Find out what's wrong with a node
NotReady, memory pressure, a flapping kubelet, pods getting OOM-killed - the signals are scattered across conditions, taints and events. Kubius gathers them into a ranked diagnostics report for each node.
Works with Google GKE, Amazon EKS, Azure AKS, DigitalOcean, self-managed & more · macOS Sonoma (14) or later
The whole picture of a struggling node
Ranked findings
A rule-based report flags the real problems - NotReady, memory/disk/PID pressure, a stale kubelet heartbeat, cordons and control-plane taints - critical issues first.
Who's eating the memory
Per-container top consumers on the node, so you can see which workload is pushing it toward pressure - not just that pressure exists.
OOM kills, attributed
Out-of-memory kills on the node are surfaced and tied to the victim pod and container, including kernel OOM events, so you know what got killed and why.
Survives node death
Kubius keeps the last per-container memory snapshot, so even after a node goes away you can still see what it looked like just before it fell over.
Node events, rolled up
NodeNotReady flapping, SystemOOM, reboots and kubelet failures are collected against the node and ordered so the trigger stands out.
Down to the pods
Jump from the node to the pods scheduled on it - Evicted pods and OOM-killed containers are highlighted so you go straight to the fallout.
Why Kubius for nodes
From "the node is unhealthy" to why
Node Diagnostics is part of Kubius Pro - included in the free trial.
- Ranked findings: Ready, pressures, taints, stale kubelet, cordon
- Per-container top memory consumers on the node
- OOM kills attributed to the victim pod/container
- Last memory snapshot preserved after the node dies
- Node events rolled up and ordered by recency
- 14-day free trial, no card required
Simple, one-time pricing
Pay once. No subscription.
One-time licence from €49 - no subscription. 14-day free trial, no card required.
Free 14-day trial · macOS Sonoma (14) or later · No agents, nothing to install in-cluster
Questions, answered
What does node diagnostics actually check?
It evaluates the node's conditions (Ready, memory/disk/PID pressure, network), a stale kubelet heartbeat, cordons and control-plane taints, pod failures on the node, and the node's events - then ranks the findings so the critical ones are first.
Can it tell me what caused an OOM kill on the node?
Yes. Kubius surfaces out-of-memory kills - including kernel OOM events - and attributes them to the victim pod and container, alongside per-container memory usage on the node.
What if the node has already gone away?
Kubius retains the last per-container memory snapshot for the node, so you can still inspect what it looked like right before it died.
Do I need Prometheus or a node agent?
No. It reads node status, events and metrics from the Kubernetes API and metrics-server - there's nothing to deploy onto the nodes.