see
And your problem status is DUAL_INFASIBLE, maybe searching “DUAL_INFASIBLE” on this forum helps you.
By the way, it seems that many of inseasible/failed problems were caused by the bad scaling of data, which means the data are too big or too small (close to zero but not zero). It would be good if CVX can track the bad data back to the original constraints that produced them, and tell us which of them caused that, Because the modeling process can be too complicated (for me) to track by hand. I don’t know if this is possible. Please forgive my lack of background.