Skip to content

CLAUDE.md

Latest commit

 

History

History
91 lines (68 loc) · 3.65 KB

CLAUDE.md

File metadata and controls

91 lines (68 loc) · 3.65 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Repository Purpose

This is a scripts repository maintained by Mattox Engineering, containing diagnostic and troubleshooting tools for Kubernetes environments. The primary tool is a logs collector supporting various Kubernetes distributions (RKE2, RKE1, K3s, kubeadm).

Repository Structure

scripts/
├── logs-collector/
│   ├── rancher2_logs_collector.sh  # Main logs collection script (~1800 lines)
│   ├── README.md                    # Usage documentation
│   └── collection-details.md        # Details on what data is collected
└── LICENSE                          # Apache 2.0

Main Script: rancher2_logs_collector.sh

Running the Script

# Basic usage (must be run as root)
sudo bash logs-collector/rancher2_logs_collector.sh

# With custom output directory
sudo bash logs-collector/rancher2_logs_collector.sh -d /var/tmp

# Override auto-detected k8s distribution
sudo bash logs-collector/rancher2_logs_collector.sh -r rke2

# Custom RKE2 data directory
sudo bash logs-collector/rancher2_logs_collector.sh -c /opt/rke2

# Filter logs by date range
sudo bash logs-collector/rancher2_logs_collector.sh -s 7 -e 5  # relative days
sudo bash logs-collector/rancher2_logs_collector.sh -S 2024-01-01 -E 2024-01-07  # absolute dates

# Obfuscate IPs and hostnames
sudo bash logs-collector/rancher2_logs_collector.sh -o

Key Configuration Variables (top of script)

  • SYSTEM_NAMESPACES: Kubernetes namespaces to collect pod logs from
  • KUBE_CONTAINERS: Container names to collect logs from (etcd, kube-apiserver, etc.)
  • JOURNALD_LOGS: Services to collect journald logs for
  • VAR_LOG_FILES: Log files to collect from /var/log
  • SPACE: Minimum disk space required (MB)
  • TIMEOUT: Command timeout in seconds

Script Architecture

The script follows a sequential collection pattern:

  1. setup() - Creates temp directory, sets up paths
  2. disk-space() - Validates available disk space
  3. sherlock() - Auto-detects OS, k8s distribution (rke/rke2/k3s), and init system
  4. system-all() - Collects OS-level info (disk, processes, sysctl)
  5. networking() - Comprehensive network data collection (see below)
  6. Distribution-specific functions (rke(), rke2(), k3s(), etc.)
  7. archive() - Creates tar.gz output in /tmp (or custom directory)

Networking Collection (networking() function)

The networking() function collects comprehensive network diagnostics:

  • Firewall: iptables (all tables including raw), ip6tables, nftables, firewalld
  • IP/Routing: addresses, routes, rules, neighbors, multicast groups
  • Sockets: ss/netstat for TCP, UDP, Unix, raw sockets
  • Namespaces: network namespace list
  • Bridge/VLAN: bridge links, FDB entries, VLAN config
  • DNS: systemd-resolved status and configuration
  • systemd-networkd: networkctl status and configs
  • Interface details: ethtool for all interfaces (driver, features)
  • Kernel tunables: filtered sysctl net.* parameters
  • CNI-specific: Calico (calicoctl), Cilium (cilium CLI) when available

Supported Kubernetes Distributions

  • RKE2 (auto-detected via rke2 binary)
  • K3s (auto-detected via k3s binary)
  • RKE1 (auto-detected via docker binary)
  • Kubeadm (manual override required: -r kubeadm)

Development Notes

  • Script is designed to run on production nodes with minimal resource impact
  • Uses nice/ionice with lowest priority by default (override with -p)
  • Output is sanitized where possible, but -o flag adds IP/hostname obfuscation
  • Collection is intentionally limited to avoid sensitive data