ed-metrics

name	ed-metrics
description	Metrics - discover metric names and run aggregation queries (timeseries and tables).
metadata	{"version":"1.0.0","author":"edgedelta","repository":"https://github.com/edgedelta/agent-skills","tags":"edgedelta,metrics,timeseries,aggregation","alwaysApply":"false"}

Edge Delta Metrics

Discover and aggregate metrics.

Prerequisites

The edx CLI must be installed and authenticated. See the ed-edx skill.

Rules That Prevent Failed Queries

1. Metric names must be EXACT - no wildcards, no regex, case-sensitive. Always discover the name first.
2. No full-text search in metric filters - field:"value" syntax only.
3. Group-by keys must be real fields - check with edx facets keys --scope metric.

Discover Metric Names

edx metrics list                    # all metric names
edx metrics list --keyword cpu      # fuzzy filter

Query a Metric

# Average request duration per service, last hour
edx metrics query --name http.request.duration --agg avg --group-by service.name

# Max CPU on one host over 24h with 5-minute rollups
edx metrics query --name system.cpu.usage --agg max \
  --filter 'host.name:"web-1"' --rollup 300 --lookback 24h

# Table instead of timeseries (current values, good for top-N)
edx metrics query --name http.requests --agg sum \
  --group-by service.name --graph-type table

Aggregations: sum, avg, min, max, count, median. Filter: CQL field syntax, "*" for none.

Under the hood this builds the CQL agg:name{filter} by {keys}.rollup(secs) and returns records keyed by formula, e.g. {"A": {"records": [...]}}.

Groupable Dimensions (Important)

Metric dimensions are a fixed indexed allowlist (service.name, host.ip, ed.*, k8s.*, ...) - discover it with edx facets keys --scope metric. Grouping by anything else (an OTLP datapoint attribute like model/type, or a custom log_to_metric field dimension) silently returns one empty group rather than an error; edx metrics query warns when a --group-by key is not indexed. For attribute-level breakdowns, query the underlying logs (which carry every attribute) or emit a dedicated log_to_metric per breakdown.

Common Investigations

Question	Command
Is CPU/memory elevated?	edx metrics query --name system.cpu.usage --agg avg --group-by host.name
Which service has the most errors?	edx metrics query --name <error-count-metric> --agg sum --group-by service.name --graph-type table
Did latency regress after deploy?	edx metrics query --name http.request.duration --agg avg --from <deploy-time> --to <now>

Empty Results Checklist

1. Verify the metric name exactly: edx metrics list --keyword <part>.
2. Verify filter values: edx facets options --scope metric --facet service.name.
3. Widen --lookback.
4. Remove --filter (use "*"), then add filters back one at a time.