Monitoring

A live introspection surface over the running JVM. Use it when you need to see CPU load, heap pressure, GC behaviour, thread state, and the underlying connection-pool counters for every data source the platform has registered.

The data is sampled from the standard java.lang.management.* MXBeans on each request - no agent, no JMX socket. CPU-load accessors come from com.sun.management.OperatingSystemMXBean via reflection so the module stays free of Sun-internal compile-time dependencies. HikariCP pool stats come from the pool's own HikariPoolMXBean.

Layout

Region	View	Purpose
Left	Metrics	Navigator - jumps between the centre views and shows the platform-wide counts (data-source pools, registered data sources).
Centre	Monitoring	JVM dashboard tiles: CPU, heap, non-heap, per-pool memory, GC, thread totals. Auto-refreshes.
Centre	Threads	Filterable list of every live thread - id, state, daemon, priority, CPU / user time, lock owners.

Access

• Perspective id: monitoring. Display order 1100 (after Operations / before Settings).
• All three views are lazyLoad: true. The Metrics navigator loads first; the centre views activate on tab click.
• Backend REST endpoints are mounted at /services/ide/monitoring/:

Endpoint	Returns
GET /services/ide/monitoring/metrics	Point-in-time MonitoringSnapshot (runtime / CPU / memory / threads / GC).
GET /services/ide/monitoring/threads	Detailed ThreadDetail[] - omitted from the snapshot to keep dashboard polls small.
GET /services/ide/monitoring/counts	Named counts: data sources, Hikari pool sizes per source.

All three are gated by @RolesAllowed({"ADMINISTRATOR","DEVELOPER","OPERATOR"}). Anonymous users see nothing.

When to use what

• A spike in p99 latency you want to diagnose - open Monitoring, scan heap and GC tiles, switch to Threads to find the blocker.
• An OutOfMemoryError - check the per-pool MemoryPoolInfo tiles (Eden, Old Gen, Metaspace) on the Monitoring tab. The Metrics navigator also surfaces Hikari pool exhaustion as a count.
• "Why is this connection-pool reporting zero idle?" - the Metrics navigator's count groups list every DirigibleDataSource; for HikariCP-backed pools the total / active / idle / threads-awaiting counts come straight off HikariPoolMXBean.

Notes

• The Monitoring perspective and the Messaging perspective are sibling devops surfaces - Monitoring covers the JVM, Messaging covers the embedded ActiveMQ broker.
• The platform-wide Tracing perspective is the OpenTelemetry-backed companion for distributed traces; use it together with this perspective when an externally observable slowdown maps back to an internal hotspot.