cockpit-temps — Cockpit Temperature Monitoring Plugin

A Cockpit plugin for Ubuntu Server that visualises hardware temperature data (CPU, NVMe, etc.) as interactive line charts with configurable thresholds. Data is collected and archived by PCP (Performance Co-Pilot) via the pmdalmsensors PMDA, giving you up to 120 days of history with automatic archive rotation and disk-budget enforcement.

This is a personal hobby project I build for my own use and publish in case it's useful to someone else. I work on it in my spare time, so issues and PRs are welcome but replies may be slow. Use at your own risk.

Features

Historical charts — query PCP archives up to 1 year back
Sensor grouping — configurable groups (CPU, NVMe, etc.) with custom labels
Threshold lines — global defaults (70 / 75 / 80 °C) plus per-sensor overrides
Date range controls — quick presets (1h, 24h, 7d, 30d, 90d, 1y) and custom range
Auto downsampling — step size scales with time range (1 min → 6 h)
Metric discovery — UI button lists all available PCP lmsensors metrics
Archive rotation — pmlogger_daily + cron-based disk budget
Zero external deps — plain JS, inline SVG chart, no build step

File Tree

cockpit-temps/
├── manifest.json              # Cockpit plugin manifest
├── index.html                 # Main HTML page
├── app.js                     # Application logic + SVG chart
├── style.css                  # Styles
└── config/
    └── sensors.json           # Sensor mapping, thresholds, retention config  ← you must edit this

scripts/
├── setup_pcp_temps.sh         # Install PCP + lmsensors PMDA + rotation
├── install_plugin.sh          # Deploy plugin into Cockpit
└── uninstall_plugin.sh        # Remove plugin (optional --purge)

Installation

Step 1 — Clone the repo

git clone <this-repo> cockpit-temps-repo
cd cockpit-temps-repo

Step 2 — Run the PCP setup script (as root)

sudo bash scripts/setup_pcp_temps.sh

This will:

Install lm-sensors, pcp, cockpit, cockpit-pcp
Run sensors-detect to load kernel sensor modules
Install the lmsensors PMDA
Configure pmlogger to archive lmsensors metrics every 60 s
Set up archive rotation (120 days retention, 10 GB max)
Enable and start all services

Configurable variables (set before running or edit the script):

Variable	Default	Description
`RETENTION_DAYS`	120	Keep archives this many days
`MAX_SIZE_GB`	10	Max total archive disk usage
`PMLOGGER_INTERVAL`	60	Logging interval in seconds

Example:

sudo RETENTION_DAYS=180 MAX_SIZE_GB=2 bash scripts/setup_pcp_temps.sh

Step 3 — Configure sensors for your hardware (required)

This step is required on every machine. The default sensors.json is configured for specific hardware (an Intel CPU and one NVMe drive at a specific PCI address). Your system will almost certainly have different metric names. The plugin will load but show no data until this is done.

Find your metric names:

pminfo -t lmsensors

Example output (your output will differ):

lmsensors.coretemp_isa_0000.package_id_0  [coretemp-isa-0000 Package id 0]
lmsensors.coretemp_isa_0000.core_0        [coretemp-isa-0000 Core 0]
lmsensors.coretemp_isa_0000.core_1        [coretemp-isa-0000 Core 1]
lmsensors.nvme_pci_0100.composite         [nvme-pci-0100 Composite]
lmsensors.nvme_pci_0100.sensor_1          [nvme-pci-0100 Sensor 1]

The naming convention is lmsensors.<chip>.<feature> where the chip name is the adapter name from sensors with dashes replaced by underscores (e.g. coretemp-isa-0000 → coretemp_isa_0000).

Common hardware differences:

Hardware	Example chip name
Intel CPU	`coretemp_isa_0000`
AMD CPU	`k10temp_pci_00c3` (address varies)
NVMe at 01:00	`nvme_pci_0100`
NVMe at 02:00	`nvme_pci_0200`
Second NVMe	`nvme_pci_0300` (address varies)

Edit cockpit-temps/config/sensors.json to match your metric names. A minimal example for one CPU package sensor and one NVMe:

