CertCities.com -- The Ultimate Site for Certified IT Professionals
Free CertCities.com Newsletter via E-mail Share share | bookmark | e-mail
  Microsoft®
  Cisco®
  Security
  Oracle®
  A+/Network+"
  Linux/Unix
  More Certs
  Newsletters
  Salary Surveys
  Forums
  News
  Exam Reviews
  Tips
  Columns
  Features
  PopQuiz
  RSS Feeds
  Press Releases
  Contributors
  About Us
  Search
 

Advanced Search
  Free Newsletter
  Sign-up for the #1 Weekly IT
Certification News
and Advice.
Subscribe to CertCities.com Free Weekly E-mail Newsletter
CertCities.com

See What's New on
Redmondmag.com!

Cover Story: IE8: Behind the 8 Ball

Tech-Ed: Let's (Third) Party!

A Secure Leap into the Cloud

Windows Mobile's New Moves

SQL Speed Secrets


CertCities.com
Let us know what you
think! E-mail us at:



 
 
...Home ... Editorial ... Columns ..Column Story Saturday: April 5, 2014


 SELECTions  
Damir Bersinic
Damir Bersinic


 Have Data, Will Travel
Transporting tablespaces in Oracle 8i.
by Damir Bersinic  
2/14/2001 -- A client recently came to me to see if there was a better way to keep their regional offices updated with pricing on over a million part numbers that they have. The current method was not working, and the problem was compounded by the fact that bill of materials explosions required to price things properly in their manufacturing process changed on a regular basis and had a symbiotic relationship with the part numbers.

Up to this point they performed bulk loads of the changed data and objects at the regional offices on a monthly basis using SQL*Loader and scripts. Although the interval for updates was not going to change, to create indexes and other objects took the better part of a weekend and left almost no time to correct problems if they occurred. In a few cases, Monday morning operations, such as processing orders, had to wait until Tuesday because of problems with the load.

This problem is really one of "What's the fastest and easiest way to get data from one Oracle database to another with the lowest possibility of failure?" In looking at the problem, at least four ways exist to solve it:

  • Using Oracle's Export and Import utilities -- This method will extract the tables and its dependent objects from the source database (using the EXP utility) and create an Oracle proprietary file that can be imported into the target database (using the IMP utility). This method can be scripted and is conceptually simple. However, it will need to drop and re-create the objects in the target system making it slow. Probably not a good choice for this problem.

  • Using scripts and the SQL*Loader utility -- Similar to Export and Import, this method can be scripted and will work. However, the client already indicated that it takes too long so it is not appropriate. However, if we needed to get data from a third-party data source into Oracle, this would be a good way of doing it.

  • Using Distributed Database and Replication -- While this method will ensure that data exists on both servers, based upon how distribution is configured, it may require permanent network connections and may generate a lot of traffic on the pipe between the head office and the regional offices. If you've tried to configure replication and/or distributed database with Oracle before, you'll know that this is not the easiest thing to do. In any case, it won't work for this client as regional offices do not have much available bandwidth, and they're not likely to get more in the short term.

  • Using Transportable Tablespaces -- In Oracle 8i you can transport an entire tablespace, including its contents (tables, indexes, etc.), from one database to another. Because the data for the part numbers, bill of materials, and other required elements were contained on a single tablespace (or could be moved there by using Oracle 8i's ALTER TABLE … MOVE command) this was the chosen method. Let's look at this in more detail.

Oracle 8i's Tablespace Enhancements
Prior to Oracle 8i, management of extents allocated to segments was done by making changes to the data dictionary every time a new extent had to be allocated to a table, index or cluster. This created inefficiency as the act of allocating additional storage to an object required a hit on the SYSTEM tablespace, which now became a potential point of contention. Ways to avoid this were to carefully size extents and allocate them manually to minimize the hit.

With 8i, it is now possible to manage the extents on a tablespace locally on that tablespace. This is done by specifying the EXTENT MANAGEMENT LOCAL option when issuing the CREATE TABLESPACE command. This tells Oracle to manage the allocation of extents to objects on the tablespace by using a set of bitmaps on the tablespace instead of in the data dictionary. This also has the added benefit of allowing the tablespace to exist as a self-contained unit with all the information about storage allocation for objects created on the tablespace within the contents of the data files that it is composed of. In other words, it now becomes possible to take the tablespace on the road and transport it.

Just Like Airfares, Transporting Tablespaces Has Rules
Before you can take the contents of a tablespace on the road, you need to make sure that all the pieces on it are O.K. -- kinda like making sure the whole family is in the car before starting to drive off (and that nobody chained the bumper to the house!). To transport a tablespace, the databases for the source and destination need to be almost identical, which means:
  • Same DB_BLOCK_SIZE for both databases.
  • Same Character Set and National Language Character Set for both databases.
  • Same operating system for both databases.
  • The tablespace to be transported must not have any function-based indexes, scoped REFs, domain indexes or use any Oracle 8.0 advanced queuing functionality.
  • All data must be self-contained: indexes and tables on the same tablespace, BLOBs and data segments on the same tablespace, parent and child tables in a primary key-foreign key relationship on the same tablespace (though you can choose not to transport all of the PK-FK data), all partitions for a table must be on the same tablespace, etc.
  • The tablespace name must not exist on the target database. If it does, it needs to be deleted before the source tablespace is attached to the destination database.

