[Bug]: Proxmox cluster issues (agent registration + cluster network detection + VM metrics)

[AlexODC]

Describe the bug

Hello,

First of all, thank you for Pulse. It works extremely well in simple environments and I really like the project.

However, I am encountering several issues when using Pulse with a Proxmox cluster.
On a single-node Proxmox lab (no cluster) everything works perfectly.

I apologize in advance for the length of this issue, but I want to provide as much information as possible to help diagnose the problems.

Environment

Pulse version: v5.1.19,
Proxmox: cluster environment

Network setup:

Proxmox Management / API network:
10.15.2.0/24
Proxmox cluster (corosync) network:
10.15.5.0/24
Pulse server:
10.15.6.10

Problem 1 – Agent registration reports failure

When installing the Pulse agent on Proxmox nodes using the setup command:

curl -sSL "http://10.15.6.10:7655/api/setup-script?type=pve&host=https%3A%2F%2F10.15.2.1%3A8006&pulse_url=http://10.15.6.10:7655&backup_perms=true" | PULSE_SETUP_TOKEN=TOKEN bash

The agent logs (systemctl status pulse-agent) show:
Proxmox token created but registration failed (node may need manual configuration)

However:

• the agent does appear in Pulse
• metrics are collected
• "last used" timestamp updates

So the node seems functional despite the error. It's actual state isn't clear.

Additionally, the node does not appear in the Proxmox page even though the installation process says the Proxmox integration is enabled.

Problem 2 – Cluster network detection

Pulse attempts to communicate with Proxmox using the cluster network (10.15.5.x) instead of the management/API network (10.15.2.x), evenf when using environment variables to allowlist/blacklist subnets.

Example:
Cluster network (corosync):
10.15.5.0/24
Proxmox API / GUI network:
10.15.2.0/24

Pulse discovers nodes using the 10.15.5.x addresses, which are not the addresses used for the API.
I think this causes instability when Pulse attempts to communicate with the cluster.

Is there a way to force Pulse to use a specific IP/network for Proxmox nodes?

For example, use 10.15.2.x instead of the discovered 10.15.5.x cluster addresses.

Problem 3 – Agent installation behaviour

The agent installation sometimes behaves differently depending the version of pulse.

When installing with the provided commands:

• the agent reports registration errors
• after running systemctl restart pulse-agent, the agent suddenly reports successful registration.
• in some pulse version it doesn't register but host show up on host page.

This causes nodes to appear in the Proxmox settings section, but not in the Proxmox dashboard.

Problem 4 – VM RAM metrics incorrect

When monitoring Linux VMs, the RAM usage appears to include cache, which leads to incorrect values.
This issue was supposedly fixed in recent updates but still seems to occur in my environment.

Problem 5 – Temperature monitoring use SSH
Logs shows that ssh is used to collect temperature sensors (documentation say that ssh isn't used with agent install)

Problem 6 – Wrong CEPH monitoring
CEPH monitors aren't detected (count 0) and only 1 manager while my cluster got 3 monitors and 2 managers
Same issue that this report: #1290

I performed many clean installations to isolate the issues of these different problems.

Steps repeated:

• reinstall Pulse container from scratch to avoid reinstallation/token conflicts
• full removal of agents (
curl -sSL "http://10.15.6.10:7655/api/setup-script?type=pve&host=https%3A%2F%2F10.15.2.1%3A8006&pulse_url=http://10.15.6.10:7655&backup_perms=true" | PULSE_SETUP_TOKEN=e0ae5701f1ad393ddd3296959d3621cd bash

curl -fsSL http://10.15.6.10:7655/install.sh | bash -s -- --uninstall --url http://10.15.6.10:7655/
)

• Install agents (all methods (agent, api only, manual) tried)

I repeated this process many times to test different installation orders:

• agent install → Proxmox API
• agent install → Proxmox integration
• Proxmox integration → agent install

Expected behaviour

• stable agent registration
• nodes appearing correctly in the Proxmox dashboard
• Pulse using the correct management network for Proxmox API
• accurate VM memory metrics
• temperature monitoring should not use ssh to be collected with agent install

Conclusion

Pulse works perfectly on non-cluster Proxmox environments.
However, with a cluster the behaviour becomes unstable and difficult to debug.

