Good morning all,
I am experiencing what looks to be a spike in CPU usage every 5 seconds (shown in the dashboard widget) when on the web GUI.
It looks like it is comprised of 5-6% system usage, 10-14% user usage, and 0% interrupt usage, and is extremely regular in occurrence - happening every 5-6 seconds. It must also be noted that there is minimum traffic at this point.
I have 3 VLANs configured - a management, IoT, and primary VLAN. I have Wireguard setup, but I am not actively using a tunnel at this point. I have approx. 5 aliases setup.
I am not logging anything from the firewall rules I have set, and am not using Netflow or any IDS/IPS.
My WAN is DHCP (not PPPoe, etc.) 1000/105.
I am running Opnsense 25.1.11-amd64 in a Proxmox VM, on a little N150 box with 4 x Intel i226-v NICs. The VM is allocated 3 x 'host' type cores, with NIC passthrough of 2 x NICs and 8Gb of memory. I am using CRC offloading, TSO offloading and VLAN hardware filtering. Proxmox is up to date and other than this performance seems fine.
I am not using traffic shaping.
I have tried using top and catching these spikes by monitoring the shell but I'm struggling to. It appears as though they are a php-cgi process and/or a python3.11 process. However, when I press 'c', it won't let me view the full path of the process.
Try this, points 16 and 17 (https://forum.opnsense.org/index.php?topic=42985.0).
The relevant part of the discussion starts here (https://forum.opnsense.org/index.php?msg=220563).
Thank you meyergru. I did try searching and couldn't find anything that replicated my issue, however your second link looks promising - I will look into it.
I can't get CPU temps as I am running on Proxmox but it looks as though the CPU method mentioned also collects usage data?
It looks as though our hardware is very similar, have you experienced this issue yourself? Or is the patch you looked at making a result of that?
Thanks again, appreciate your help
I use bare metal. The mentioned sysctl call will be done regardless of if it returns anything and thus cause the spikes. The limiting patch could not be applied, as discussed on Github, so there will be spikes regardless. However, this is not a problem in and of itself and it only turns up when you actually use the dashboard.
Thanks meyergru, I just looked into applying this manually but couldn't see temperature.sh within /usr/local/opnsense/scripts/system.
It's not a huge problem just annoying to see up to 20% CPU usage across 3 x cores (n150) for such a task every 5 seconds. Although, as you say I guess this sysctl call is only made when using the dashboard?
What else is it used for other than CPU temps/ usage?
IDK, it has been a while since then and maybe the code has been changed. It can all openly be scrutinized on Github.