Hello,
A few days ago I updated my Proxmox server, and since then the fan has been much louder than usual, even audible from several meters away.
I spent all morning today looking into what it could be. It was clear that if the fan was running "full blast," it was due to the temperature. And the temperature was due to the CPU usage.
Indeed, the CPU on all four cores was at 25% (the CPU0 core was always at 100%). And it wasn't a Proxmox process or one of the services I run on it. It was something in the kernel.
Investigating, I saw that going back to the previous kernel everything works correctly. Then I checked the microcode and saw that it wasn't updated. I updated the Intel microcode, and it still worked fine with the previous kernel (the CPU goes down even with the LXC and VMs running).
The definitive test was to go back to the latest kernel and verify that the problem had been solved. At least with my CPU, microcode mitigations are necessary so that kernel 6.8.12-14 doesn't cause CPU0 to remain in the C0 state all the time.
I've documented it here (in Spanish):
https://www.manelrodero.com/blog/kernel-681214pve-provoca-100-en-cpu0
But the Google Translate translation is quite good if you want it in English:
https://www-manelrodero-com.transla..._sl=es&_x_tr_tl=en&_x_tr_hl=en&_x_tr_pto=wapp
Best regards, and thanks to everyone who helps in the community.
A few days ago I updated my Proxmox server, and since then the fan has been much louder than usual, even audible from several meters away.
I spent all morning today looking into what it could be. It was clear that if the fan was running "full blast," it was due to the temperature. And the temperature was due to the CPU usage.
Indeed, the CPU on all four cores was at 25% (the CPU0 core was always at 100%). And it wasn't a Proxmox process or one of the services I run on it. It was something in the kernel.
Investigating, I saw that going back to the previous kernel everything works correctly. Then I checked the microcode and saw that it wasn't updated. I updated the Intel microcode, and it still worked fine with the previous kernel (the CPU goes down even with the LXC and VMs running).
The definitive test was to go back to the latest kernel and verify that the problem had been solved. At least with my CPU, microcode mitigations are necessary so that kernel 6.8.12-14 doesn't cause CPU0 to remain in the C0 state all the time.
I've documented it here (in Spanish):
https://www.manelrodero.com/blog/kernel-681214pve-provoca-100-en-cpu0
But the Google Translate translation is quite good if you want it in English:
https://www-manelrodero-com.transla..._sl=es&_x_tr_tl=en&_x_tr_hl=en&_x_tr_pto=wapp
Best regards, and thanks to everyone who helps in the community.
Last edited: