Open altserg opened 10 months ago
Looks like the issue is related to https://github.com/aio-libs/aiohttp/issues/7252 Flag enable_cleanup_closed=True in iohttp.TCPConnector. When I removed the flag in init.py, the memory consumption is not increasing so rapidly. I'm testing now, will report the result. However, if this works, should some Python version check and quick be implemented in init.py ? Note that in the linked issue people are saying that the fix should be implemented in Python 3.11.4, I had the memory leak in self-compiled later Python 3.11 versions. Debian bookworm has Python 3.11.2, so the quirk has to be implemented there anyway. This is what HASS implements aiohttp_client.py:ENABLE_CLEANUP_CLOSED = not (3, 11, 1) <= sys.version_info < (3, 11, 4) aiohttp_client.py: enable_cleanup_closed=ENABLE_CLEANUP_CLOSED,
So far so good with enable_cleanup_closed=False
Hi all,
I'm using HASS Core on RPI, and I'm moving to the latest Debian OS version. I have noticed that with Python 3.11.2 (also with Python 3.11 manually compiled on previous RPI OS version), the memory usage raises very quickly when Dahua integration is enabled. Refer to attached memory usage log with peaks at the end corresponding to enablement of Dahua integration and HASS restart
I'm wondering if anyone else has noticed something similar, and I would appreciate any help in debugging the issue - how to track all memory allocations in Dahua component ? Or maybe it's something with asyncio that goes wrong ?