scripts

Diagnostic scripts (read-only)

These scripts collect a snapshot of cluster state to help you triage an incident. They are strictly read-only.

🔒 Safety contract: every script here runs only non-mutating kubectl verbs (get, describe, logs, top, version, api-resources). None of them create, patch, delete, drain, cordon, uncordon, scale, apply, or taint anything. The shared helper (lib-common.sh) actively refuses to run a mutating verb.

Requirements

• kubectl configured for the target cluster (set KUBECTL=/path/to/kubectl to override the binary)
• bash 4+
• Optional: metrics-server for kubectl top output

Scripts

Script	Purpose
collect-cluster-diagnostics.sh	Broad cluster snapshot: versions, nodes, events, component health, resource usage
export-events.sh	Export sorted events (optionally warnings-only)
gather-pod-logs.sh	Logs (incl. --previous) + describe for pods in a namespace
gather-node-info.sh	Node conditions, capacity, taints, and scheduled pods
collect-ingress-info.sh	Ingress, Services, Endpoints, and controller logs
storage-diagnostics.sh	StorageClasses, PVs, PVCs, VolumeAttachments, CSI drivers
namespace-diagnostics.sh	Everything in one namespace: workloads, events, quotas, policies

Examples

# Broad snapshot into a timestamped directory
./scripts/collect-cluster-diagnostics.sh

# Focus on one namespace
./scripts/namespace-diagnostics.sh -n payments

# Pod logs (including crashed containers' previous logs)
./scripts/gather-pod-logs.sh -n payments -l app=checkout

# Only warning events, written to a file
./scripts/export-events.sh --warnings-only -o warnings.txt

A note on sensitive data

Output may contain pod names, node names, IPs, and config. The scripts never print Secret values (only names), but review any bundle before sharing it outside your organization.