Metrics
Metrics are the numeric signals your applications and infrastructure emit — counters, gauges, and histograms. Owlpane stores them as columnar series and gives you two ways to query them: the point-and-click Metrics explorer at /observability/explorer, and the expression language PromQL at /observability/promql. Runtime CPU, memory, heap, and event-loop metrics surface in the Profiler at /observability/profiler.
What it is
A metric is a number over time with labels attached. That shape makes metrics ideal for rates, thresholds, and aggregation — how many requests, how slow, how much memory — where logs and traces are too high-cardinality to summarise cheaply.
Owlpane keeps metrics queryable in two dialects so both audiences are served: analysts who want to pick a metric, a function, and a grouping from menus, and engineers who already think in PromQL expressions.
Key capabilities
- Browse every metric an application reported in the selected range, then chart it.
- Aggregate and group — choose an aggregation, group by a label, and filter on others. The whole view is encoded in the address bar, so you can send a chart to a colleague as a link.
- Distribution and breakdowns — see how a histogram's readings spread across its buckets with estimated percentiles, and which values a label takes on the metric.
- PromQL subset with live validation — a metric name with label matchers,
rateandincreaseover a range,sum/avg/max/min/count(withby), andhistogram_quantile. The query is checked as you type. - Runtime profiling — process CPU, memory (RSS), heap used, event-loop utilisation and lag, and garbage collection for Node APIs, plus browser page performance for front-end apps.
Where it lives
| Surface | Console route | What it does |
|---|---|---|
| Metrics explorer | /observability/explorer | Chart any metric: pick an aggregation, group by a label, and filter. |
| PromQL | /observability/promql | Write and run a PromQL subset, validated as you type. |
| Profiler | /observability/profiler | Runtime CPU, memory, heap, event loop, and GC; browser page performance. |
| GPU profiler | /observability/profiler/gpu | GPU-oriented profiling views. |
How data gets in
Metrics arrive over OTLP (/v1/metrics) as counters, gauges, and histograms, or through Prometheus remote-write-compatible pipelines via the OpenTelemetry Collector, authenticated with an owl_ing_* key.
- Metric ingestion — signal types and querying metrics in dashboards.
- OTLP ingest reference — endpoints and content types.
- Ingest keys & authentication — passing and rotating keys.
- SDKs — language packages and OpenTelemetry-native guides.
- Integration library — Prometheus, host, database, and cloud scrape packs.
- OTLP contract — metric naming and attributes.
Concepts
| Term | Meaning |
|---|---|
| Counter | A monotonic value that only increases, useful for rates. |
| Gauge | A value that can rise or fall, such as memory in use. |
| Histogram | A distribution bucketed by value, used for latency percentiles. |
| Label | A key/value dimension that splits a series, such as service.name. |
| Aggregation | A function applied over a window, such as a rate or a percentile. |
| Series | One metric plus one set of label values. |
Choosing a query surface
Use the Metrics explorer when you want to pick a metric, an aggregation, and a grouping from menus and share the result by URL. Reach for PromQL when you already have an expression in mind or need a range function such as rate or histogram_quantile. Both read the same stored series.
Sharing a view
Because the explorer encodes its metric, aggregation, grouping, filters, and range in the address bar, a copied URL reopens the same chart for a colleague. That makes an ad-hoc chart a durable link without saving a dashboard first.
Related
- Logs — the message-level companion to numeric signals.
- Traces — request-level detail behind a latency spike.
- Dashboards — put metric panels beside logs and traces.
- Alerts and on-call — threshold and anomaly rules over metrics.
- Infrastructure — host, database, and cloud metrics in context.
Last updated