But if there is two 'first photon' which is not coherence each other, then after some actions(induced emissions) there will be two family of photons inside of laser.
This is entirely normal. The "first photons" that start the laser action are generated by spontaneous emission. They are not coherent with each other, and there are normally more than one present in the cavity at a time.
how can we make two incoherent families of photons coherent and make real laser?
What happens is, each time a photon from group "1" stimulates a new emission, that puts the excited atom involved back to the ground state. This reduces the number of excited atoms available to provide gain for group "2".
This produces a kind of positive feedback mechanism so that if group "1" has slightly more photons than group "2", then group "1" will quickly starve group "2" of gain, and group "1" will come to dominate.
This is related to the gain pinning effect. If you increase the pump power of a laser above the threshold level, typically this doesn't increase the population inversion, because stimulated emission increases and the excited population is locked at the threshold level (while output power increases). This means the gain spectrum is also "pinned" at the threshold level, so if the group "1" photons have a frequency at the peak of the gain spectrum, but group "2" are slightly off the peak, then group "2" will not have sufficient gain to overcome the cavity loss and will quickly die away.
However, it is also possible that group "1" and "2" are not interacting with exactly the same set of atoms in the gain medium. For example, if group "1" are in a different spatial mode than group "2" so that they overlap a different part of the gain material. In this case both groups can lase simultaneously, and you have a multi-mode laser.