OpsRadar

Problem: Maintaining production Kubernetes clusters requires constant manual auditing — checking resource requests/limits against actual usage, spotting misconfigured RBAC or security settings, and catching stale deployments before they cause incidents or waste spend. This work is repetitive, error-prone, and rarely done systematically because it takes hours per cluster and doesn't scale across dozens of namespaces. Engineers end up reacting to incidents instead of preventing them. Solution: K8s Ops Radar is a dashboard that connects to a real Kubernetes cluster (via kubeconfig or in-cluster deployment) and runs IBM Bob 2.0 agents in parallel — one subagent per namespace/deployment — to analyze manifests against live metrics. Each subagent checks for over/under-provisioned resources, security misconfigurations, and outdated dependencies, then reports findings back to an orchestrator. Results are visualized as a radar/heatmap showing cluster health at a glance, with a History timeline tracking how issues are resolved over time. For each finding, the dashboard can generate a ready-to-review pull request with the exact fix (e.g., corrected resource limits). Target users: Platform engineers, SREs, and DevOps teams managing production Kubernetes clusters who currently do this auditing manually or not at all. It's built to be self-hosted via Docker Compose (quick trial) or a Helm chart with least-privilege, read-only RBAC (for teams that want it running inside their own cluster permanently). What makes it unique: Most cluster-monitoring tools show metrics; they don't reason about them or propose fixes. K8s Ops Radar treats each namespace as an independent unit of work, dispatching Bob 2.0 subagents concurrently rather than running one slow sequential scan — turning what used to be a multi-hour manual audit into a live scan you watch happen in minutes, ending in an actionable PR rather than just a dashboard full of numbers. The RBAC design (read-only, scoped, no write access to secrets) makes it something a security-conscious enterprise team could actually adopt, not just a hackathon demo. How it addresses the problem: by automating the exact workflow one of our team members previously did by hand in production (auditing resource right-sizing across live Kubernetes clusters), we cut audit time from hours to minutes, remove human error from the process, and produce a concrete, reviewable artifact (a PR) instead of a report someone has to act on manually later.