Orkestra Metrics Analysis

Managing 170+ Resources Across 5 CRDs with Zero Code

Executive Summary

This document analyzes the Prometheus metrics from a running Orkestra instance managing 5 CRDs (3 built-in Kubernetes resources + 2 custom resources) with a total of 170+ live resources. The metrics demonstrate Orkestra’s ability to handle production-scale workloads with minimal resource footprint and consistent performance across both built-in and custom resource types.

Key Findings:

• 170+ resources managed across Pods (69), Secrets (70), Deployments (30), Websites (3), and OrkApps (1)
• 4,200+ reconciliations processed with 99.6% success rate (only 15 errors out of 4,215 total reconciliations)
• Consistent reconciliation latency averaging 173ms across all CRDs
• Efficient memory usage at 97MB RSS for a process managing 170+ resources
• No performance degradation between built-in resources (Pods, Secrets, Deployments) and custom CRDs

1. Environment Overview

2. Resource Distribution

┌─────────────────────────────────────────────────────────────────┐
│  Resource Count per CRD                                         │
├─────────────────────────────────────────────────────────────────┤
│  Secret           ██████████████████████████████████████████ 70 │
│  Pod              ██████████████████████████████████████████ 69 │
│  Deployment       ██████████████████ 30                         │
│  Website          ██ 3                                          │
│  OrkApp           █ 1                                           │
└─────────────────────────────────────────────────────────────────┘

Insight: Orkestra handles high-volume built-in resources (70 Secrets, 69 Pods) with the same ease as low-volume custom CRDs. The 2:1 ratio between Secrets and custom resources demonstrates balanced workload distribution.

3. Reconciliation Performance

3.1 Total Reconciliations by CRD

3.2 Reconciliation Latency (Histogram)

Key Finding: All reconciliations complete in under 5ms.

11. Conclusions

11.5 The Zero-Programming language Promise is Fulfilled

All 5 CRDs — including the built-in Pod, Secret, and Deployment watchers — were defined entirely in YAML. No code was written to:

• Create clients for built-in resources
• Write informers for Pods/Secrets/Deployments
• Implement reconciliation logic for any resource
• Configure metrics or health endpoints

Orkestra v0.1 — Declarative Operators for Kubernetes Metrics captured: March 22, 2026

• Next: RoadMap