Ah, might I suggest that you don't actually want random selection in a case like this? I am reading what you said as doing a full end to end inventory every six months while at some unspecified interval in between you want to make some spot checks. Suppose for the sake of discussion that your spot checks are once a month. So, on month 1 you check item XYZ and then on month 2, since you are doing a random selection, XYZ might come up again while some other item might go months and months without being checked.
There is an art to cycle counting. One of the drivers is often the warehouse setup, i.e., it can be easier to count everything on one set of racks and then rotate which racks are checked each month. But, the best strategy is to do something purposive. Best I have seen and done is to do a classification using one of the ABC types of classification which figures in the rate of change and the value of the item. I.e., A is all of the items which have the most value moving through the inventory. Actually, Gordon Graham recommends classification into 13 levels ... you might look up his books. The point being that cycle counting the stuff in A is way more important than counting the stuff in the bottom category (often called X regardless of how many levels). In fact, there may be so little value and movement in the X category that you don't count it at all until the next full sweep.
The other consideration which is often thrown in here is stock levels. If there is high turn over and stock levels can get low prior to being replenished, e.g., if it is a publisher and one needs to wait for a new print run or it is simply an item with a long lead time, then it is good to count it when the level gets low. Good, both because it gets you an accurate level when you don't have many left and because there aren't as many to count.
This means setting up a system in which you record the date of the last count and come up with a selection mechanism based on classification, stock level, and time since last count. That's more work than making a random selection, but it will be far better for the business.