Let $G$ be a finite group, $\mathcal{O}$ a complete discrete valuation ring, and $F$ the residue field. The Brauer construction for a fintely-generated $\mathcal{O}G$-module $M$ with respect to a $p$-subgroup $P$ is defined to be the $F[N_G(P)/P]$-module
$$
M(P):= M^P/\left( J(\mathcal{O}M^P + \sum_{Q<P}\mathrm{tr}_Q^P(M^Q)\right)
$$

where $tr_Q^P:M^Q \to M^P$ is the relative trace map. The Brauer construction is an additive functor which fits into the following diagram which commutes up to natural isomorphism:

[![enter image description here][1]][1]


  [1]: https://i.sstatic.net/fswXa.png
where $\mathbf{triv}$ is the category of $p$-permutation modules. 

I remember reading somewhere that the Brauer construction "does not work" over $\mathcal{O}$. I think they meant that there is no functor making the diagram commute if you replace the $F$ by $\mathcal{O}$. Does anyone know what goes wrong?