{
    "groups": [
        {
            "id": "cpu",
            "label": "CPU",
            "sensors": [
                {
                    "id": "cpu_package",
                    "label": "CPU Package",
                    "metric": "lmsensors.coretemp_isa_0000.package_id_0",
                    "default": true,
                    "thresholds": null
                }
            ]
        },
        {
            "id": "nvme",
            "label": "NVMe",
            "sensors": [
                {
                    "id": "nvme_composite",
                    "label": "NVMe Composite",
                    "metric": "lmsensors.nvme_pci_0100.composite",
                    "default": true,
                    "thresholds": [
                        { "value": 75, "label": "NVMe Composite Warning",  "color": "#FFA726" },
                        { "value": 80, "label": "NVMe Composite Critical", "color": "#E53935" }
                    ]
                }
            ]
        }
    ],
    "thresholds": {
        "global": [
            { "value": 70, "label": "Warning 70\u00b0C",  "color": "#FFA726", "style": "dashed" },
            { "value": 75, "label": "Smartd 75\u00b0C",   "color": "#FF7043", "style": "dashed" },
            { "value": 80, "label": "Critical 80\u00b0C", "color": "#E53935", "style": "dashed" }
        ]
    },
    "retention": {
        "days": 120,
        "maxSizeGB": 10
    },
    "archiveBase": "/var/log/pcp/pmlogger"
}

Remove sensor entries whose metrics do not exist on your hardware — they will simply produce no data but cause no errors.

Step 4 — Install the plugin

sudo bash scripts/install_plugin.sh

This also disables PCP archive compression ($PCP_COMPRESSAFTER=never), which is required because pmrep cannot read .xz-compressed archives when given a directory argument. See PCP Archive Compression for details.

Re-run this command any time you change sensors.json.

Step 5 — Open Cockpit

Navigate to https://<server-ip>:9090 and click Temperatures in the menu.

Sensor Configuration Reference

Global thresholds

Defined under thresholds.global in sensors.json. These horizontal lines are drawn on every chart regardless of which sensors are selected.

Per-sensor thresholds

Each sensor entry can include a thresholds array to add sensor-specific lines:

{
    "id": "nvme_composite",
    "label": "NVMe Composite",
    "metric": "lmsensors.nvme_pci_0100.composite",
    "thresholds": [
        { "value": 75, "label": "NVMe Composite Warning",  "color": "#FFA726" },
        { "value": 80, "label": "NVMe Composite Critical", "color": "#E53935" }
    ]
},
{
    "id": "nvme_sensor1",
    "label": "NVMe Sensor 1 (controller chip)",
    "metric": "lmsensors.nvme_pci_0100.sensor_1",
    "thresholds": [
        { "value": 82, "label": "NVMe Sensor 1 Warning",  "color": "#FFA726" },
        { "value": 90, "label": "NVMe Sensor 1 Critical", "color": "#E53935" }
    ]
}

Set "thresholds": null to inherit only the global thresholds.

Thresholds — NVMe sensor rationale

NVMe drives expose multiple temperature sensors that measure different physical locations and have very different normal operating ranges:

Sensor	What it measures	Normal range under load
Composite	Drive-level aggregate	50–70 °C
Sensor 1	Controller chip	65–82 °C
Sensor 2	NAND flash	50–70 °C

Samsung PM9A1 (MZVL8512HELU, OEM 980 Pro Gen4): The controller chip (Sensor 1) routinely runs 15–20 °C hotter than the composite temperature. This is expected behaviour — the drive begins thermal throttling only when the composite temperature approaches 82 °C. The authoritative check for thermal problems is nvme smart-log:

nvme smart-log /dev/nvme0 | grep -E "Warning Temperature Time|Critical Composite Temperature Time|Thermal Management T[12] Trans Count"

If all three fields report 0, the drive has never throttled and the higher Sensor 1 readings are normal. A global threshold calibrated against composite temperatures (e.g. 75 °C) will fire false alarms for Sensor 1. Use sensor-specific thresholds as shown above to avoid this.

The `default` flag

Sensors with "default": true are pre-checked when the plugin loads. All others must be selected manually.

PCP Archive Compression (Important)

PCP archive compression must be disabled for this plugin to work.

pmlogger_daily compresses rotated PCP archives by default (.0 → .0.xz, .meta → .meta.xz) but leaves .index files uncompressed. When pmrep is given a directory as its -a argument, it tries to open all archive sets found there. The orphaned .index files (pointing to compressed .0.xz/.meta.xz that pmrep cannot read) cause pmrep to fail with PM_ERR_NAME Unknown metric name for every metric.

The install script handles this automatically by setting $PCP_COMPRESSAFTER=never in /etc/pcp/pmlogger/control.d/local.

