Skip to content
GreenKube

Smart Recommendations

GreenKube’s recommendation engine analyzes your cluster metrics to produce actionable optimization suggestions that reduce both costs and carbon emissions.

How It Works

Section titled “How It Works”

The recommendation engine runs a multi-pass analysis:

Pass 1: Resource Utilization Analysis

Section titled “Pass 1: Resource Utilization Analysis”
  • Compare actual usage vs. requests/limits over a configurable window
  • Detect consistent over-provisioning (usage < 30% of request)
  • Detect under-provisioning (usage > 90% of limit)

Pass 2: Behavioral Pattern Detection

Section titled “Pass 2: Behavioral Pattern Detection”
  • Identify zombie pods — running but doing no useful work
  • Detect periodic workloads — candidates for cron jobs or autoscaling
  • Find idle namespaces — entire namespaces with minimal activity

Pass 3: Optimization Modeling

Section titled “Pass 3: Optimization Modeling”
  • Calculate optimal resource requests/limits based on P95 usage
  • Model HPA configurations for variable workloads
  • Estimate savings in $, kWh, and kgCO₂e

Pass 4: Carbon-Aware Insights

Section titled “Pass 4: Carbon-Aware Insights”
  • Suggest schedule shifts for batch workloads to low-carbon periods
  • Recommend region migration for workloads where carbon intensity varies significantly

Recommendation Types

Section titled “Recommendation Types”
TypeIconDescription
Zombie Pod🧟Pod with near-zero utilization
Rightsizing📏Over or under-provisioned resources
HPA Candidate⚖️Workload suitable for autoscaling
Carbon Scheduling🌍Shift to lower-carbon time/region
Namespace Cleanup🧹Idle or unused namespace
Storage Optimization💾Orphaned PVCs or over-sized volumes

Confidence Levels

Section titled “Confidence Levels”

Each recommendation includes a confidence score:

  • High — Strong signal, >90% confidence in estimated savings
  • Medium — Good signal, 60-90% confidence
  • Low — Weak signal, worth investigating but may be a false positive

Accessing Recommendations

Section titled “Accessing Recommendations”

Dashboard

Section titled “Dashboard”

The Recommendations tab shows all active suggestions with:

  • Priority ranking
  • Affected resources
  • Estimated savings
  • One-click details

CLI

Section titled “CLI”
Terminal window
# View all recommendations
greenkube report --recommendations


# Filter by type
greenkube report --recommendations --type=zombie


# Export as JSON
greenkube report --recommendations --format=json

API

Section titled “API”
Terminal window
# All recommendations
GET /api/v1/recommendations


# Filtered
GET /api/v1/recommendations?type=rightsizing&severity=critical

History Tracking

Section titled “History Tracking”

Recommendations are tracked over time so you can:

  • See when a recommendation first appeared
  • Track whether it was resolved
  • Measure the impact of optimizations you’ve applied
Section titled “Related”