Telemetry

Every field the outbound streams can carry, named. Each row was read from the source that emits it.

Property	Value
Engine version	1.1.74
Verified against	the shipped 1.1.74 release and images
Verified	2026-09-02
Machine-readable	JSON

Labels

These labels ride on every engine telemetry series.

Label	Type	Example	Source	PII
__name__	string	tenx_pipeline_up	Prometheus series name	No
tenx_reported_name	string	edge-receiver	Pipeline display name from the config	No
tenx_pipeline_uuid	string	a1b2c3d4-e5f6-...	Generated per pipeline run	No
tenx_host_name	string	edge-node-1	The host the engine runs on, or Unknown when the lookup fails	Hostname
tenx_parent_uuid	string	9f8e7d6c-...	Parent pipeline id, present only when one is set	No

Three labels appear on specific series rather than on all of them.

Label	Type	Example	Appears on	PII
tenx_unit_name	string	receive	tenx_launch_failure_total	No
tenx_pipeline_factory	string	run	tenx_launch_failure_total, when a unit is missing from the factory	No
tenx_failure_reason	string	missing unit	tenx_launch_failure_total	Review

Two of those need a word more. tenx_failure_reason is missing unit or unit init error for the two classified failures, and otherwise the launch exception's message, which can name a config path or a unit, which is why its PII column reads Review. It still carries no log content, because launch fails before any event is read.

runtimeAttributes is the one field a deployment fills itself: name:value pairs supplied at launch, each name prefixed with tenx_ if it is not already, and the contents sent as given.

Metrics

The series names below are the wire names, after the suffix that micrometer's Prometheus renderer adds: _seconds on timers, _total on counters.

Metric	Type	Value	Measures
tenx_pipeline_up	gauge	1	Pipeline is running. Emitted on every send, including the empty keepalive
tenx_pipeline_bootstrap_time_seconds	timer sum	seconds	Time from launcher construction to the pipeline starting
tenx_pipeline_input_time_seconds	timer sum	seconds	Time the input stage ran
tenx_pipeline_runtime_seconds	timer sum	seconds	Total pipeline runtime
tenx_pipeline_invocation_time_seconds	timer sum	seconds	Runtime of one invocation, for invocation-shaped launchers such as Lambda
tenx_launch_failure_total	counter	count	A pipeline failed to launch

Every value is a number: a count or a duration. No metric value carries a string, so no metric value can carry log content.

Pipeline metrics

The series above are the engine reporting on itself. Per-pattern metrics (all_events_summaryVolume_total, all_events_summaryBytes_total, emitted_events_summaryBytes_total and the rest) are a separate stream that goes wherever the pipeline's metric output points: Prometheus, Datadog, CloudWatch, Elastic, or the log10x hosted backend.

The shipped configs for the Receiver, the Reporter and the Retriever index and stream apps include the log10x metric output, so on those configs this stream shares the single backendEndpoint switch with the two calls above: no endpoint means no destination, and an endpoint sends both. Metric outputs are additive, so adding one of your own sends this stream there as well, until run/output/metric/log10x is commented out.

Per-pattern metrics are labelled with the enrichment fields the pipeline configures, which typically include a recurring pattern name, severity, Kubernetes namespace and container, and HTTP status. The pattern name is a template derived from log statement structure, with placeholders where the variable data was, so it carries the shape of a log line and none of its values.