Metrics Overview

kpod-metrics exports 50+ metrics across 8 categories. All metrics are labeled with namespace, pod, container, and node.

Metric Categories

Category	Source	Key Metrics
CPU	eBPF	Run queue latency, context switches
Network	eBPF	TCP bytes, retransmits, drops, RTT, connections
Memory	eBPF + Cgroup	OOM kills, page faults, cache stats, cgroup usage
Syscalls	eBPF	Per-syscall count, errors, latency
Disk I/O	eBPF + Cgroup	Block I/O latency, read/write bytes
L7 Protocols	eBPF	HTTP, DNS, Redis, MySQL, Kafka, MongoDB
Interrupts	eBPF	Hardware/software IRQ latency
Process	eBPF	Exec, fork, exit events

Metric Naming

All metrics use the kpod. prefix (exported as kpod_ in Prometheus format):

• kpod.cpu.* — CPU scheduling metrics
• kpod.net.* — Network metrics (TCP + interface)
• kpod.mem.* — Memory metrics (OOM, page faults, cache, cgroup)
• kpod.syscall.* — Syscall metrics
• kpod.disk.* — Disk I/O metrics
• kpod.fs.* — Filesystem metrics
• kpod.irq.* — Interrupt metrics
• kpod.proc.* — Process lifecycle metrics

Labels

Every metric includes these standard labels:

Label	Description
namespace	Kubernetes namespace
pod	Pod name
container	Container name
node	Node name

Some metrics include additional labels (e.g., device, interface, syscall, mountpoint). See individual metric pages for details.