Reported IO spikes like these are likely the result of the ONTAP API returning incorrect data to Harvest, which it in turn passes to Graphite. Sometimes the root cause is an ONTAP bug, and since you are running a quite old release (8.2) upgrading ONTAP might resolve. I also see you are using an older version of Harvest (1.2.2) so upgrading it is also a good idea but I doubt updating only Harvest will solve the issue.
I have a theory for what causes this issue and would like to try a potential workaround for it. If you are interested to trial this with me please send a private message.
Cheers, Chris Madden
Solution Architect - 3rd Platform - Systems Engineering NetApp EMEA (and author of Harvest)