We study the percolation problem on the Apollonian network model. The Apollonian networks display many interesting properties commonly observed in real network systems, such as small-world behavior, scale-free distribution, and a hierarchical structure. By taking advantage of the deterministic hierarchical construction of these networks, we use the real-space renormalization-group technique to write exact iterative equations that relate percolation network properties at different scales. More precisely, our results indicate that the percolation probability and average mass of the percolating cluster approach the thermodynamic limit logarithmically. We suggest that such ultraslow convergence might be a property of hierarchical networks. Since real complex systems are certainly finite and very commonly hierarchical, we believe that taking into account finite-size effects in real-network systems is of fundamental importance.