Since spectrum below 6 GHz bands does not suffice to meet the high bandwidth requirements of 5G use cases, 5G networks expand their operation to mmWave bands with multi-GHz of available spectrum. However, operation at mmWave bands raises two challenges: high penetration loss and susceptibility to objects blocking the line-of-sight. As a remedy, transmitters point their antennas to their receivers using beamforming to contain the signal in the direction of the intended receiver. To avoid loss of connectivity due to blockers or other factors such as rain, users can be connected to multiple base stations (BS). While these two features, beamforming and multi-connectivity, can mitigate the challenges of mmWave bands, it is not trivial to find the optimal user association which can maintain the minimum rate requirements of the users. In this paper, we first formulate a throughput-maximizing user association scheme considering beamforming and multi-connectivity, and investigate the behavior of the optimal scheme to devise a simple, yet efficient heuristic that does not require global network information. Our proposal, dubbed beam-align, which only uses local information at a BS, performs close to the optimal in terms of per-user capacity and user satisfaction levels and has only polynomial complexity. Our numerical analysis shows that beam-align outperforms the frequently-used association schemes based on SNR. To investigate the performance of beam-align also under various challenging settings, we consider three scenarios: presence of blockers, different levels of rain intensity, and presence of clustered users. From simulations, we observe that beam-align is robust against different conditions, as it maintains a stable performance under the studied scenarios.