Wouldn't want to add much to bookbinder's link, although of course feel free to come back with questions.
Two things I will add, though:
1. A review article should be designed to help anybody unfamiliar with the topic under review to know what is going on with the field -- they can't be expected to be made experts based solely on your review, of course, but they should at least know where to start!
2. A review article should also be designed to convince people who are as expert as, or more expert than, you are, that you know what you're talking about. It's impossible to miss nothing, but you should be sure to miss nothing important. Use your judgement about what's important: seminal papers from years past that are important to the field will be highly-cited, and are more or less necessary in any review; for more recent work, it's probably more relevant to give an impression of what's going on now than including everything modern.
3. It is impossible to read every paper you're intending to review, but make sure that if you're going to discuss a particular paper in more detail that you've actually read and understood it.