Taffrail: Kubernetes Metrics Aggregation for Python

Taffrail is a python package that automatically discovers metrics providers in a Kubernetes cluster and exposes them via a simple interface. You can easily add custom metrics providers as well as override existing metrics endpoints.

Installation

Install taffrail using pip:

$ pip install taffrail

Usage

Create the client:

from kubernetes import client, config
from taffrail.taffrail_client import MetricsClient

config.load_kube_config()

metrics_client = MetricsClient(client)

Get all available metrics:

response = metrics_client.get_metrics()

Get all available sources:

response = metrics_client.get_sources()

Get metrics for a specific source:

response = metrics_client.get_metrics_with_source('prometheus')

Custom kube-state-metrics endpoint

By default, taffrail will look for kube-state metrics endpoint at: /api/v1/namespaces/kube-system/services/kube-state-metrics:http-metrics/proxy/metrics. To override with a custom endpoint simply set the KUBE-STATE-METRICS-PATH env var.

Example:

$ export KUBE-STATE-METRICS-PATH="/api/v1/namespaces/default/services/kube-state-metrics:8080/proxy/metrics"

Available Sources

Currently, taffrail supports the following metrics providers:

• prometheus
• kube-state-metrics
• heapster
• metrics-server

Running inside Kubernetes

See Taffrail Server