Memory leak and memory profiler no-longer working

1

I try to keep relatively current with the HA releases. A couple of weeks ago I did an update and you can see from the following that I now have a memory leak issue.

image

I was hoping maybe one of the more recent updates would fix it, but no go. I then attempted to use the memory profiler via the developer tools screen:

image
image1460×685 30.4 KB

When I first attempted to use the memory profiler I was getting some errors related to openai, so I removed the openai stuff I was playing with under voice assistance. That cleared the openai error, but still doesn’t allow the memory profiler to run. When I start the memory profiler now it simply causes the HA python process to kick to 100% CPU for about 5 minutes, with no memory profile output being generated.

I’m using a supervised install of HA running on Debian bookworm on an odroid N2+ SBC.

Does anyone have input on how to get the memory profiler to run and generate some output. I used it in the past and was able to discover the component with the memory leak. Any suggestion on things to try would be appreciated. Maybe there is a way to get the memory profiler to output some log messages?

Kicked the memory profiler off about 3 minutes ago and here you can see the HA python process pegging a CPU at 100%, with the web interface spinning as it waits for any kind of response back from the HA server.

image
image1797×684 55.7 KB

I’m ready and willing to help debug this HA issue, if I could just use some guidance on how to get HA to spit out something related to the issue.

Turns out with the CPU pegged HA will actually die and restart. Nothing useful in the logs.

At some point in this process this error shows up on the GUI

image
image1396×711 32.9 KB

Using the command

docker logs hassio_supervisor

the entries show where the supervisor can’t communicate with HA core and so it is what ends up restarting HA. No other useful information in that log.

The log entries showing supervisor restarting HA core:

2024-08-29 20:14:34.559 INFO (MainThread) [supervisor.updater] Fetching update data from https://version.home-assistant.io/stable.json
2024-08-29 20:14:43.307 INFO (MainThread) [supervisor.api.proxy] Home Assistant WebSocket API error: Cannot proxy websocket message of unsupported type: 257
2024-08-29 20:14:43.308 INFO (MainThread) [supervisor.api.proxy] Home Assistant WebSocket API for a0d7b954_appdaemon closed
2024-08-29 20:15:13.312 ERROR (MainThread) [supervisor.homeassistant.api] Timeout on call https://172.30.32.1:8123/api/core/state.
2024-08-29 20:15:13.313 WARNING (MainThread) [supervisor.misc.tasks] Watchdog missed an Home Assistant Core API response.
2024-08-29 20:15:19.308 ERROR (MainThread) [supervisor.homeassistant.api] Timeout on call https://172.30.32.1:8123/api/core/state.
2024-08-29 20:15:55.335 ERROR (MainThread) [supervisor.homeassistant.api] Timeout on call https://172.30.32.1:8123/api/core/state.
2024-08-29 20:16:31.333 ERROR (MainThread) [supervisor.homeassistant.api] Timeout on call https://172.30.32.1:8123/api/core/state.
2024-08-29 20:17:07.309 ERROR (MainThread) [supervisor.homeassistant.api] Timeout on call https://172.30.32.1:8123/api/core/state.
2024-08-29 20:17:43.312 ERROR (MainThread) [supervisor.homeassistant.api] Timeout on call https://172.30.32.1:8123/api/core/state.
2024-08-29 20:17:44.305 ERROR (MainThread) [supervisor.homeassistant.api] Timeout on call https://172.30.32.1:8123/api/core/state.
2024-08-29 20:17:44.305 ERROR (MainThread) [supervisor.misc.tasks] Watchdog missed 2 Home Assistant Core API responses in a row. Restarting Home Assistant Core!
2024-08-29 20:17:44.313 INFO (SyncWorker_7) [supervisor.docker.manager] Restarting homeassistant

Sadly the memory just keeps leaking, all the drop downs are restarts.

image

2

As clear by the thread I never received any input on this issue. It seems that a couple versions back someone fixed something in the core as my memory leak has gone away. Here’s a current memory graph. Prior to removing integration in my attempt to look for a problematic integration I did a backup. Then over time I noticed after an HA update a couple weeks back the issue appeared to clear. So restored the version that demonstated the memory leak, with no integration removed and then updated to the latest release. Here’s the memory graph now:

image

The initial flat line is the result of rolling the system back. But as you can see the memory leak appears to have been addressed.

I still can’t run the memory profiler so if anyone knows how to fix that I welcome input.

3

I am also attempting to monitor RAM usage. I am still running on a rpi3 so I assumed oom was the cause, but your post is telling me otherwise.

Profile: Memory won’t work. The process will be restarted by the supervisor the second I send the command.

The start logging objects (and much later, stop), seem to work, tho. Maybe that’s enough…