Module runtime_metrics 

Per-runtime observability for the tokio runtimes the node runs on.

The node splits work across separate tokio runtimes (notably a node-core runtime and a client-facing serving runtime). These metrics make it possible to tell whether one runtime is starving another for worker threads: the scheduler_lag_seconds heartbeat and a non-zero global_queue_depth are the direct signals of a runtime whose workers cannot keep up with ready tasks.

Structs

RuntimeMonitorMetrics

Functions

start_runtime_monitor

Starts background monitoring of the tokio runtime identified by handle, labelling all metrics with runtime.