Originally posted by: 9050_Zhihai_Zhang

Using cplex 1251, I solved a quadratic program that has a linear objective function and two conic quadratic constraints. CPLEX reported a feasible solution. After that, I checked the slack variable associated with one of the conic quadratic constraints and found that the corresponding constraint (>=0) is infeasible, that is, the value of LHS obtained manually is less than zero (slack variable >0). What's problem?

How did you check the slack, and how large was the violation?

Originally posted by: 9050_Zhihai_Zhang

I exorted the model to a .lp file and read it into the interactive optimizer of CPLEX. After solving it, I displayed the solution of qcslacks. The violation is about -0.2~-0.4.

That indeed looks odd. Can you export your model to a SAV file and post it here?

Originally posted by: 9050_Zhihai_Zhang

I have  found my error. Thanks.