13.3 Host metrics — CPU, RAM, disk
Before fancy stacks, know what the host is telling you.
| Metric | Healthy habit | Rough red flag (lab intuition) |
|---|---|---|
| CPU | Check load vs core count over time | Stuck near 100% with rising latency |
| RAM | Watch used vs available; swap thrash | OOM kills in dmesg / journal |
| Disk space | Free % on / and data mounts |
Climbing toward full with no cleanup |
| Disk I/O | Slow writes under load | API waits on disk-bound work |
| Inodes | Many tiny files can exhaust inodes while space remains | df -i shows 100% |
# Quick host pulse (Amazon Linux / Ubuntu habits)
uptime
free -h
df -h
df -i
ss -tuln # which published ports (ports) are listening — logical ports, not CPU pins
What you see: load average, memory summary, filesystem free space, and listening TCP/UDP ports as logical service numbers (published ports on one building IP) — same ports analogy as earlier chapters.
For deeper Linux process and journal work, refresh the AWS Cloud Linux chapters rather than re-teaching them here.