When a server is on fire, you need a fast, repeatable triage checklist. These 10 commands give you a full picture of CPU, memory, disk I/O, network, and process health in under 60 seconds. Based on Brendan Gregg's USE Method.
The 60-Second Checklist
1. System Load Overview
uptime — are things getting worse?bash
1
uptime
Look at the three load averages (1, 5, 15 minutes). If the 1-minute average is higher than the 15-minute, things are getting worse. Compare to your CPU count (nproc).
2. Kernel Messages
dmesg — hardware or kernel problems?bash
1
dmesg -T | tail -20
Look for OOM kills, disk errors, or hardware faults. The -T flag gives human-readable timestamps.
3. Overall System Stats
vmstat — CPU, memory, I/O at a glancebash
1
vmstat 1 5
Key columns to watch:
Column
Meaning
Red Flag
r
Processes waiting for CPU
Greater than CPU count
b
Processes blocked on I/O
Any non-zero value
si / so
Swap in/out
Any non-zero means swapping
us
User CPU %
High = app is busy
sy
System CPU %
High = kernel overhead
wa
I/O wait %
High = disk bottleneck
st
Steal time %
High = noisy neighbor (VM/cloud)
4. CPU Usage Per Core
mpstat — is one core pegged?bash
1
mpstat -P ALL 1 3
A single core at 100% while others are idle usually means a single-threaded bottleneck in your app.
5. Top Processes
pidstat — what's eating CPU?bash
1
pidstat 1 3
Use pidstat -d 1 3 for per-process disk I/O, and pidstat -r 1 3 for per-process memory usage.
6. Disk I/O
iostat — which disk is busy?bash
1
iostat -xz 1 3
Column
Red Flag
%util
Above 80% means the disk is saturated
await
Average I/O latency in ms — high means slow disk
r/s / w/s
Read/write operations per second
7. Memory
free — how much memory is actually free?bash
1
free -h
available is the real number — it includes reclaimable cache
If available is near zero and swap is in use, you're memory-starved
8. Network Connections
ss — connection states and countsbash
1
ss -s
For detailed per-connection info: ss -tunapl. The flags stand for TCP, UDP, numeric, all states, processes, listening — easier to remember as "tuna please".
9. Disk Space
df — any filesystems full?bash
1
df -h
A 100% full / filesystem can cause cascading failures — logs can't write, temp files fail, databases crash. Set up monitoring alerts at 80%.
