The new 1.5 version of p4pcm.pl supports percentage thresholds. Default settings are 12% for low and 20% for high. The 12% allows a chance for the script to run before the disk space gets into the 10% or less range which seems to be where monitoring alerts typically come into play.

These coded defaults are based on observations of a production proxy for a customer doing unreal development. The suggested 3% low 8% high in the job description is unlikely to allow for enough "play" in the typical unreal world to detect before exhaustion. With unreal, hundreds of cached files can change in a short period of time which quickly fills the cache. Of course a very large proxy cache is an alternative. Regardless of the coded defaults, with a roughly 300G cache, p4pcm runs every 2 hours with a 20% low and 35% high. With this 2 hour cycle, low trigger is often between 16% and 18%. The original once a day script run (the current crontab default) was often finding 7% to 8% low trigger and failed once with exhausted disk space. Current steady state is a need for cache cleans about once every 10 days. On average, the last delete is a file last accessed roughly 15 days ago. Just some additional input on setting defaults.