Title: HAOS crashes with OOM – possible Mosquitto/AWTRIX trigger

Hi,

I’m trying to diagnose recurring Home Assistant crashes/freezes, possibly related to MQTT/AWTRIX.

System: Raspberry Pi 3 (64-bit), HAOS 18.2, Core 2026.6.4, ~925 MB RAM + 1 GB swap, Mosquitto add-on, Ulanzi/AWTRIX via MQTT.

The system sometimes becomes unresponsive within a few minutes until HA, SSH and even the Observer are no longer reachable and I have to power-cycle the Pi.

The HAOS kernel log confirms global OOM events, e.g.:

oom-kill: … global_oom … task=mosquitto
Out of memory: Killed process 4712 (mosquitto)

About 10 minutes later another OOM killed ttyd. I understand that Mosquitto being killed does not necessarily mean it caused the OOM.

The interesting part: After stopping Mosquitto completely, HA ran for more than 24 hours without a crash. Before that, crashes occurred repeatedly.

I use several AWTRIX MQTT automations. A complex HVV/public transport automation is currently my main suspect. I’m now testing with Mosquitto enabled but HVV and another complex sports automation disabled, then I’ll enable them one by one.

My questions:

  1. Has anyone seen AWTRIX/MQTT automations cause sudden memory spikes or message buildup leading to OOM?
  2. Is there a way in HAOS to record per-container/process memory usage shortly before an OOM? System Monitor only shows total RAM usage, and the SSH add-on doesn’t provide normal Docker access.

Thanks for any ideas!

A pi3 with the 1gb of memory is no longer recommended to run current versions of HA. Not enough memory or cpu, and when you get it to run it will be unstable and prone to issues like you are explaining.

This ^^

The mqtt issue is a symptom, not a cause. You’d have had the same OOM errors with a different integration.

Upgrade your pi3. 1gb is no longer sufficient nor recommended for HA. You need 2gb at a bare minimum