Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.pastures.farm/llms.txt

Use this file to discover all available pages before exploring further.

When an issue diagnosed by Pastures AI matches problems on other clusters in your fleet, a fleet correlation banner appears in the response.

​
What it looks like

The banner is styled with an amber background and appears directly in the AI response: 
⚡ This issue is also active on 3 other clusters: prod-east, prod-west, staging-01
This signals that the problem isn’t isolated — it may indicate a systemic issue across your fleet, such as a CVE, a shared misconfiguration, or an upstream bug affecting multiple environments.

​
How it works

Fleet correlation is powered by the Oracle backend’s cross-cluster matching:
  1. When a diagnosis is generated, Oracle checks the query and root cause against data from all registered clusters
  2. If the same issue signature is detected on other clusters, the response includes a fleetCorrelation object
  3. The AI drawer renders this as a warning banner with the count and names of affected clusters
interface FleetCorrelation {
  count: number;       // Number of other clusters affected
  clusters: string[];  // Names of the affected clusters
}

​
Mode availability

ModeFleet correlation
OracleFully supported — cross-cluster matching from Engine data
DemoSimulated on select responses (etcd, Longhorn queries)
BYO ModelNot available — requires fleet-wide data from the Engine
Fleet correlation requires the Oracle backend to have visibility into multiple clusters. It is not available when using a BYO model, since your own endpoint has no access to fleet-wide cluster data.

​
Why it matters

Without fleet correlation, operators troubleshoot clusters in isolation. A Longhorn degradation on one cluster might be treated as a local issue, when in reality the same problem is hitting three other clusters due to a shared storage driver version. Fleet correlation surfaces these patterns automatically, enabling:
  • Faster triage — understand blast radius immediately
  • Coordinated remediation — fix all affected clusters, not just the one you noticed
  • Pattern detection — identify systemic issues before they escalate