I would appreciate any guidance on:

• How to install proxmox cluster on pulse (install proxmox api of all server ? only one ? on a VIP IP that serve all server behind ?)
• forcing the IP/network used for Proxmox nodes
• resolving agent registration inconsistencies
• troubleshooting cluster detection

Thank you for your help and for this excellent project.

To reproduce

1. Deploy pulse instance
2. Copy the command in the installation page (or any other command)
3. Paste it on a host
4. Host pulse agent service say registration failed but show up in agent settings and agent dashboard but not on proxmox page.

Expected behavior

When executing the scripts provided by Pulse Server to add a Proxmox node, it should do what is expected:
Install the agent on the server.
Upload the correct login information and token.
Connect the agent to a Proxmox host.
Ensure that all metrics are reported correctly.

Pulse version

v5.1.19 (latest)

Agent version

v5.1.19

Image tag or digest

rcourtman/pulse:latest

Installation type

Docker Compose

Relevant logs or diagnostics

● pulse-agent.service - Pulse Unified Agent
     Loaded: loaded (/etc/systemd/system/pulse-agent.service; enabled; preset: enabled)
    Drop-In: /etc/systemd/system/pulse-agent.service.d
             └─override.conf
     Active: active (running) since Thu 2026-03-05 11:27:59 CET; 11s ago
 Invocation: 0f7e1a4b439144b1a5e6df0e81bc5602
   Main PID: 956056 (pulse-agent)
      Tasks: 23 (limit: 629145)
     Memory: 24.8M (peak: 89M)
        CPU: 4.331s
     CGroup: /system.slice/pulse-agent.service
             └─956056 /usr/local/bin/pulse-agent --url http://10.15.6.10:7655 --interval 30s --enable-host --enable-proxmox --insecure

Mar 05 11:27:59 pve1 pulse-agent[956056]: {"level":"info","time":"2026-03-05T11:27:59+01:00","message":"Command execution disabled (use --enable-commands to enable)"}
Mar 05 11:27:59 pve1 pulse-agent[956056]: {"level":"info","time":"2026-03-05T11:27:59+01:00","message":"Host agent module started"}
Mar 05 11:27:59 pve1 pulse-agent[956056]: {"level":"info","component":"host-agent","time":"2026-03-05T11:27:59+01:00","message":"Proxmox mode enabled, checking setup..."}
Mar 05 11:27:59 pve1 pulse-agent[956056]: {"level":"info","component":"host-agent","types":["pve"],"time":"2026-03-05T11:27:59+01:00","message":"Auto-detected Proxmox products"}
Mar 05 11:27:59 pve1 pulse-agent[956056]: {"level":"info","component":"host-agent","type":"pve","time":"2026-03-05T11:27:59+01:00","message":"Setting up Proxmox type"}
Mar 05 11:28:02 pve1 pulse-agent[956056]: {"level":"info","component":"host-agent","token":"pulse-10-15-6-10","time":"2026-03-05T11:28:02+01:00","message":"PVE monitoring token already exists; rotating token in place"}
Mar 05 11:28:03 pve1 pulse-agent[956056]: {"level":"info","component":"host-agent","type":"pve","token_id":"pulse-monitor@pam!pulse-10-15-6-10","time":"2026-03-05T11:28:03+01:00","message":"Created Proxmox API token"}
Mar 05 11:28:03 pve1 pulse-agent[956056]: {"level":"info","component":"host-agent","ip":"10.15.2.1","time":"2026-03-05T11:28:03+01:00","message":"Using user-specified ReportIP for Proxmox registration"}
Mar 05 11:28:03 pve1 pulse-agent[956056]: {"level":"warn","component":"host-agent","error":"auto-register returned HTTP 401: Invalid or expired setup code","type":"pve","time":"2026-03-05T11:28:03+01:00","message":"Failed to register Proxmox node with Pulse; if the node does not appear, delete /var/lib/pulse-agent/proxmox-pve-registered (or proxmox-pbs-registered) to >
Mar 05 11:28:03 pve1 pulse-agent[956056]: {"level":"warn","component":"host-agent","type":"pve","host":"https://10.15.2.1:8006","time":"2026-03-05T11:28:03+01:00","message":"Proxmox token created but registration failed (node may need manual configuration)"}

