Metrics reference

The server registers the Micrometer meters below; standard JVM and process meters are available beside them.

Exposure

Setting	Default	Effect
`management.endpoints.web.exposure.include`	`health,info,metrics`	Endpoints exposed under `/actuator`. `GET /actuator/metrics` lists meters; `GET /actuator/metrics/{name}` reads one
`management.endpoints.web.exposure.include` containing `prometheus`	off	Serves `/actuator/prometheus`; the Prometheus registry is on the `boot` class path
`management.otlp.metrics.export.enabled`	`false`	Pushes metrics over OTLP
`management.otlp.metrics.export.url` (env `MANAGEMENT_OTLP_METRICS_EXPORT_URL`)	Spring Boot default	OTLP metrics endpoint, for example `http://collector:4318/v1/metrics`
`management.tracing.export.enabled`	`false`	Exports traces; the endpoint is `management.opentelemetry.tracing.export.otlp.endpoint`

yukta.execution.duration and yukta.execution.queue.wait publish the percentiles p50, p95 and p99.

Meter	Type	Tags	Meaning
`yukta.executions.running`	gauge	none	Executions holding a slot
`yukta.executions.queued`	gauge	none	Executions waiting for a slot
`yukta.executions.rejected`	counter	`reason`	Starts refused; one series per reason below
`yukta.execution.queue.wait`	timer	none	Time an execution waited in the queue before running
`yukta.execution.duration`	timer	`status`	Run time of ended executions, by terminal status

Values of the reason tag on yukta.executions.rejected, from ExecutionScheduler.RejectReason, lower-cased.

Value	Meaning
`shutdown`	The server is shutting down
`limit`	An execution limit was reached (including the child-run limit)
`queue_full`	The start queue is full
`workflow_queue_full`	The queue of that workflow is full
`queue_timeout`	The start waited in the queue past its limit
`not_leader`	This server is not the leader
`rate_limit`	The start rate limit was exceeded
`draining`	The server is draining for a restart

Meter	Type	Tags	Meaning
`yukta.node.executions`	counter	`module`, `status`	Nodes that reached a terminal status; `module` is the plugin type
`yukta.node.duration`	timer	`module`	Node run time
`yukta.node.retries`	counter	`module`	Messages retried
`yukta.edge.dropped`	counter	`type`	Messages dropped by an edge buffer; `type` is the node’s plugin type
`yukta.plugin.events`	counter	`plugin`, `outcome`	Counts reported by plugins through `PluginMetricsReporter`; `plugin` carries the metric name (for example `mcp_tool_call`, `llm.budget.exceeded`)
`yukta.workflow.cache.size`	gauge	none	Prepared workflows held in memory

Meter	Type	Tags	Meaning
`yukta.tracker.queue.depth`	gauge	none	State events accepted and not yet applied
`yukta.tracker.dropped`	counter	`type` (`log`, `state`) or `kind` (`supervisor`, `feed`, `analytics`)	Dropped log lines, lost state events, and events a slow consumer missed
`yukta.tracker.log.append.failed`	counter	`type` (`node`, `run`)	Failed writes to the execution log
`yukta.execution.log.lost`	counter	`reason` (`queue_full`, `write_failed`)	Execution log events that never reached PostgreSQL (Plus persistence)
`yukta.checkpoint.bytes`	gauge	none	Node output held by the checkpoint store
`yukta.checkpoint.errors`	counter	none	End markers that could not be written

Meter	Type	Tags	Meaning
`yukta.workers`	gauge	none	Workers connected to the control plane
`yukta.worker.slots`	gauge	none	Free node slots across connected workers
`yukta.leader`	gauge	none	`1` on the leader, else `0`; registered when `yukta.ha.mode=postgres`

Meter	Type	Tags	Meaning
`yukta.input.requests`	counter	`kind`, `risk`, `outcome`	Input requests by kind, risk and outcome
`yukta.input.open`	gauge	none	Open input requests
`yukta.input.wait`	timer	`tier`	Time requests waited for an answer
`yukta.input.escalations`	counter	`tier`	Escalations
`yukta.input.denied`	counter	`rule`	Answers refused by an authorization rule
`yukta.input.taken-over`	counter	none	Requests taken over
`yukta.input.sweep`	timer	none	Duration of the expiry sweep
`yukta.input.token`	counter	`outcome`	Answers by token (`accepted`, `failed`, others by outcome)
`yukta.input.delivery`	counter	`responder`, `outcome`	Deliveries to responders: `sent`, `retry`, `failed`
`yukta.control.approval`	counter	`outcome`	Parked supervisor commands awaiting approval, by outcome

Meter	Type	Tags	Meaning
`yukta.supervisor.commands`	counter	`id`, `outcome`	Commands issued by each supervisor attachment (outcomes include `dry-run` and `awaiting-approval`)
`yukta.supervisor.workflow`	counter	`outcome`	Runs of workflow-engine supervisors
`yukta.supervisor.workflow.slots.used`	gauge	none	Reserved supervisor slots in use
`yukta.supervisor.restored`	gauge	`result` (`started`, `failed`)	Attachments the last restore started or could not start
`yukta.supervisor.declarations.failed`	counter	`operation`	Failed supervisor declaration operations

Meter	Type	Tags	Meaning
`yukta.mcp.call`	timer	`tool`, `outcome`	`MCP_TOOL_CALL` duration; `outcome` is `SUCCESS` or `FAILURE`
`yukta.mcp.heavy.refused`	counter (function)	none	MCP server tool calls refused by the lower limit for `start_workflow` and regex log reads
`llm.tokens`	counter	`provider`, `direction`	Model tokens used; `direction` is `input` or `output`
`llm.cost`	counter	`provider`	Model cost, when the node reports it