Metrics explorer
Chart any metric, split it by tag, host or service, and combine series.
Charting a metric #
The explorer charts one metric over a time range. Choose the metric, the aggregation and the group-by dimensions, and the chart follows.
- Metric name — what is being measured.
- Aggregation — average, sum, minimum or maximum.
- Group by — host, service, or any tag on the series.
- Time range — from the last 15 minutes to a custom window.
Aggregation semantics #
How a metric is aggregated depends on how it is reported. A metric that counts something over an interval is summed across a bucket; every other kind is averaged.
Why the difference
A gauge sampled every 10 seconds averaged over a minute gives the right value. A count incremented between reports summed over a minute gives the right value too. Averaging a count would understate it; summing a gauge would overstate it.
Metric kinds #
- Gauge — a value at a point in time, such as memory used.
- Sum — a cumulative counter.
- Count — increments over an interval.
- Histogram — a distribution, stored so percentiles can be computed.
Splitting a metric #
Group by host to see which machine is responsible; group by service to see which service is; group by a tag to find the dimension that explains the shape.
Start ungrouped. A single line that is easy to read tells you whether the metric is worth splitting at all.
Combining metrics #
The explorer can plot several metrics together, each with its own letter. This is how you put a rate next to a latency on one chart without a dashboard.
Limits #
- Up to 64 labels per series.
- Label names up to 128 characters, values up to 2,048.
- Up to 1,000,000 active series per organisation.