Skip to main content

Infrastructure Placement & Capacity

The Infrastructure dashboard (/infrastructure) gives platform engineering and SRE teams complete visibility into physical and virtual compute capacity, node placement density, and workload bin-packing efficiency across all connected Kubernetes clusters and bare-metal nodes.

Infrastructure Capacity Dashboard​

The infrastructure dashboard aggregates real-time telemetry from the Nife Cluster Agent, visualizing CPU, Memory, Disk, and Pod utilization against maximum allocatable limits.

Nife Infrastructure Placement Dashboard

Core Infrastructure Metrics​

The dashboard monitors key operational thresholds:

1. Compute Allocation vs. Usage​

  • CPU Allocatable vs. Requested: Compares total physical vCPUs reserved by Kubernetes pod resource requests against actual real-time CPU core usage.
  • Memory Saturation: Tracks Resident Set Size (RSS) and cache memory consumption across worker nodes to prevent Out-Of-Memory (OOM) pod terminations.
  • Node Allocation Density: Identifies nodes running near capacity or those with low utilization suitable for consolidation.

2. Regional & Zone Distribution​

  • Visualizes cluster node topology across availability zones (e.g., us-east-1a, us-east-1b, eu-west-1a).
  • Identifies single points of failure in multi-zone high-availability (HA) configurations.

3. Workload Bin-Packing & Optimization​

  • Idle Headroom Alerts: Flags underutilized worker instances where node autoscaling or pod rescheduling can reduce cloud infrastructure spend.
  • Node Taints & Tolerations: Inspects GPU, spot instance, and dedicated workload node placement constraints.

Managing Node Pools & Placement Rules​

From the Infrastructure view, administrators can:

  • Drain & Cordon Nodes: Safely evict running pods to perform kernel upgrades or maintenance without application downtime.
  • Inspect Pod Allocation: Drill down into individual worker nodes to view active containers, namespace allocations, and restart counts.
  • Trigger Cluster Autoscaling: Adjust minimum and maximum node pool boundaries based on current placement demand.