Telemetry Poller and Feedback Loops

The telemetry_poller.go operates as a continuous background service (the "Fortress Telemetry Poller"), maintaining a global, real-time state of the dLAN infrastructure.

Polling Mechanisms

The poller executes at regular intervals (typically 5 minutes) to gather metrics from disparate infrastructure endpoints.

Data Sources

Proxmox: Queries the hypervisor for container counts, CPU/RAM usage, ZFS pool metrics, and host uptime.
Pelican Panel: Queries game server management for total node counts and online server status.
cPanel / WHM: Tracks offline nodes, load averages, and determines the overall cluster health status (e.g., ALL_CLEAR vs DEGRADED).
Virtualizor: Monitors running VM statistics for VPS hosting environments.
Financial (CoinStats): An hourly check of the Crypto Fear & Greed index and BTC dominance.

Infrastructure Feedback Loop

The gathered telemetry isn't just logged; it is injected into the active memory context of relevant agents (like Sasha or Sysadmin).

When an agent is queried about system health or receives a NOC alert, the latest telemetry snapshot provides immediate, real-world context, preventing the agent from needing to perform time-consuming manual checks before formulating an initial hypothesis.

Mermaid Diagram: Telemetry Loop

graph LR
    subgraph "Infrastructure"
        P[Proxmox]
        C[cPanel]
        V[Virtualizor]
    end
    
    P --> T[Telemetry Poller (5m)]
    C --> T
    V --> T
    
    T --> M[(Global Context Cache)]
    
    M --> A[Agent: Sasha/Sysadmin]
    A --> D[Diagnose Issues]

Guidance for AI Agents