Not very common. Impractical: one may need 1010 or more numbers for a large simulation.
Measurement of some random or chaotic physical process. Throw dice, observe turbulence in air flow, observe fluctuations in a lava lamp (yes, really), use radioactivity. Usually difficult to set up, and even more difficult to verify that the distribution is not biased in some way.
Design an algorithm that produces a series of seemingly random numbers given some starting value(s) (seed).
Random number generators, RNGs. Often called pseudo-random number generators to emphasize that the algorithm is really deterministic, although the result is supposed to look random.