System Monitoring

When you SSH into a box and need to know what's going on right now.

TUI / interactive

• ★ btop — beautiful, mouse-aware, the modern default; replaces htop / glances for many.
• ★ htop — classic; everywhere; still great.
• glances — Python, broader-scope (also network / disk); web mode.
• atop — captures historical data; great for "what was the system doing at 3 AM?"
• bashtop — older btop predecessor.
• gtop — Node-based; nice graphs.
• bottom (btm) — Rust; modern alternative to btop.

Disk usage

• ★ ncdu — interactive disk usage. The default.
• ★ dust — modern Rust replacement; faster, prettier.
• duf — replaces df; better tables.
• df -h, du -sh * — built-in classics.
• gdu — Go alternative to ncdu.

I/O

• iotop — what's writing / reading?
• iostat (sysstat) — disk I/O metrics.
• dstat (or dool) — combined CPU / disk / network.
• biotop (BCC) — eBPF I/O analysis.

Process / memory

• ps aux — universal.
• pgrep / pkill — find / signal by name.
• ss — sockets; replaces netstat.
• free / vmstat — memory and VM stats.
• smem — accurate per-process memory (PSS).
• lsof — open files / sockets.

Whole-host dashboards

• ★ Netdata — beautiful per-node real-time + 1-second metrics. Free; cloud option for fleet aggregation. The "install on every box" pick.
• Cockpit (Red Hat) — web UI for managing servers; included with most modern distros.
• Webmin / Virtualmin — older web UI; still around.
• Zabbix / LibreNMS — heavier monitoring platforms.
• Munin — old-school; still works.

Linux performance toolkit

• perf — hardware perf counters; the deepest profiling tool on Linux.
• strace / ltrace — syscall / library call tracing.
• bpftrace / bcc — eBPF magic; see eBPF Tools.
• Brendan Gregg's flamegraph.pl — generate flame graphs from perf samples.

Containers / k8s

• lazydocker — TUI for Docker.
• ctop — top-style container view.
• k9s — see k8s CLI tools.
• kube-state-metrics + node-exporter — for Prometheus.

Hardware / sensors

• lm-sensors — temperature / fan / voltage.
• smartctl (smartmontools) — disk health.
• dmidecode — hardware inventory.
• hwinfo / lshw / lspci / lsblk / lsusb — hardware enumeration.

Patterns to adopt

• ★ Install Netdata on every box. Two minutes; free; gives you 90% of monitoring.
• htop or btop muscle memory. Pick one and learn its keybindings.
• Use atop for after-the-fact triage. "Box was slow at 02:14" — atop has the answer.
• Disk monitoring with smart + zfs/btrfs scrubs scheduled.
• OOM-killer logs in dmesg — check there before you assume the app crashed itself.

Pick this if…

• Default install on every server: Netdata + htop / btop + ncdu.
• Need historical per-second data: atop.
• Web UI for managing boxes: Cockpit.
• Heavy multi-host monitoring: Zabbix or LibreNMS or Prometheus stack.
• Performance investigation: perf + bpftrace + flamegraph.