In the case of my client, not all of these were true, but the major ones (same operating systems, dB_BLOCK_SIZE, character set, etc.) were, so all we had to do was move indexes to the tablespace from their previous location and ensure that all the elements needed were in one spot.

The Car is Loaded, Lets Move that Tablespace!!
In order to transport a tablespace from one database to another, you perform the following steps:

  1. Ensure that the data on the tablespace is self-contained. This is done by running the DBMS_TTS.TRANSPORT_SET_CHECK procedure passing it the name of the tablespaces you want to transport. This procedure populates a view called TRANSPORT_SET_VIOLATIONS that you can query to see if you have any errors that would cause the transport to fail. If you do, correct them and try again.
  2. Generate a Transportable Tablespace Set that includes a list of all the objects on the tablespace that can be imported on the destination database. This is done by first altering the tablespace to be transported to READ ONLY and making a copy of the data files belong to the tablespace after doing so.
  3. You then run the EXP utility to extract the data dictionary information to be imported into the destination database. To do this, pass the EXP utility the TRANSPORT_TABLESPACE=y parameter, along with the names of tablespaces to be transported, and whether or not to export constraints, security, and other settings. You now have a set of data files (with all the data) that can be copied to the destination database, as well as an Oracle export file that can used to create the data dictionary information about the objects in the datafiles.
  4. Alter the tablespace back to READ WRITE mode in the source database so that the next round of updates can be started.
  5. Transport the export file and data files to the destination database. You can use CD, tape, ZIP file or any other method that makes sense, based upon your data size.
  6. If the tablespace to be attached to the destination database already exists, drop it along with its contents. This will be true where you are frequently updating data transferred from a central office to a branch office, as is the case here. Issue the DROP TABLESPACE command with the INCLUDING CONTENTS CASCADE CONSTRAINTS parameters.
  7. Copy the data files from the source system to the hard disk on the destination (make a note of where they are as you will need this later).
  8. Run the IMP utility to import data dictionary information. You need to specify the TRANSPORT_TABLESPACE=y parameter and the DATAFILES parameter, followed by the physical location of the datafiles on the destination computer's hard disk. If the users who own the objects in the tablespace do not exist in the destination database, you may also need to use the FROMUSER and TOUSER parameters. This is a relatively quick process, as only the data dictionary information about the objects on the tablespace is added since the data is already in the datafiles.

If you want to make modifications on the destination database, you can alter the tablespace to READ WRITE, although in this client's case we left the tablespace READ ONLY as all updates were made at the central office.

Still Seems Like a Lot of Work To Me!
While the steps required to move the entire contents of a tablespace from one database to another may appear restrictive, keep in mind that you can move more than one tablespace at a time to ensure all data is self-contained. It is also easier to copy files and run an IMP command on the destination database than to use SQL*Loader or a series of export files, and SQL scripts to create the indexes after the fact. Furthermore, the central office can now specify the storage parameters for tables and indexes ensuring that space is used efficiently and not leave this up to the regional admins or Oracle defaults.

The end result in my client's case? More control and fewer errors. It also allowed them to keep a copy of each update sent to the regional offices on a CD-ROM, along with the export files in case they needed to go to an earlier incarnation of the database structure -- an added bonus.

By the way, did you notice that the database instances never had to be shut down to do this in either the source or destination?

Damir Bersinic is an Infrastructure Consultant with Trecata Corporation, a systems integration consultancy in Toronto, Canada. He has more than 20 years of industry experience and has written a number of books on Oracle, Windows, SQL Server and Active Directory. His most recent book is "Portable DBA: SQL Server" from Osborne McGraw-Hill. He is also an Oracle Certified Professional (DBA), MCSE, MCDBA and MCT. He can be reached by e-mail at .

 


More articles by Damir Bersinic:

-- advertisement --


There are 32 CertCities.com user Comments for “Have Data, Will Travel”
Page 1 of 4
3/1/01: Manoj says: Damir, Its really Great !!!
8/26/01: Raj says: The Provided information is very helpfull in all the ways.This is really a good site..
9/10/01: mohamed says: i look for oracle developer training cource may you help me?
10/24/01: Eman says: i look for oracle8i materials?
12/12/01: R.G.Senthil says: Respectsd sir, Iam from india,my name R.G.Senthil kumar Please tell the detailse about oracle administrator , Gide ,And tell the features ,And exam detailse Thanking Yoy, urs faithfully R.G.Senthil kumar
1/11/02: Yash from India says: Excellent
2/3/02: Vivek Arora from Hardwar, India says: Respected Sir, Please send more features of Oracle 8i that can enhance working in Oracle 8i. Thank You
2/28/02: Senthilnathan from Malaysia says: I have few doubts in oracle datafile storage, First I have to create database with no tables except system tables. After that I created one table and reload data from backup, then I checked datafile size is 2 GB. Then I deleted that table records used truncate method, then i saw the datafile size remain 2 GB. I would like to know why datafile not shrinking after deleted records. What is the mechnism using oracle for storing reocrds into datafile(users01.dbf). Please give me the answer.
11/14/02: Shahrukh Khan from India says: Superb, and I am looking for oracle8i materials?
2/1/03: Srinivasan from Doha - Qatar says: For the Question of Senthilnathan from Malaysia you have to specify the option of autoshrink for your datafiles to automatically reducde after deleting records
First Page   Next Page   Last Page
Your comment about: “Have Data, Will Travel”
Name: (optional)
Location: (optional)
E-mail Address: (optional)
Comment:
   

-- advertisement (story continued below) --

top