We consider scheduling of colored packets with transition costs which form a general metric space. We design a $1 - O(\sqrt{MST(G) / L})$ competitive algorithm. Our main result is a hardness result of $1 - Ω(\sqrt{MST(G) / L})$ which matches the competitive ratio of the algorithm for each metric space separately. In particular, we improve the hardness result of Azar at el. 2009 for uniform metric spaces. We also extend our result to weighted directed graphs which obey the triangular inequality and show a $1 - O(\sqrt{TSP(G) / L})$ competitive algorithm and a nearly-matching hardness result. In proving our hardness results we use some interesting non-standard embedding.