Temperatures not displayed in dashboard even though sensors -j works (AMD CPU/GPU)

[kitopopo]

Body:

Pulse version: v4.25.0
Platform: LXC container on Proxmox VE
Hardware: AMD CPU (k10temp), AMD GPU (amdgpu), NVMe, WiFi

Problem description:

Pulse can connect to the Proxmox host via SSH and run sensors -j, returning a JSON with all temperature readings. However, none of these temperatures are shown in the Pulse dashboard.

From the Pulse container, running:

ssh -i /root/.ssh/pulse_rsa root@<Proxmox_IP> sensors -j

produces the following output:

{
  "nvme-pci-0300": {
    "Adapter": "PCI adapter",
    "Composite": { "temp1_input": 49.85 }
  },
  "amdgpu-pci-0400": {
    "Adapter": "PCI adapter",
    "vddgfx": { "in0_input": 0.831 },
    "vddnb": { "in1_input": 0.912 },
    "edge": { "temp1_input": 52.0 },
    "PPT": { "power1_input": 8.0 },
    "sclk": { "freq1_input": 200000000.0 }
  },
  "k10temp-pci-00c3": {
    "Adapter": "PCI adapter",
    "Tctl": { "temp1_input": 53.875 }
  },
  "iwlwifi_1-virtual-0": { "Adapter": "Virtual device", "temp1": {} }
}

This demonstrates that Pulse does receive temperature data via SSH, but the dashboard does not display it.

Expected behavior:

Pulse should display CPU, GPU, and NVMe disk temperatures in the dashboard.

Sensors returning null or empty values can be ignored, but valid sensors (k10temp, amdgpu, nvme) should be displayed.

Additional notes:

AMD hardware may be relevant, as Pulse might not correctly map k10temp or amdgpu sensors.

The error from some virtual sensors (e.g., iwlwifi_1-virtual-0) is normal and should not affect the other readings.

[BatisteDONDOGLIO]

I have the same issue.

My Pulse instance is on a physical machine tho. But I can ssh into my Proxmox node and run sensors -j just fine.

But it isn't displayed on the Pulse dashboard.

I even tried to remove the Pulse config from the Proxmox node and re-add it, but the issue remains.

I wonder if it's possible to use the new proxy feature from a physical machine?
@kitopopo This should be the default in your case, if I understand correctly. As LXC instances of Pulse seem to now use it in the latest version. So it should only fall back to using SSH if the proxy isn't working correctly. Unless I'm mistaken

[devachnid]

@kitopopo - I had this problem previously. The cause was that pulse was not running as root in the LXC, but as a user 'pulse', meaning the SSH keys need to be created for that user. I can see from your details above that the SSH keys are in root's home directory.

[BatisteDONDOGLIO]

@kitopopo - I had this problem previously. The cause was that pulse was not running as root in the LXC, but as a user 'pulse', meaning the SSH keys need to be created for that user. I can see from your details above that the SSH keys are in root's home directory.

I think this is the intended behavior, actually, and shouldn't be the cause of the issue.
See: #595 (comment)

Pulse falls back to a legacy SSH path for temperature collection when the proxy is unavailable, so it keeps trying to reach each node as the sensors key user (root by default)

[kitopopo]

@kitopopo - I had this problem previously. The cause was that pulse was not running as root in the LXC, but as a user 'pulse', meaning the SSH keys need to be created for that user. I can see from your details above that the SSH keys are in root's home directory.

I think this is the intended behavior, actually, and shouldn't be the cause of the issue. See: #595 (comment)

Pulse falls back to a legacy SSH path for temperature collection when the proxy is unavailable, so it keeps trying to reach each node as the sensors key user (root by default)

Hi, thanks for your help and advice.

I’ve followed your instructions and checked everything step by step:

-Pulse runs as user pulse inside the LXC container, not as root.

-The SSH keys are located in /opt/pulse/.ssh inside the container (id_rsa, id_rsa.pub, and known_hosts_sensors).

-The public key (id_rsa.pub) has been copied into Proxmox /root/.ssh/authorized_keys with the forced command:

command="sensors -j",no-port-forwarding,no-X11-forwarding,no-agent-forwarding,no-pty ssh-rsa AAAAB3...

