AWS CodeBuild Project
Does ZopNight manage AWS CodeBuild Project?
CodeBuild bills per build-minute at a rate set by the compute type each project selects, so a build's cost is duration times instance size. ZopNight discovers projects on the 6-hour cycle, attributes per-project build spend from Cost Explorer or CUR 2.0, and recommends rightsizing oversized build environments.
Rules that fire on AWS CodeBuild Project
At a glance
| Field | Value |
|---|---|
| Scheduling notes | discovery, cost tracking, and recommendations only. |
AWS CodeBuild runs build and test workloads on managed compute, billed per build-minute by instance size. Oversized build environments and long-running builds inflate CI cost linearly.
Duration times compute type
CodeBuild’s meter is per build-minute, with the rate fixed by the compute type the project declares: general-purpose sizes from small through 2xlarge, plus GPU and ARM variants. Cost is therefore a clean multiplication: minutes consumed times the size chosen. Both factors drift upward independently. Compute types get bumped once during a slow-build investigation and never revisited; build durations creep as dependency trees and test suites grow. Neither shows up anywhere except the bill, because CI is judged on green-versus-red, not on minutes.
Per-project spend through ZopNight
A dedicated provider discovers build projects on the 6-hour cycle, and per-project build cost is attributed from Cost Explorer or CUR 2.0. The rightsizing recommendation compares the compute type against what the builds do: a project on a 2xlarge fleet running lint jobs is the caricature, but the honest version, medium builds that would finish acceptably on small, is common and invisible without per-project numbers. Build projects have nothing running between builds, so there is nothing to schedule; the levers are size, caching, and frequency.
CI patterns that burn minutes
Missing caches dominate: every build re-downloading dependencies or rebuilding Docker layers pays full-length minutes for work a cache would skip. Retry storms on flaky tests multiply spend by the retry count. Webhook triggers configured too broadly rebuild everything on every branch push, including branches nobody will merge. And provisioned fleets (reserved capacity for faster starts) bill continuously, converting a per-minute service back into an always-on one for projects that build twice a day.
Reading project economics natively
The CodeBuild console lists projects with their compute type and recent build history, including per-build durations. Multiplying average duration by builds per week by the compute rate gives each project’s run rate in one line, and the projects worth attention rise straight to the top.