DSXchange

Search is your friend. There are lots of posts discussed here which will help you set up the right properties. Also, there is a property help button on the stage itself which will help you identify the purpose of that particular property. You can go over the DB2/UDB Load Stage documentation. Try to ...

You can also send a Delete from Table where <conditioin>, in the Before Sql tab of your database stage.

You can do it yes. But there are limitations. If your database is partitioned then it becomes even more complicated. Do a select max(id) from table and store the result in a hashed file keyed on a dummy key. Reference that key in your transformer and specify the constraint In.Key > Max.key where In....

I wrote a script for a friend of mine a while back to get the job elapsed time for all the jobs from a text file. The result is a comma delimitted file with job name and its elapsed time. Its fairly simple. If you are familiar with unix you will be able to figure it out. All you need to do is change...

This postwill be of some help.

Another reason might be that the database was down during that time. Try running the job again, if it still fails with the same error then try to log into the same database using your favourite tool. See if that connects.

Never kill a job that accesses a database, especially if its doing something like "clear table then insert rows". The job will abort, but the thread accessing the database stays alive and abandoned, hence keeping the table locked. If, at anytime, you have to stop a job, ask the dba to kill the threa...

True, there is an SCD stage in DataStage 8.0, but its not that hard to implement. As Ray pointed out, search this forum or even google to get the pseudocode. Then turn that into computational logic by use of different stages.

There are clients that expect job sequencers containing 10 jobs to finish in 2 seconds, if not they blame the design, as well as the developer and at worst datastage. Hate such clients which expect rocket science from an ETL tool. They are better off writing thousands of lines of sqls and spending ...

Test it out. The processing will not start untill both source and hashed file for the transformer are ready. But as a standard practice, build hashed files in a seperate job. This way if you have to restart your job, the hashed files wont have to be built from scratch.

talk2shaanc wrote:Note: 1. char(13) is carriage return

char(13):char(10) is the CR for dos.

Or to simplify your life, set the transaction size to 0 which means all or nothing. In other words, if the job fails, the inserted rows will be rolled back. This way every time you restart your job, it will start loading afresh.

Herewould be a good start

Time shall tell. A person never knows. Demand is increasing day by day. I hear my boss asks me to do stuff equalent to rocket science. Tool/hardware limitations set in. But once this barrier breaks, you never know where we will head off too.

I agree. You need to test it out. Performance is a very vague term. It really depends upon your data size, processing window.