Today’s Linux monitoring session was inspired by my Coursera coursework and the realization that my unattended Aryncore MCP server deserved a proper health audit.
I used `htop`, `ps`, and `systemctl` to evaluate load, memory, services, and rogue processes, then killed off a stale `node` instance hogging CPU. This practical check-in helped bridge the gap between academic shell concepts and my real-world infrastructure.
top
Moderate background load observed (1.8+). Docker and Prometheus behaving as expected. A rogue `node` process stood out.
ps aux --sort=-%cpu | head -n 20
Used to confirm the source of CPU drain. Docker, Prometheus, and a high-spike `node` instance.
ps aux --sort=-%mem | head -n 20
Memory bloat traced to GUI layers like LightDM and Unity — not essential for SSH-only usage.
systemctl list-units --type=service --state=running
System services healthy. No signs of stuck units.
sudo apt install lm-sensors
sudo sensors-detect
sensors
Collected hardware temps for future dashboarding.
sudo apt install nemo
nemo &
Swapped out buggy Nautilus for Nemo GUI — file manager works again under X11.
ollama list
Verified local models; Ollama was not the source of the drain.
kill -9 <PID>
Killed off the rogue `node` process, restoring CPU equilibrium.
This kind of live system evaluation is a critical aspect of running AI toolchains and container-based infrastructure. Lightweight GUI tools should be replaced on the server. Future improvements could include `glances`, `netdata`, and a basic audit script.
Signed: Lorelei Noble
Back to DevLogs