Limitations & Scope

This page documents the intentional boundaries of the current public agent bundle.

Discovery Agent limits

• Node pages show capacity and allocatable snapshots, not live usage sampling.
• Provider detection is best effort and depends on cloud metadata access or useful node labels.
• CronJob execution history depends on child Jobs remaining visible and not being excluded by filters.
• Failed execution logs are stored only as a tail of up to 200 lines, not as full pod log archives.
• Namespaces and resources excluded through ignore rules are completely omitted from discovery and related history.
• Discovery helps explain runtime state in Moonin, but it is not a full observability pipeline or a replacement for dedicated metrics, traces or long-term log retention tools.

Scaling Rules Agent limits

• The current implementation targets Deployments through HPAs.
• Scaling windows are driven by template state from Moonin plus the agent’s local time-based evaluation at reconcile time.
• Reverts happen on the next reconcile cycle, so disablement or expiration is not instant to the millisecond.
• If a target Deployment has no HPA, the agent may create a provisional managed HPA for the execution window.
• Rollback quality depends on the original baseline captured before Moonin took control of the HPA.
• External HPA or replica changes performed after that baseline is captured can affect the eventual rollback target.
• Only one managed template ownership context can control a given HPA at a time.

Bundle boundaries

• The bundle does not proxy end-user requests into the cluster.
• The bundle does not create a durable local data store.
• The public documentation explains supported behavior, not internal commercial heuristics of the wider Moonin platform.