Launching today

Shoalcheck
Read-only Kubernetes audit engine for drains, upgrades, CI
2 followers
Read-only Kubernetes audit engine for drains, upgrades, CI
2 followers
Read-only Kubernetes audit that runs from your laptop. One reading lists reliability, security and cost findings from 89 detection rules and 15 stack analyzers, each tied to its object. Preflight answers a change before you make it: which node won't drain, what a lost zone strands, what blocks the next upgrade. CI/CD gate with SARIF, findings filed in Jira or Linear. Free preflight; CLI and desktop app for macOS, Windows and Linux.






Free Options
Launch Team

CloudflareYour next app belongs on Cloudflare Workers. Deploy for free
Promoted
Hi Product Hunt! I'm Illia, a DevOps engineer. I work with production Kubernetes clusters for a living, and the changes that hurt were never exotic: a drain that hung on a PodDisruptionBudget at zero, an upgrade that stalled because one node pool couldn't roll, a zone outage that left a StatefulSet with nowhere to go.
Shoalcheck answers those before you act:
Preflight for a drain, a lost zone or an upgrade names the object that stops the change. The exit code is the answer: 0 nothing stops it, 2 something does, 1 partial.
The full reading: 104 rules across reliability, security and cost, each finding tied to the object it rests on.
A CI/CD gate that fails the build on new high findings and judges a pull request's manifests before they land, with SARIF for GitHub code scanning.
Tickets: the findings you pick, filed in Jira or Linear.
It's read-only: get and list only, nothing installed in the cluster, and an RBAC denial is reported as "not read", never as a pass.
Preflight is free. Try it in two minutes: brew install shoalcheck/tap/shoalcheck, then shoalcheck preflight drain <node> against any context you can read. Or open the sample reading on shoalcheck.com without a cluster.
What it doesn't do: it models less than the scheduler, so it can say a change will stop, not promise one will succeed. I'd love to hear where the model is wrong, and which change bit you last.