Cached subsystem metrics for the tm_* Prometheus/OpenMetrics families, mirroring the tm.
More...
#include <prom/prom.hpp>#include <memory>
Go to the source code of this file.
Namespaces | |
| namespace | threadman |
| namespace | threadman::metrics |
Functions | |
| std::shared_ptr< prom::Scope > & | threadman::metrics::scope () |
The shared threadman scope; applies the tm_ prefix to every metric. | |
| prom::Counter & | threadman::metrics::pool_tasks_submitted () |
| Tasks submitted to a thread pool (all submit variants). | |
| prom::Counter & | threadman::metrics::pool_tasks_completed () |
| Tasks that ran to completion on a pool worker. | |
| prom::Counter & | threadman::metrics::pool_tasks_failed () |
| Tasks whose body threw an exception. | |
| prom::Counter & | threadman::metrics::pool_tasks_cancelled () |
Tasks cancelled before running (drained by shutdown_now). | |
| prom::Counter & | threadman::metrics::pool_queue_rejected () |
| Submissions rejected because a bounded queue was full. | |
| prom::Counter & | threadman::metrics::pool_submit_rejected () |
| Submissions rejected because the pool was shutting down or terminated. | |
| prom::Counter & | threadman::metrics::pool_scale_ups () |
| Non-core workers spawned by a scale-up decision. | |
| prom::Counter & | threadman::metrics::pool_scale_downs () |
| Non-core workers retired after idling out. | |
| prom::Gauge & | threadman::metrics::pool_queue_depth () |
| Tasks currently waiting in a pool's queue (set from the summary tick). | |
| prom::Gauge & | threadman::metrics::pool_workers () |
| Live workers in a pool (set from the summary tick). | |
| prom::Gauge & | threadman::metrics::pool_workers_active () |
| Workers currently executing a task (set from the summary tick). | |
| prom::Gauge & | threadman::metrics::pool_workers_idle () |
| Workers currently idle on the queue (set from the summary tick). | |
| prom::Histogram & | threadman::metrics::pool_task_execution_seconds () |
| Wall-clock seconds a task body spent executing. | |
| prom::Histogram & | threadman::metrics::pool_task_queue_wait_seconds () |
| Seconds a task waited in the queue before a worker picked it up. | |
| prom::Counter & | threadman::metrics::threads_created () |
| ManagedThread bodies that entered (started running). | |
| prom::Counter & | threadman::metrics::threads_completed () |
| ManagedThread bodies that returned normally. | |
| prom::Counter & | threadman::metrics::threads_failed () |
| ManagedThread bodies that exited by throwing. | |
| prom::Gauge & | threadman::metrics::threads_live () |
| ManagedThread bodies currently running (inc on enter, dec on exit). | |
| prom::Counter & | threadman::metrics::manager_threads_registered () |
| Thread control blocks registered with the manager. | |
| prom::Counter & | threadman::metrics::manager_summaries_published () |
| Periodic ManagerSummary payloads published to listeners. | |
| prom::Gauge & | threadman::metrics::manager_pools_live () |
| Pools currently registered with the manager (set from the summary tick). | |
| prom::Counter & | threadman::metrics::tasks_stuck_detected () |
| Distinct tasks newly detected as stuck (running past the threshold). | |
| prom::Gauge & | threadman::metrics::tasks_stuck_current () |
| Tasks currently flagged as stuck (set from the housekeeper tick). | |
| prom::Counter & | threadman::metrics::futures_created () |
| Promise/Future pairs created. | |
| prom::Counter & | threadman::metrics::futures_satisfied_value () |
| Futures satisfied with a value. | |
| prom::Counter & | threadman::metrics::futures_satisfied_exception () |
| Futures satisfied with an exception. | |
| prom::Counter & | threadman::metrics::future_continuations_registered () |
.then/.on_error continuations registered against a future. | |
| prom::Counter & | threadman::metrics::future_continuations_dispatched () |
| Continuations dispatched through an executor. | |
| void | threadman::metrics::warm_up () |
| Eagerly construct the scope and every metric handle (and, transitively, prom's process-wide adapter cell / NullAdapter). | |
Detailed Description
Cached subsystem metrics for the tm_* Prometheus/OpenMetrics families, mirroring the tm.
* logging channels in <threadman/log.hpp>.
Each accessor returns a long-lived metric handle created on first call via a shared prom::Scope named "threadman" (metric prefix "tm_"). The handle is cached in a static-local, so subsequent calls are a no-op load. Per-pool series are selected at the call site with .labels(prom::Labels{{"pool", opts_.name}}); the only dynamic labels are pool (operator-bounded pool name) and pool_id (numeric) — never task ids, thread ids, or exception text, which would be unbounded cardinality.
Metrics are always-on and safe by default: until a host application installs a backend via prom::Registry::global()->set_adapter(...), every metric op is a noexcept no-op through prom's NullAdapter — the same safety profile as a logger with no sinks. This header never installs an adapter; it just declares the scope and its metrics.
Histograms observe seconds (convert nanosecond durations via /1e9).
