A cylinder $S^1\times \mathbb{R}$ will do the job !
In general, isometries of a product Riemannian manifolds are not the "product of isometries of the factors", but in this simple case I hope we can agree it's true: an isometry of the cylinder is of the type $$(x,y)\mapsto (r(x),\pm y+a),$$ where $r\in O(2)$ and $a\in \mathbb{R}$. If you want a rigorous proof of this fact, check this answer. Clearly the isometries of the cylinder map tangent vectors like $(v,0)$ in tangent vectors of the same kind (the last coordinate stays equal to $0$), so the cylinder is not isotropic, but it's homogeneous.
I want to add an interesting curiosity. Isotropic connected riemannian manifolds are completely classified up to isometry and are:
- Euclidean spaces $\mathbb{E}^n$,
- Spheres $\mathbb{S}^n$,
- Hyperbolic (real) spaces $\mathbb{H}^n$,
- Hyperbolic complex spaces $\mathbb{H}^n\mathbb{C}$,
- Hyperbolic quaternionic spaces $\mathbb{H}^n\mathbb{H}$,
- Hyperbolic octonionic plane $\mathbb{H}^2\mathbb{O}$,
- Projective real spaces $\mathbb{P}^n\mathbb{R}$,
- Projective complex spaces $\mathbb{P}^n\mathbb{C}$,
- Projective quaternionic spaces $\mathbb{P}^n\mathbb{H}$,
- Projective octonionic plane $\mathbb{P}^2\mathbb{O}$.
While homogeneous Riemannian manifolds are much more! They are precisely the manifolds $G/H$ where $G$ is a Lie group and $H\leq G$ is a compact subgroup (basically a complete classification is near impossible).