Skip to content
Start free

Find any metric, without a query language

Every metric in the time window appears as a card under its OpenTelemetry name, whether it came over OTLP, from a Prometheus scrape or from an integration. Filter, break any chart down by a dimension and open the detail, and switch to read-only SQL when you need more.

Start freeBook a demoRead the metrics docs
  • OTLP · Prometheus scrape
  • Read-only SQL
  • Monitor from any metric

How it works

From a metric name to a monitor in a few clicks

The explorer reads the metric store in your workspace's own database, and every metric keeps the name and attributes it was sent with.

  1. 01SendOTLP from SDKs and collectors, Prometheus scrapes and integrations all land in one store.
  2. 02DiscoverEvery metric that reported in the window appears as a card with its type, unit and a chart.
  3. 03FilterNarrow by text, type, service, dataset or any attribute, with negation and wildcards.
  4. 04Break downSplit every chart by a dimension, or open one metric to see its attributes and series.
  5. 05QuerySwitch to SQL mode for read-only ClickHouse queries, drafted for you if you like.
  6. 06AlertCreate a threshold monitor from any metric, or let an anomaly job learn its normal range.

Metric details

Open a metric to see what is inside it

The detail panel shows a metric's unit, temporality, datapoints, services and datasets, and every attribute it carries with how many values it has. Pick an aggregation that fits its type and break it down by any attribute.

  • Histograms offer p99, p95, p90, p50 and mean; counters offer rate and sum; gauges offer average, maximum, minimum and last
  • Each series shows its last, average and maximum value for the window
  • Create a monitor for the metric from the same panel

SQL mode

Read-only SQL when the cards are not enough

SQL mode runs ClickHouse queries over the metric store, with guardrails that make it safe to open to the whole team. Describe what you want in plain language and the assistant drafts the SQL for you to review. It never runs a query itself.

  • One statement

    A single SELECT or WITH query. Anything that writes or changes settings is refused.

  • Allowed tables

    Queries read only allow-listed tables, never system tables or remote sources.

  • Bounded

    Results stop at 2,000 rows, and a query gets 30 seconds.

  • Reviewed drafts

    Drafting uses the AI model you configure, and every draft is written to the audit log.

Sources

OTLP and Prometheus, in one store

Services and collectors send OTLP to your workspace, and a collector can scrape any Prometheus /metrics endpoint on the interval you set. Integrations from the catalog add database, Kubernetes, cloud and network metrics to the same store.

  • A Prometheus scrape takes targets, a path, a scheme, a timeout and an optional bearer token
  • Metrics keep their OpenTelemetry names and attributes, such as http.server.request.duration
  • The same metrics feed dashboards, monitors and anomaly jobs

Better together

Every metric, ready to watch

The metric store is the one your monitors, anomaly jobs and infrastructure views read, so a card you find here can become an alert in the same workspace.

Aggregations

Each metric type gets the aggregations that fit it

The explorer reads the OpenTelemetry type of every metric and offers only the aggregations that make sense for it. The first one is the default.

Metric typeAggregationsDefault
GaugeAverage, maximum, minimum, lastAverage
Counter (monotonic sum)Rate, sumRate
Up-down counter (non-monotonic sum)Average, maximum, minimum, lastAverage
Histogramp99, p95, p90, p50, meanp99
Exponential histogram and summaryMeanMean

Bucketing is explicit and in UTC. SQL mode reads the same tables, so any aggregation ClickHouse supports is a query away.

FAQ

Questions teams ask before they move their metrics

Talk to an engineer
Does Tracix support Prometheus's query language?
No. Explore mode needs no query language at all, and SQL mode runs read-only ClickHouse SQL over the same metrics.
Can a Prometheus server push metrics to Tracix?
No. A collector scrapes Prometheus /metrics endpoints itself; Tracix does not accept metrics pushed from a Prometheus server. Point a collector at the same endpoints instead.
How do metrics get into Tracix?
Over OTLP from SDKs and collectors, from Prometheus endpoints a collector scrapes, and from the integrations you turn on. They all land in one store under their OpenTelemetry names.
Is SQL mode safe to open to everyone?
It runs one read-only query at a time over allow-listed tables, stops at 2,000 rows or 30 seconds, and refuses anything that writes. The assistant only drafts queries for you to review.

See your first metrics in minutes

Point an OpenTelemetry SDK or collector at your workspace, or add a Prometheus target to a collector policy, and the cards appear as metrics arrive.

Start freeRead the metrics guide