Confirmations

• I verified the exact Pulse version and image reference from the running instance.
• I searched existing issues for duplicates.

[raregtp]

Seeing the same issue, even after updating to 5.1.20. I'm running Proxmox clustered, and running Pulse in Docker Swarm fronted by Traefik.

Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"warn","error":"server responded with status 404 Not Found","time":"2026-03-05T22:24:00-06:0>
Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"info","version":"v5.1.20","pulse_url":"https://pulse.domain.com","host_agent":true,"docker_a>
Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"info","addr":":9191","time":"2026-03-05T22:24:00-06:00","message":"Health/metrics server li>
Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"info","time":"2026-03-05T22:24:00-06:00","message":"Command execution disabled (use --enabl>
Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"info","time":"2026-03-05T22:24:00-06:00","message":"Host agent module started"}
Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"info","component":"host-agent","time":"2026-03-05T22:24:00-06:00","message":"Proxmox mode e>
Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"info","component":"host-agent","types":["pve"],"time":"2026-03-05T22:24:00-06:00","message">
Mar 05 22:24:00 server pulse-agent[3198999]: {"level":"info","component":"host-agent","type":"pve","time":"2026-03-05T22:24:00-06:00","message":"S>
Mar 05 22:24:08 server pulse-agent[3198999]: {"level":"info","component":"host-agent","type":"pve","token_id":"pulse-monitor@pam!pulse-pulse-domain>
Mar 05 22:24:08 server pulse-agent[3198999]: {"level":"warn","component":"host-agent","error":"auto-register returned HTTP 401: Invalid or expired>
Mar 05 22:24:08 server pulse-agent[3198999]: {"level":"warn","component":"host-agent","type":"pve","host":"https://10.10.5.210:8006","time":"2026->
Mar 05 22:32:10 server pulse-agent[3198999]: {"level":"warn","component":"host-agent","error":"pulse responded with status 401 Unauthorized","time>
Mar 05 22:32:40 server pulse-agent[3198999]: {"level":"warn","component":"host-agent","error":"pulse responded with status 401 Unauthorized","time>
Mar 05 22:33:10 server pulse-agent[3198999]: {"level":"warn","component":"host-agent","error":"pulse responded with status 401 Unauthorized","time>
Mar 05 22:33:40 server pulse-agent[3198999]: {"level":"warn","component":"host-agent","error":"pulse responded with status 401 Unauthorized","time>

[AlexODC]

Hello,

First, thank you for your work and your reactivity that's insane.
With 5.1.20 update, i managed to configure the cluster in one command. CEPH cluster works now too. And temperatures seems to detect from the agent and not SSH and that's very good.

Problem that persists:

• RAM not detected proprely (linux cache included)
• Disk not detected proprely on some VM while guest agent is running and VM entirely restarted (Windows & Linux VM)
• Disks metrics disappears on all VM repeatedly like there is a cache problem (pulse detect disk on a lot of VM, and sudently they disappears).

Also, i added my first host of my cluster and it detected 7 nodes but now the Automatic failover enabled only detect the first node.

Here the logs to help for the debug:

pulse-support-bundle-20260306-095413.zip

Thank you !