If archives are already compressed

If you installed PCP before running the install script, some archives may already be compressed. To fix this:

Move .xz files out of the archive directory:

ARCHIVE_DIR=/var/log/pcp/pmlogger/$(hostname)
mkdir -p /tmp/pcp-compressed-backup
mv "$ARCHIVE_DIR"/*.xz /tmp/pcp-compressed-backup/ 2>/dev/null

Remove orphaned .index files (those whose matching .0 file is missing):

for idx in "$ARCHIVE_DIR"/*.index; do
    base="${idx%.index}"
    if [[ ! -f "$base.0" ]]; then
        rm -v "$idx"
    fi
done

Restart pmlogger:
```
sudo systemctl restart pmlogger
```

Archive Rotation & Disk Budget

How it works

pmlogger_daily runs via systemd timer (or cron) once per day:
- Creates a new daily archive
- Removes archives older than RETENTION_DAYS (-k N)
pcp-archive-budget cron script (/etc/cron.daily/pcp-archive-budget):
- Checks total archive size per host
- If over MAX_SIZE_GB, removes oldest archive sets until under budget

Verification commands

# Check archive size
du -sh /var/log/pcp/pmlogger/$(hostname)/

# List archive files
ls -lhS /var/log/pcp/pmlogger/$(hostname)/ | head -20

# Check pmlogger_daily timer
systemctl list-timers | grep pmlog

# Check retention config
cat /etc/default/pmlogger    # or /etc/sysconfig/pmlogger

# Manually trigger budget enforcement (safe to run)
sudo bash /etc/cron.daily/pcp-archive-budget

Adjusting retention after install

Edit /etc/default/pmlogger (or /etc/sysconfig/pmlogger):

PMLOGGER_DAILY_PARAMS="-E -k 180 -x 0"   # 180 days
PCP_MAX_SIZE_GB=3                          # 3 GB budget

Then restart: sudo systemctl restart pmlogger

Troubleshooting

Plugin doesn't appear in Cockpit menu

ls -la /usr/share/cockpit/cockpit-temps/
cat /usr/share/cockpit/cockpit-temps/manifest.json
sudo systemctl restart cockpit.socket

Chart shows no data / "No data points found"

Work through this checklist in order:

Check that PCP services are running:
```
systemctl status pmcd pmlogger
```

Check that archives exist and contain data:

ls /var/log/pcp/pmlogger/$(hostname)/
pmrep -a /var/log/pcp/pmlogger/$(hostname)/ -o csv -H -t 60sec \
  -S "-10minutes" lmsensors.coretemp_isa_0000.package_id_0

Check that the metric names in sensors.json match your hardware:
```
pminfo -t lmsensors
```
If the names differ, update sensors.json and re-run install_plugin.sh.
If archives are empty, wait a few minutes — pmlogger needs time after startup to write the first data points.

pminfo shows no lmsensors metrics

# Reinstall the PMDA
cd /var/lib/pcp/pmdas/lmsensors
sudo ./Remove
sudo ./Install
sudo systemctl restart pmcd
pminfo lmsensors

Archive disk usage growing too large

du -sh /var/log/pcp/pmlogger/$(hostname)/
sudo bash /etc/cron.daily/pcp-archive-budget
sudo sed -i 's/-k [0-9]*/-k 90/' /etc/default/pmlogger
sudo systemctl restart pmlogger

Acceptance Tests

1. Sensors detected

sensors
# Expected: output showing your hardware adapters (coretemp, nvme, k10temp, etc.)

2. PCP lmsensors metrics available

pminfo -t lmsensors
# Expected: list of lmsensors.* metric names matching your hardware

3. Live data via pmval

# Replace the metric name with one from your pminfo output
pmval -s 3 -t 2sec lmsensors.coretemp_isa_0000.package_id_0
# Expected: 3 temperature samples, e.g. 55.000, 56.000, 55.000

4. Archive data exists

# Wait a few minutes after setup, then:
pmrep -a /var/log/pcp/pmlogger/$(hostname)/ -o csv -H -t 60sec \
  -S "-10minutes" lmsensors.coretemp_isa_0000.package_id_0
# Expected: CSV rows with timestamps and temperature values

5. UI smoke test (manual)

Open Cockpit → Temperatures
Select a sensor checkbox
Choose a time preset → click Fetch data
Expected: line chart renders with data points
Hover over the chart → tooltip shows timestamp and temperature value
Threshold lines are visible

6. Survives reboot

sudo reboot
# After reboot:
systemctl is-active pmcd pmlogger cockpit.socket
# Expected: all "active"

Uninstall

# Remove plugin only
sudo bash scripts/uninstall_plugin.sh

# Remove plugin + PCP config (archives are preserved)
sudo bash scripts/uninstall_plugin.sh --purge

License

MIT

Part of a Cockpit plugin suite

I build a small family of Cockpit plugins for home servers, all dependency-light and made to be readable at a glance:

cockpit-temps — hardware temperature history with thresholds (this plugin)
cockpit-smart — S.M.A.R.T. disk health with trend tracking
cockpit-pcloud — pCloud storage quota and backup folder status
cockpit-tailscale — plain-language Tailscale network overview

Browse them all via the cockpit-plugin topic.

Name		Name	Last commit message	Last commit date
Latest commit History 29 Commits
cockpit-temps		cockpit-temps
docs		docs
scripts		scripts
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation