So the way they want it is the od of the bottom cylinder is the datum feature. Not the groove or top od. The axis of only that cylinder is datum a. They do not want the whole axis of the part as it will not be concentric.
if there are other features for that snap ring groove (im assuming) then you would include those features also to associate it for datum a. (construct it as a cylinder)
It's for an o ring. There is actually 4 of them down the part. They only want the axis of the od cylinder as the datum. The part is like 14 inches long.
They want a od cylinder and a construction line though the axis of only that cylinder. They do not want the axis of the part just of thay od cylinder.
It's both. On its own, your Datum would be the derived center axis of the measured OD of the groove. Because this is also a feature of size, material boundary comes into play when this datum is used in a DRF. That's when things become interesting. For ASME Y14.5 (not familiar with ISO) the datum becomes a cylinder of the minimum circumscribed measurement of said OD. It will then derive the axis from that measurement option. If this datum was called out at MMB, then you would get more relief (bonus) as the OD got smaller.
If the majority of this part is turned, then it makes a bit more sense to have it established as the primary datum. I would imagine there are also implied or hard callouts for concentricity or position to this feature.