Sudden high memory (RAM) usage

Configuration
1

Occasionally I encountered high memory (RAM) usage suddenly. It will continue to go up until the whole RPi freezes.

I am wondering is there anyway to find out what causes this? Please help

2

I am seeing this as well recently. Running Hassbian 0.37.1 and my memory usage is around 95%. Logging into the Pi3 can take a while when it is this high and running htop shows that Hass is using most of the memory.

Reboot brings it back down but it starts climbing up again reaching over 90% in a few hours.

Running a few automations but less than 10 and I have 1 sensor attached (DHT22).

I know that the memory is there to be used but it doesn’t free up even when I restart HA.

3

Seems as if its the database file for history and logs that is causing the issue. Deleted home-assistant_v2.db and the memory usage dropped 30%. Will change my conf.yaml to limit the amount of history & logs saved and see if that helps keep it down.

1 Like
4

What are you doing to limit the log file size? Are you creating a script that looks at the file size or are you just configuring the log in a different manor?

5

I delete it when it gets too big :wink:

6

I believe these steps I have taken have solved the high memory issue…

  1. Move the recorder component to external Mysql server and the use of include and exclude options to limit the data that is being recorded.
  2. Limit the logger activities to reduce the log file size.

For example…

logger:
  default: error
  logs:
    homeassistant.core: critical
    homeassistant.components.switch.orvibo: critical

You have to see which components that gave out constant errors but doesn’t really affect anything in your log.

1 Like
7

When you say external, are you saying on another system, I.E. a separate server? Or do you mean another service running on the pi?

8

Separate server. The MYSQL server is in my NAS.

9

I see memory usage peaks every 6 hours with 0.97 (Hass.io on raspberry 3).

Screenshot_20190808-164311
Screenshot_20190808-164311720Ă—1440 38.5 KB

I didn’t find yet what could be the cause. Any clue to investigate? Do you know if there is some process planned at 0, 6, 12, 18 and 24?

My config is on github

10

go through your automations and find out which ones fire every 6 hours

1 Like
11

Thank you. A quick review point out the Speedtest integration and there is a warning on memory usage in the documentation.

12

Another peak is due to InfluxDB add-on. 10% additional memory from my raspberry pi 3 is used at 6:10 AM everyday during 15 minutes. Don’t know why.