The secondary would be the go. You'd want to leave the primary alone as much as possible - largely as its more of pain to get right on a big scope.
However it may not be enough just to adjust the secondary, in which case, I would "zero" both out (get them as close as possible to their original positions), lock the primary and start from there.
Also, I found that doing the collimation while the scope was at, say, a 45-60 degree angle made a difference.
I assume the weight of the primary, combined with the rather weak springs, helps to drag itself out of collimation when moving through its arc.
This is just what worked for me.
I'm a long way from being an expert at collimating newts, so anybody with more cogent advice, please sing out!
