Metrics Analysis

This tool can be used to analyze Prometheus metrics endpoints for high cardinality. If you want to contribute your own changes please take a look at the CONTRIBUTING.md file.

Usage

Get a sample of all metrics data from the current cluster and namespace

This will store the metrics outputs in the directory metrics-<current-context>

./get-all-pods-metrics.sh

You can use the skip_until_prefix variable to skip reading already read metrics in case of an error (e.g. a pod disappears while reading)

Build the tool

make

Usage

Use ./metrics-analysis -h to see the help.

List of Metrics (with summary)

./metrics-analysis -f metrics-.../podname.metrics -n

Output format: <name> <label_count> (value_count, ...)

Show specific metric

./metrics-analysis -f metrics-.../podname.metrics -m response_latency_ms

Summarize everything

./metrics-analysis -f metrics-.../podname.metrics