From the Pulse container, running as the pulse user:

su -s /bin/bash pulse -c "ssh -i /opt/pulse/.ssh/id_rsa root@192.168.1.100 sensors -j"

Result: The command correctly returns the JSON with temperatures for the AMD CPU, NVMe, and GPU:

"amdgpu-pci-0400": {
  "edge": {"temp1_input":51.0},
  "vddgfx": {"in0_input":0.831},
  "vddnb": {"in1_input":0.906}
},
"k10temp-pci-00c3": {"Tctl":{"temp1_input":53.625}}

Therefore, the SSH connection and keys work perfectly. The issue seems to be inside Pulse when processing or displaying AMD temperatures, because although sensors -j returns valid data, the Pulse dashboard still does not show them.

My Pulse version is v4.25.0.

@rcourtman , could you please take a look and advise if there’s a known issue with AMD temperature reporting in LXC containers?

thanks in advanced.

Best regards

[arunas-devoro]

Same here AMD Epyc CPUs. no erros. but GUI does not show temp readings. I'm running Pulse on LXC

systemctl status pulse-sensor-proxy.service
● pulse-sensor-proxy.service - Pulse Temperature Proxy
Loaded: loaded (/etc/systemd/system/pulse-sensor-proxy.service; enabled; preset: enabled)
Active: active (running) since Sat 2025-10-25 09:40:40 EEST; 2h 4min ago
Docs: https://github.com/rcourtman/Pulse

Oct 25 11:44:56 pve1 pulse-sensor-proxy[3105744]: 11:44AM DBG Fetching temperature via SSH corr_id=53e20163-1660-4b9a-b00e-3e1426850eb1 method=get_temperature node=x.x.x.x pid=4003713 uid=100999
Oct 25 11:44:56 pve1 pulse-sensor-proxy[3105744]: 11:44AM INF Temperature data fetched successfully corr_id=53e20163-1660-4b9a-b00e-3e1426850eb1 method=get_temperature node=x.x.x.x pid=4003713 uid=100999
Oct 25 11:44:56 pve1 pulse-sensor-proxy[3105744]: 11:44AM INF Request completed corr_id=53e20163-1660-4b9a-b00e-3e1426850eb1 method=get_temperature pid=4003713 uid=100999

[appleimperio]

Same issue here on AMD Ryzen 7
Pulse version: v4.25.0
Platform: LXC container on Proxmox VE
if I run ssh root@your-proxmox-node "sensors -j"

get result {"spd5118-i2c-1-51":{"Adapter":"SMBus PIIX4 adapter port 0 at 0b00","temp1":{"temp1_input":34.000000,"temp1_max":55.000000,"temp1_min":0.000000,"temp1_crit":85.000000,"temp1_lcrit":0.000000,"temp1_max_alarm":0.000000,"temp1_min_alarm":0.000000,"temp1_crit_alarm":0.000000,"temp1_lcrit_alarm":0.000000}},"k10temp-pci-00c3":{"Adapter":"PCI adapter","Tctl":{"temp1_input":52.125000}},"nvme-pci-0500":{"Adapter":"PCI adapter","Composite":{"temp1_input":39.850000,"temp1_max":82.850000,"temp1_min":-273.150000,"temp1_crit":84.850000,"temp1_alarm":0.000000}},"acpitz-acpi-0":{"Adapter":"ACPI interface","temp1":{"temp1_input":20.000000}},"amdgpu-pci-7600":{"Adapter":"PCI adapter","vddgfx":{"in0_input":0.885000},"vddnb":{"in1_input":1.055000},"edge":{"temp1_input":46.000000},"PPT":{"power1_input":8.126000},"sclk":{"freq1_input":2200000000.000000}},"spd5118-i2c-1-50":{"Adapter":"SMBus PIIX4 adapter port 0 at 0b00","temp1":{"temp1_input":39.500000,"temp1_max":55.000000,"temp1_min":0.000000,"temp1_crit":85.000000,"temp1_lcrit":0.000000,"temp1_max_alarm":0.000000,"temp1_min_alarm":0.000000,"temp1_crit_alarm":0.000000,"temp1_lcrit_alarm":0.000000}},"nvme-pci-0100":{"Adapter":"PCI adapter","Composite":{"temp1_input":35.850000,"temp1_max":80.850000,"temp1_min":-273.150000,"temp1_crit":84.850000,"temp1_alarm":0.000000},"Sensor 1":{"temp2_input":45.850000,"temp2_max":65261.850000,"temp2_min":-273.150000},"Sensor 2":{"temp3_input":35.850000,"temp3_max":65261.850000,"temp3_min":-273.150000}},"r8169_0_200:00-mdio-0":{"Adapter":"MDIO adapter","temp1":{"temp1_input":50.000000,"temp1_max":120.000000}}}

