A small exercise in practical problem-solving: intermittent website errors prompted an investigation of PHP workloads, scheduled jobs and hosting limits. The result is a lightweight shell monitor and a PHP dashboard that makes the resulting measurements visible.
Scheduled shell-script sampling, structured log parsing, basic data aggregation, PHP 5.5 compatibility, browser-based plotting and separation of public and private diagnostics. The collector doesn't execute the monitored analysis script or inspect the contents of its PSK archive.
This is a deliberately limited public view: statistics are averaged by hour and delayed by at least one hour. Internal file paths, commands, process details and credentials are never published.
22 hourly data points from 128 samples. Last published hour: 16 Oct 2026 21:00 UTC.
Whole shared server; not an account-specific CPU measurement.
Instantaneous process samples averaged hourly.
Only file metadata is collected.
The investigation identified duplicate scheduled requests and a historical-data analysis that could scan a growing archive multiple times. We reduced redundant scheduling and added observation rather than immediately rewriting live scientific calculations. The monitoring does not establish the precise cause of a 502 error, nor does it measure CloudLinux CPU or I/O limit faults.