upgrade cluster from 5.6 to 5.7, problem with index?

We have 3 nodes cluster (5.6) and recently we’ve tried to upgrade cluster to 5.7 version. It’s been used below manual:

After upgrade of one node we’ve executed some query on newly upgraded node and time of execution was 100x longer than on older version. We observed strange situation that it is depend on DESC mode, if we use ASC it is ok.
Have anyone similar problem?