I hope you can help me at the following problem. First, regarding the installation:
My HA is installed on Proxmox as a virtual machine: 8GB RAM / 40 GB HDD. For some time now, I have noticed that the CPU load, which can be seen both under Proxmox and under Home Assistant, occasionally spikes significantly.
Unfortunately, I don’t have a good starting point for where to look. The error log does not provide any direct clues.
I would have expected increased activity during nights on backup times, but you do not seem to have those. At what times are backups scheduled, and do these get uploaded to e.g. Nabu Casa or a NAS? If you do not backup to anything or only to SD, that would be way more worrysome.
And also my OpnSense shows nothing dangerous. Only internal IPs and Adguard. But what I can see, thats Home Assistant contact the following ARPA addresses:
xxx.60.0.10.in-addr.arpa
But why do Home Assistant that??? I have in this VLAN only 3 valid addresses.
Thank you for your hint. The sensor data are stored in a MariaDB outside of HA in a separate LXC Proxmox Container. Each container or VM in Proxmox will be backuped each week. And before each Update in HA I always make a backup on my Synology DS918+.
I think it works very well.
I think, I found the problem: I use the Reolink integration and have also a dashboard with picture overview card. We my wife or I use this dashboard, which shows the camera live view, the CPU raises up. Any idea for making it better???
in-addr-arpa addresses are for reverse-DNS, which means IP to hostname lookups.
Some integrations and addons do those lookups and there is nothing dangerous in it.
I think some integrations/addons might be using it as a sort of discovery method.
First my day job used to be a performance engineer.
Are you actually seeing any issues magnifying glass to your chart
That graph shows a machine idling at best. You don’t have an issue. It’s single digit mostly of the time and ten to fifteen in a burst. That’s normal use of an idle box. There was one high pop and it lasted for nearly no time. Unless the machine froze during that one spike
Machines are ‘spiking’ when they jump and sustain utilization at an abnormal level that impacts performance. Your machine is simply responding to a request and handling it and clearing it out
The reason I bring this up is… As a perf guy I don’t see anything here’s concerning… Hell I don’t even concern myself with a machine less than 50% loaded if there’s no apparent latency…
If the box isn’t doing anything weird I’d leave it alone… If you spend your time chasing anomalies like this forever you’ll never actually use the thing. So is it actually acting up?
Also thanks to you, for your help and reply. When I have two cams and the network throughput of 2 MB bringt a 4 core VM in Proxmox up 30-50% of utilization, I like to fix it a little bit.
I completly understand your points of sight, but when I sit here nearby my MiniPC with Proxmox and get some wind by the fan, it isn’t so nice .
Nevertheless I will try to change the configuration a little bit, so the that the cam stream on Dashboard will doesn’t requires so much CPU load.
Thank a lot to all, for helping me and have a nice weekend.
Thanks to all for helping me. I have changed the resolution on dashboard from high definition (2560x1440 // 20FPS // 2048 Bitrate) to standard fluent (640x480 // 15FPS // 512 Bitrate).
For the dashboard perfectly adequate. So my CPU don’t raise up higher than 5% CPU usage. Perfekt for me. If I need a higher resolution, I will find another way, with popup or something.
Always do backups!
The sensor data are stored in a MariaDB outside of HA in a separate LXC Proxmox Container. Each container or VM in Proxmox will be backuped each week. And before each Update in HA I always make a backup on my Synology DS918+.
.