liveBook · Manning

Up metric

As explained in Machine Learning Platform Engineering

The up metric is a Prometheus metric whose value indicates whether a monitored target is successfully providing scrape data.

The up metric indicates whether Prometheus can successfully scrape a monitored target: 1 means available, and 0 means down or unavailable. It provides a direct signal for alerting on BentoML deployments. section 11.1.4

Alerting on a BentoML service

Define ServiceDown with up(...) == 0 and MissingUpMetric with absent(up(...)) == 1, then load both rules through the Prometheus Helm values configuration.

How Machine Learning Platform Engineering covers it

How deep it goes
Defines binary values and alert expressions, then extends detection with absent for deleted or terminated deployments.
Examples
Demonstrates BentoML API-server and runner selectors, ServiceDown, MissingUpMetric, and five-minute alert rules.
What it assumes
Assumes Prometheus deployment and introduces alert expressions, rule groups, labels, annotations, and missing-metric detection.
What you can do after
Operate/configure Prometheus alerts that detect unavailable or missing BentoML deployment scrape data.
How it connects
Connects Prometheus, PromQL, alert rules, the absent function, Helm, and Alertmanager.
Cover of Machine Learning Platform Engineering
liveBookmanning.com
Read Machine Learning Platform Engineering on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machine Learning Platform Engineering. The section links point to the parts of the book it draws on.