ESXi CPU Temperature Monitor: A Shell Script for Real-Time IPMI Alerting

Date: May 2, 2026
Author: IT Solutions USA
Category: VMware / ESXi
Tags: ESXi, VMware, IPMI, Shell Script, BMC, Thermal Monitoring, DevOps, Infrastructure

ESXi does not ship with a continuous, color-coded temperature dashboard. If you manage bare-metal hosts in a lab or production environment and want immediate visual feedback when a CPU runs hot — without standing up a full monitoring stack — this script covers the gap in under 30 lines of POSIX shell.

Script Variables Reference

IPMI Facts

Why IPMI?

VMware ESXi exposes hardware sensor data through the esxcli hardware ipmi sdr list command, which queries the host's BMC (Baseboard Management Controller) via the in-kernel IPMI driver. This is the same data source your server's iDRAC, iLO, or IPMI web interface reads — except here it's accessible directly from the ESXi shell without any separate network configuration or out-of-band management port.

The SDR (Sensor Data Repository) contains every hardware sensor the BMC knows about: CPU temperatures, inlet/exhaust temps, fan speeds, voltages, power draw. The script filters this down to CPU temperature rows only and presents them in a live-refreshing view.

The Complete Script

#!/bin/sh
# CPU Temperature Monitor for ESXi
# Continuously monitors CPU temps, prints in red if >70C

THRESHOLD=70      # Temperature threshold for alert (C)
INTERVAL=10       # Seconds between checks

while true; do
    clear
    echo "CPU Temperature Monitor - $(date -u)"
    echo "---------------------------------------"

    # Get CPU temperatures from IPMI sensor list
    esxcli hardware ipmi sdr list | grep "CPU" | while read -r line; do
        # Extract sensor name and temperature value (numbers only)
        name=$(echo "$line" | awk '{print $4,$5}')
        temp=$(echo "$line" | awk '{print $7}' | grep -o '[0-9]\+')

        # Check if temperature is numeric
        if [ -z "$temp" ]; then
            temp=0
        fi

        # Display in red if above threshold
        if [ "$temp" -gt "$THRESHOLD" ]; then
            echo -e "$name: \033[31m${temp}C\033[0m"
        else
            echo "$name: ${temp}C"
        fi
    done

    echo "---------------------------------------"
    sleep $INTERVAL
done

How the Color Logic Works

The highlight mechanism is a simple integer comparison feeding into ANSI escape codes — the same codes used by terminal emulators since the 1970s.

When a temperature exceeds the threshold, the script wraps the value in \033[31m (set foreground to red) and \033[0m (reset all attributes). The ESXi BusyBox shell's echo -e flag is required to interpret these escape sequences — plain echo without -e will print the literal escape string instead of rendering the color.

The session stays alive because the outer while true loop never exits. The sleep $INTERVAL call at the end of each cycle prevents the loop from consuming a full CPU core.

Deploying the Script

Save the script to a persistent datastore path — the ESXi ramdisk (/tmp) is wiped on reboot. A recommended location is your local datastore:

# Save to a persistent datastore
vi /vmfs/volumes/datastore1/scripts/cpu_monitor.sh

# Make it executable
chmod +x /vmfs/volumes/datastore1/scripts/cpu_monitor.sh

# Run it
/vmfs/volumes/datastore1/scripts/cpu_monitor.sh

To start it automatically after a host reboot, add the call to /etc/rc.local.d/local.sh using nohup so it persists without an interactive session:

# In /etc/rc.local.d/local.sh — runs at boot
nohup /vmfs/volumes/datastore1/scripts/cpu_monitor.sh \
  > /vmfs/volumes/datastore1/scripts/cpu_monitor.log 2>&1 &

With this in place, the monitor starts silently at boot, logs to the datastore path (where you can grep for past thermal events), and you can attach to its output with tail -f at any time. The log is on the datastore so it survives host reboots — do not redirect it to /tmp, which is a ramdisk wiped on every reboot.

Tuning for Your Environment

Two variables cover most customisation needs without touching the script logic.

THRESHOLD=70
- Intel Xeon E5/E7 series: set to 75–80 (against Tcase; check your specific SKU datasheet). - AMD EPYC: check AMD's published Tccd and Tctl limits — these CPUs report die temperature (Tdie) which runs higher than traditional Tcase. - Consumer CPUs in a lab setting: 65 is a conservative starting point.

INTERVAL=10
- 5 s — active troubleshooting during a workload spike. - 10 s — normal monitoring, default. - 30 s — background logging without clear, append output to a file instead.

If your server has more than two CPUs or uses a non-standard naming convention, change grep "CPU" to grep -i "temp" to catch all temperature sensors regardless of label.

Limitations and What This Doesn't Replace

This script is a lightweight operational tool, not a replacement for a proper monitoring stack. It has no persistent alerting, no trend history, no multi-host view, and no integration with incident management. For production environments, pair it with Prometheus + node_exporter (IPMI exporter), vCenter alarms, or your organisation's SIEM.

Where this script excels is in situations where none of the above are available: a standalone lab ESXi host, a quick post-maintenance thermal check, or an environment where deploying Prometheus is more overhead than the task warrants.