Smoothing out RRDs because of errors?

I have been monitoring some network devices over a congested link. So lately I have been gettings lots of false positive alarms. Also I see traffic graphs have become messed up, with high peaks (similar to 32bit counter error).

Is there a way you can squash these peaks in Librenms rrd files?
example :point_up_2:t4:

Not sure what I want to replace the peaks with, but essentially I am wondering if there are any best practices to “fix” these issues. Its hard looking at the “real” data when you have peak errors like this.

Hi there, I have a script that i run that takes care of the spikes , however the bigger question is why is it happening. Its happening becuase the pollers cannot keep up or they over lap or miss. How many pollers do you have and what are they running for minions?