Memory Leak in 2025.10x?

jshep321 · January 10, 2026, 3:15pm

Hi,
I believe have some type of memory leak issue. I’ve been running on a pi5 8GB with PCIe SSD for quite some time. As shown in the graphs, this only started occurring in October 2025. I’m not confident, but I think most/all of the reversions to baseline are due to restarts of the core and/or updates. I have attached the glances screenshot also which shows the OOM issue and that the HA core is the primary culprit, growing slowly but surely.

Currently running 2025.12.5 and HAOS 16.3.

Any ideas on the best way to debug this?

Thanks!

tom_l · January 10, 2026, 9:45pm

Update to the latest version. Is it fixed?

If not, restart in safe mode. Is if fixed?

If yes - then it is a third party integration.

If not then check your config for unterminated loops (repeats in automations or scripts that never end).

jshep321 · January 11, 2026, 5:27pm

Thanks. Per the 2nd paragraph (which I added shortly after the post), I am on the latest version – seeing this issue since 2025.10.x. If you examine the graph timeline, the memory leak is extremely slow – 2-4% daily. Running in safe mode would break the functionality. Most of my automations are in node red, so I think that rules them out. Any other debugging suggestions?

tom_l · January 11, 2026, 7:44pm

Currently 2026.1.0 is the “latest version”.

jshep321 · January 12, 2026, 3:35pm

updated to 2026.1.0. Trendline is the same.

tom_l · January 12, 2026, 8:56pm

Ok that’s step one of what I wrote.

NathanCu · January 12, 2026, 9:34pm

And if you make it through Tom’s list then next is disable half your integrations

If it stops congrats it’s in that half. Reenable half of them. Continue until you identify it.

If the problem persists after you disabled half the it’s not in that half. re-enable the half you originally turned off and disable the other half. Keep chopping your enabled integration set in half until you isolate it.

Time for good old fashioned system detective work now…