Tomcat Manager status and JMX monitoring alert overlap

Setting useful alerts for Tomcat without notification noise involves an integration decision between the Manager application status endpoint and JMX monitoring. The goal is reliable alerting for operational conditions such as thread saturation and session anomalies while avoiding duplicate or noisy notifications from routine events.

The Manager application exposes status metrics like activeSessions, reloadCount and processingTime via XML/JSON. JMX monitoring exposes ThreadPool MBeans with attributes such as currentThreadsBusy and maxThreads. Both interfaces are consumed by external alerting systems and have different update semantics, security exposure for remote access, and overhead implications from polling.

Which operational conditions should be sourced from Manager status versus JMX MBeans to avoid duplicate alerts? How should alert thresholds be coordinated across the two interfaces to reduce noise from routine reloads or transient thread spikes? What integration pattern supports correlating session changes reported by Manager with thread pool saturation reported by JMX?