You can model the functioning of each hard disc as a "toss of a coin" that has 0.97 chance to land on "working" and 0.03 chance to land on "not working", i.e. the probability of a hard disc working can (in the context of this problem at least) be model as single event with a 0.97 probability of working. With 3 hard discs you then have 3 such independent events, each with the same probability of working.
If you look at the Binomial distribution, it gives the probability P(X = k), where X is a stochastic variable representing the number of successes when you have n independent events each having probability p for success. In your case, n is 3 and if we choose to model "success" as "hard disc is working", then p is 0.97, and the answer to the problem is P(K >= 2) = P(X = 2) + P(X = 3).
Notice, that since you can replace k with n-k and p with 1-p in the Binomial distribution and get same result, you may also map "success" in the Binomial distribution to "hard disc not working" with probability p = 0.03, and then you get the answer as P(K' < 2) = P(K' = 0) + P(K' = 1), where K' now represent the number of hard discs that doesn't work, i.e. K' = 3 - K. This last formulation corresponds to the solution you found in post #3.