Test Cases for PX jobs

[sathyanveshi]

Hi,

I have developed a sequence and a set of DS PX jobs. I need to write the test cases for them. Can I have a list of indicative test cases that can be run on the sequence and the jobs?

Cheers,
Mohan

[ArndW]

Hello Mohan,

I must admit that I'm not quite sure what you are asking for - I would think that the testing phase is almost completely different depending on your job, data and environment. Are you testing stability, speed or correctness/completeness? But perhaps I misunderstood your question and someone else will assist.

[sathyanveshi]

Hi,

Thanks for the response.

Let me explain what I need. We have developed the sequence (and a set of jobs) catering to the requirements of the client. Now, as part of the deliverables, we need to submit the test case document. So, if you can let me know the different test cass (cases/conditions which can be tested), then it would be great.

If you still need more explanation, please let me know.

Cheers,
Mohan

[ArndW]

Mohan,

the test cases should be catered to the customer's requirements. I wouldn't know where to start - take my comments earlier as a start. Do you get data from flat files? If yes, how does your process react if files are missing or you get a file with the same name but different format? What happens if the Database is down when you start. The list goes on and on (ad absurdum) and the failure modes should be part of the design. Usually if data correctness is part of your testing procedure then you need someone OUTSIDE of the development team and INSIDE the business to check that aspect.

I think you need to play a devil's advocate role and think of everything you would intentionally break during normal processing.

[sathyanveshi]

Hi,

Thanks a lot for your inputs.

Let me consider your inputs as a starting point and proceed further.

Thanks again..

Cheers,
Mohan

[ray.wurlod]

Creating test data is an art. It helps if you're cynical, suspicious, or even outright paranoid.

Ideally you test every possible condition. Practically, you test all, or most, likely combinations of conditions.

You create a test plan showing the test data to be used in each case, the expected outcome and the actual outcome.

The decision is Boolean. All tests passed, or not.

[vmcburney]

Sequence jobs
When testing a sequence job you need to have a test case where each part of the sequence job fails. This tests the error handling of the job, how does it abort? How does it report the error?

Another test case is to place each called job into a status of aborted and then run the sequence job to confirm that it can reset and run those jobs.

You need test cases around failover restart. What is your strategy for restarting a failed job? Get the sequence job to fail at each stage and then test your restart capability.

Consider test cases for your manually entered job parameters, in particular how does the job react when the user makes an error and enters the wrong parameter. Check the error handling of this scenario.

Parallel Jobs
The test cases here are obvious, run each job and check the results. Your tests need to audit the row counts. I like setting up custom test data that has a wide range of data scenarios such as null field, blank fields, invalid dates, text fields with non standard characters etc. These files are good for most stages of testing and especially good for regression testing.[/b]

[s_boyapati]

Hi,

I have developed a sequence and a set of DS PX jobs. I need to write the test cases for them. Can I have a list of indicative test cases that can be run on the sequence and the jobs?

Cheers,
Mohan

You have to write and test all test cases to cover atleast the following creteria:

a) all input column definitions must be tested.
b) look ups testing
c) parameters testing
d) restartability features, if any you used
e) should be allowd or not allow partial data acceptance
f) data delays.

etc...etc to say. All above creteria falls under unit testing, system testing, integration testing etc... in testing terminology.

Sree

[scottr]

u r correct sree.my previous developer design a job which loads in to fact table using a wrong hash file.The job is running for couple of months and now the fact table has 46 mil records of wrong data.for the last one month i am working on it.now we are updating the entire fact table.

i think each stage,each source,each lookup can contribute N number of test cases. am i right??

[s_boyapati]

u r correct sree.my previous developer design a job which loads in to fact table using a wrong hash file.The job is running for couple of months and now the fact table has 46 mil records of wrong data.for the last one month i am working on it.now we are updating the entire fact table.

i think each stage,each source,each lookup can contribute N number of test cases. am i right??

Yes. Depends up on complexity. Some Stages we use for business logic control and some as apart of Algorithm( which is also apart of business logic, but mostly driven by technology). You are right.

Sree