Catching Kubernetes Cloud Cost Overruns Early with OpenCost and Prometheus Alerts

·2 views

A monthly cloud invoice landing 40% over forecast often reveals only that spending was too high, not which team or workload caused it. Cloud billing consoles don't map to Kubernetes namespaces, and billing data typically arrives 6 to 24 hours late, so runaway workloads can consume days of budget unnoticed. The article proposes deploying OpenCost, a CNCF project, which estimates per-namespace costs from resource allocations and exports them as Prometheus metrics at scrape-interval resolution. It then describes building recording rules to convert allocations into hourly dollar figures and setting PrometheusRule alerts to fire during the month. On OpenShift, the walkthrough notes that user workload monitoring must be enabled so custom ServiceMonitor and PrometheusRule objects are picked up.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Half the JavaScript on a static FAQ page never runs, and much of it is framework server code

A developer found that a prerendered FAQ page, which has no interactive elements, loads 19 JavaScript chunks, and about half of their code never executes in the browser. Using V8 precise coverage via the Chrome DevTools Protocol, they measured roughly 49 percent execution across repeated runs, with timing variations from idle work and router prefetches. Legal pages served as a control and showed a higher proportion of executed code because they contained mostly framework code. Two chunks containing about 98 KB of framework server-side rendering machinery, which the browser never uses, were identified by their string tables. The author attributes this to a single bundle graph where server and client code share modules, rather than to any bug in the app's own code.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Guide to tracking a Xianyu seller's shop and prices with a scraper

Xianyu, known in English as Goofish, is China's largest second-hand marketplace, where resellers and collectors often spot good deals early. Monitoring a busy seller by hand is difficult because the site is in Chinese, designed for the mobile app, and its inventory changes daily. The guide explains how to convert a seller's shop into a structured table and receive a daily list covering only new listings, price or title changes, and removed items. Scheduled runs with a named watch list return the full shop on the first pass and only changed rows afterward, with billing based on the number of rows stored. The scraper works from seller or item links rather than keyword search, since Xianyu limits search results to logged-in users.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Turns AI Editor's Wait Time Into a $0.99 Monthly Add-On

The creator of an AI-assisted code editor found that users were frustrated by the delay while the model warmed up. A free pre-warm feature proved costly in API credits, so the developer shifted to a paid fast-track option. The add-on costs $0.99 per month and uses a higher-tier endpoint, with a feature flag and Stripe-based checkout added in under 300 lines of code. After launch, the conversion rate reached 4.2%, churn after the trial was 1.8%, and the add-on generated about $1,200 in monthly recurring revenue.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Hidden-text attack on a real LLM agent: what the test revealed

A developer writing on Dev.to describes an experiment in which hidden text was planted to attack a live LLM agent. The piece reports how the agent behaved under this manipulation and was shared alongside free Kaggle notebooks for reproducing the test. The article is framed as a security-oriented look at how agents can be misled by content they process. The version circulated here is a blog repost from ShVaSa Lab that adds generic commentary on latency, hardware constraints and deterministic validation. That added material includes a sample code snippet and a latency claim of up to 45% lower operational overhead, but these are the lab's own assertions rather than findings from the original experiment.

0 comments Read more at DEV Community