add_file(d1, 'test1.dmp', 'dmpdir'); dbms_datapump.metadata_filter(d1, 'schema_expr','in (''OE'')'); dbms_datapump.start_job(d1); dbms_datadump.detach(d1); end;
importing	declare d1 number; begin d1 := dbms_datapump.open('import','full',null, 'test1'); dbms_datapump.add_file(d1, 'test1.dmp', 'dmpdir'); dbms_datapump.metadata_remap(d1, 'remap_schema', 'oe', 'hr'); dbms_datapump.start_job(d1); dbms_datadump.detach(d1); end;