I'm guessing given that you've asked this question twice you're not able/planning to provide the additional details, but a comments too short to explain the problem with your question, so here's some more information untill this one gets closed :)
Your question is essentially unanswerable. The reason for this is that there are a huge range of security assurance techniques that get lumped under the term "penetration testing" and the best answer depends on which of those you're looking for.
Which one you're looking for will depend on what your goals are. Now I'm going to assume that by asking this question your organisation hasn't had a lot of "pen tests" before and base this advice on that premise (but you see here where if you'd provided some more details in the question I wouldn't have had to make what could well be an erroneous assumption which might make this answer irrelevant)
There are essentially two approaches to security testing (I dislike the term pen testing), white box and black box. In a white-box test you provide the tester with credentials and knowledge of your environment. This allows for much greater levels of information to be extracted in a safe unintrusive fashion and will provide you much more information (if done correctly) than a black box test. In my opinion, if it's your first test, this is likely the best approach. A white-box test is good for answering the question "what security problems do we have in this environment"
A black box test takes a more "adversarial" standpoint, where the tester is provided no credentials/access and has to try to get them. This is appropriate once you think you have a good level of security and are looking for a test to prove that you have achieved that level of security. This is good for answering the question "have we achieved the level of security that we think we have"
Another relevent question is "what's my goal with this security test" realistically you may want the tester to find things to improve your security or you may want to minimize findings (for example where this test is compliance driven and you want to have a "clean report" for the organisation that mandated you have the test).
In the latter case I recommend that you definitely go for "black box" and restrict the scope as much as you can, as that will likely limit what the tester can find :)
Realistically that is however unlikely to actually help you improve your security. If you want to improve your security, I'd recommend providing the tester with as much information as possible and working with them to target the areas you're actually concerned about so that you get the best benefit from the testing you're doing.