You were not far from the result.
A parametric description of this line is
$$\pmatrix{x\\y\\z}=P_0+tV_1=\pmatrix{0\\1\\2}+t\pmatrix{3\\-1\\-2}$$
Why this expression of $V_1$ ?
Because vector $V_1$ ,which has to be orthogonal to the normal vector to the plane $\pmatrix{1\\1\\1}$ and to the direction vector $\pmatrix{1\\-1\\2}$ has thus to be proportional to the cross product
$$\tag{*}\pmatrix{1\\1\\1} \times \pmatrix{1\\-1\\2}=\pmatrix{3\\-1\\-2}.$$
Edit: Why where you close to the result ? Your relationships
$$\begin{cases}(2) : a_1+a_2+a_3=0\\(2) : a_1-a_2+2a_3=0\end{cases}$$
constitute an underdetermined system of 2 equations with 3 unknowns ; for this kind of system, there is a standard procedure.
Write this system as a square system of 2 equations with $2$ unknowns, byconsidering the third variable $a_3$ as a parameter, i.e. write it under the form
$$\begin{cases}(1) &:& a_1+a_2=-a_3\\(2)& :& a_1-a_2=-2a_3\end{cases}$$
The solutions to this parametric system are:
$$\begin{cases}a_1=\frac12(-3a_3)\\a_2=\frac12(a_3)\end{cases}$$
Thus the general solution is
$$\pmatrix{a_1\\a_2\\a_3}=\pmatrix{\frac12(-3a_3)\\\frac12(a_3)\\a_3}$$
which is in fact proportional to the cross product we have found in (*)