Thanks for the info, the current plugin is only using vmstat so no per node/zone but it seems like it would make sense to have a plugin that parses `/proc/zoneinfo`. If you are interested in this you could open a [feature request] on GitHub.
The things are a bit more complex. While the information in */proc/zoneinfo* gives a detailed breakdown, the goal here is to be able to have a monitoring option to capture memory fragmentation at an early stage. I am sure, your are very well familiar with the [proc filesystem](https://www.kernel.org/doc/Documentation/filesystems/proc.txt) - I refer to section 1.2.
*/proc/buddyinfo* gives an overview, but the actual details, if there are enough available pages to serve high block requests, is in */proc/pagetypeinfo*.
So for what you want to capture you would mostly be looking at pagetypeinfo, or do you need to cross reference several files?
Either way it may make sense to collect this, at least initially, with the `exec` input and a custom script. The formatting of these files would require a custom parser and I think this is probably a fair bit more detailed than most others are monitoring.