Skip to main content

Infeasible solution to a feasible problem

Awaiting user input



  • Mario Ruthmair
    Gurobi Staff Gurobi Staff

    Hi Kaushik,

    If you want to find out why some model is infeasible, here is an article on that topic.

    However, your log output reveals some numerical issues with your model. Here are some general guidelines how to deal with numerical issues.

    Although some heuristics found feasible solutions in the beginning, the root relaxation turns out to be infeasible. This should not happen from a theoretical point of view but can happen if the numerics of your model are bad (with respect to the limited machine precision). Note also that the start solution you hand over is not considered as feasible solution to this model.

    Here are some suggestions:

    • Your matrix coefficient range goes down until 1e-7. Note that the feasibility tolerance is by default 1e-6, all bound or constraint violations below this value might be ignored by the solver. You could either decrease this value with the according parameter linked above, or better try to revise, cleanup, and scale the coefficients you are using.
    • Do you also experience this issue when using the default settings, i.e., not forcing barrier with Method=2?
    • You could put more emphasis on numerics by using parameter NumericFocus=1 (or 2 or 3).
    • Also note that for feasible problems the default number of solutions to keep in the solution pool is 10. Only the best 10 solutions found so far are kept and worse ones are discarded. You can control this with parameter PoolSolutions.

    Best regards,

  • Mario Ruthmair
    Gurobi Staff Gurobi Staff

    I just want to add that you are using an old Gurobi version. The latest one is 10.0.1 that potentially deals in a better way with numerically challenging models.


Please sign in to leave a comment.