this is just a note about this script,
as I am working with a munin plugin to proper present ZoL stats,
I found that arc_sumary.py is original written for FreeBSD,
and some of the arcstats it use to summarize the result is not actual value on ZoL.
p is stand for target size of MRU
so it use p as MRU size ,
size-p as MFU size,
on ZoL we do have independent mru_size and mfu_size value,
on my system MRU size is far off p which result in wrong conclusion for ARC Size Breakdown,
ARC Size Breakdown
so I think this script require some review.
p is the MRU target size, not the actual size
I suggest if you are writing an agent to go directly to the arc.c source for details and get the values directly from the /proc/spl/kstat/zfs/* rather than an intermediary script such as arc_summary
that's what I am doing right now,
as arc_summary.py is included in current ZoL,
I just get some reference from it in order to find out what's matter,