im running a Raspi 3b with some load peaks every 1.5h at “CPU usage”, “load (1m)”, “load (5m)” and “load (15m)”, all simultaneously.
the CPU isn’t the problem i guess. Normally its less than 10% and those peaks reach 30% ish.
The load (1m) is normally around 0.1-0.7 but at the peaks, it reaches up to 7.
The SD card has a large impact on the load. With a Samsung Pro 64GB the load (1m) peaks reach up to 30.
At the moment im using a Sandisk Ultra 32GB
Before i was using a Sandisk Endurance Max 64GB for half a year until my HA slowly startet to lose performance (i don’t think the SD is dying, i guess its the growing config). The peaks reached 10 ish.
The “CPU usage” don’t change with different SD cards.
Now i bought a Sandisk Extreme Pro 128GB (its Application Class 2 like recommended by HomeAssistant) which will arrive soon. Even if the new SD card don’t show that peaks, the problem is not gone.
So i still want to know which process causes that peaks? What process is executed every 1.5h?
Does anybody know? Or can give me a tip on what i can check?
I wonder why people always say that. For my understanding thats only true if you using one
-without wear leveling (most newer cards have this feature)
-with less write cycles
-which is to small
But for example my 64GB card had at leased 50GB free space.
Lets say HA recorder stores each 5 sec (default setting) 1MB of sensor data.
After 1 day 17GB is written. After 3 days the whole 50GB free space is used once.
For Application Class 2 SD cards you can repeat that 2000 times.
→ 6000 days → 16 years
SSH to the machine and run the command “Top” and leave it running. It will show the current running tasks and the amount of processor and memory use by task… My experience with other Linux systems is that iowait can be a killer as you have diagnosed already. So top will confirm if that is the culprit.
Sorry for the delayed reply.
I tried the SSH command “Top”. But since my problem is not continuous, i didnt manage it to see a the values during a load peak.
I wanted to have the values written to a log file on a regular interval, but I wasn’t successful on the first try.
To be honest, I have now given up on the problem and bought a tiny PC to run HA on (as a proxmox VM). As expected, my problem is solved. Load_1m is 0.1 on average now. Highest peak was 0.8 since 2 days on the new HW.