and when I run this command curl -s http://localhost:7655/api/monitoring/scheduler/health \ | jq '.instances[] | select(.key | contains("temperature")) | {key, pollStatus}'

I get this error jq: error (at <stdin>:0): Cannot iterate over null (null) if you need any logs to help figure it out this issue let me know. Thank You

[Knallebal]

i have the same issue. v.4.25 amd cpu
pulse displays my cpu temp at 20 degrees C at all times. even when i mouse-over the range is 20 minimum and 20 maximum.

when i run sensors on the node-shell it displays the correct cpu-temp:

[BatisteDONDOGLIO]

Personally I don't even see the temps at all:

Despite "Temperature" being reported as monitored in the settings:

[xukhanhh]

Personally I don't even see the temps at all:

Despite "Temperature" being reported as monitored in the settings:

Hmm, I'm in you situation also but I don't have the 'Temperature', what need to be done to archive it, I just use the command generate by UI and run it on proxmox host. My Pulse server running on a VM

[kitopopo]

Thanks @rcourtman for your excellent work and for this great application.

After updating Proxmenux, we now have full visibility of sensors and temperatures directly from the host, which works perfectly. However, the Pulse container is still attempting SSH connections to the host every few seconds.

Is there any update or fix planned regarding this behavior?

Thanks again for all your effort and the continuous improvements to the project.

[Theetjuh]

Personally I don't even see the temps at all:

Despite "Temperature" being reported as monitored in the settings:

I have exactly the same issue, tried reinstalling the plugins multiple time, also an AMD CPU/iGPU, running latest 4.26.4 version.

[rcourtman]

4.26.4 already contains commit a145ed3 with the amdgpu chip parser, so lack of temps means either the running containers are still on an older cached build or lm-sensors is exposing the iGPU under a different chip name (k10temp, radeon, zenpower, etc.) that the new code does not cover yet. Reinstalling plugins will not refresh the main Pulse container or the host agent where the parser runs.

First confirm the versions that are actually live by running docker exec pulse cat /opt/pulse/VERSION in the server container or pulse-host-agent --version on any bare metal agents.

Second capture the full sensors -j output from the affected node and include it so the chip identifiers (amdgpu*, k10temp*, radeon*, anything else) are visible.

Third pull the host agent logs around startup or any sensor polling (for example journalctl -u pulse-host-agent --since "today" or the Docker logs) and look for lines mentioning chip detection or JSON parsing failures. Send those snippets.

Once those three items are in hand it will be clear whether the fix is simply not running or if the parser needs to be extended for the chip names your hardware uses.

[Theetjuh]

pulse-host-agent --version

Within the LXC I got Pulse running I don't see any pulse-host-agent, but /opt/pulse/VERSION shows 4.26.4
sensors -j shows: {"amdgpu-pci-0300":{"Adapter":"PCI adapter","vddgfx":{"in0_input":1.406000},"vddnb":{"in1_input":0.937000},"edge":{"temp1_input":71.000000},"PPT":{"power1_input":23.000000},"sclk":{"freq1_input":400000000.000000}},"nvme-pci-0200":{"Adapter":"PCI adapter","Composite":{"temp1_input":51.850000,"temp1_max":89.850000,"temp1_min":-273.150000,"temp1_crit":94.850000,"temp1_alarm":0.000000},"Sensor 1":{"temp2_input":51.850000,"temp2_max":65261.850000,"temp2_min":-273.150000},"Sensor 2":{"temp3_input":61.850000,"temp3_max":65261.850000,"temp3_min":-273.150000}},"k10temp-pci-00c3":{"Adapter":"PCI adapter","Tctl":{"temp1_input":73.000000}}}