The mount.lustre mount process verified that the block device scheduler is using one of the acceptable scheduler types, since using an incorrect scheduler can cause significant performance loss. The currently-allowed schedulers are "noop" and "deadline". With newer kernels, the block multi-queue devices have a new "mq-deadline" scheduler that can also be used, otherwise a message like the following is printed:

onyx-32vm7: mount.lustre: change scheduler of /sys/block/vda/queue/scheduler from mq-deadline to deadline

This isn't strictly harmful (I don't think it will actually accept "deadline" as a valid scheduler in this case), but it is confusing and should be fixed.