[VM] Dynamic RAM allocation going haywire.

[haldi4803]

Describe the issue you are experiencing

I'm running TrueNAS Scale VM with HAOS... i've changed from 6GB RAM to 12GB RAM in settings when the HAOS machine was down.
Now i'm stuck with 95% RAM usage and 90% SWAP usage and have absolutely NO idea why.
Glances doesn't show anything abnormal, htop doesn't show memory usage.
Obviously i tried Rebooting the VM, but didn't change anything.

  ~ free -h
               total        used        free      shared  buff/cache   available
Mem:            11Gi        11Gi       109Mi       4.0Mi       540Mi       526Mi
Swap:          3.9Gi       3.5Gi       363Mi

Using top -b -o +%MEM | head -n 22 in the terminal on the Host does not show anything using too much RAM.
Using free -h says everything is full.

What other way do i have to debug/log RAM usage?

What operating system image do you use?

generic-x86-64 (Generic UEFI capable x86-64 systems)

What version of Home Assistant Operating System is installed?

15.0

Did the problem occur after upgrading the Operating System?

No

Hardware details

TrueNAS Scale: ElectricEel-24.10.2
64GB RAM, 12GB given to VM.
5 Cores 10 Threas given to VM.

Steps to reproduce the issue

No Clue... seriously.

Anything in the Supervisor logs that might be useful for us?

nope, nothing RAM related.
OR should there be?

Anything in the Host logs that might be useful for us?

nope, nothing RAM related.
OR should there be?

System information

System Information

version	core-2025.3.3
installation_type	Home Assistant OS
dev	false
hassio	true
docker	true
user	root
virtualenv	false
python_version	3.13.2
os_name	Linux
os_version	6.12.18-haos
arch	x86_64
timezone	Europe/Zurich
config_dir	/config

Home Assistant Community Store

GitHub API	ok
GitHub Content	ok
GitHub Web	ok
HACS Data	ok
GitHub API Calls Remaining	5000
Installed Version	2.0.5
Stage	running
Available Repositories	1604
Downloaded Repositories	14

Home Assistant Cloud

logged_in	false
can_reach_cert_server	ok
can_reach_cloud_auth	ok
can_reach_cloud	ok

Home Assistant Supervisor

host_os	Home Assistant OS 15.0
update_channel	stable
supervisor_version	supervisor-2025.03.3
agent_version	1.7.2
docker_version	28.0.1
disk_total	33.7 GB
disk_used	24.4 GB
healthy	true
supported	true
host_connectivity	true
supervisor_connectivity	true
ntp_synchronized	true
virtualization	kvm
board	ova
supervisor_api	ok
version_api	ok
installed_addons	Whisper (2.4.0), Piper (1.5.2), ESPHome Device Builder (2025.2.2), openWakeWord (1.10.0), File editor (5.8.0), Advanced SSH & Web Terminal (20.0.2), Studio Code Server (5.18.3), Glances (0.21.1), ESPHome (dev) (dev), InfluxDB (5.0.2)

Dashboards

dashboards	2
resources	9
views	6
mode	storage

Network Configuration

adapters	lo (disabled), enp0s3 (enabled, default, auto), docker0 (disabled), hassio (disabled), vethd7d7598 (disabled), vetha3bda91 (disabled), veth2705565 (disabled), veth891784e (disabled), veth541da13 (disabled), vethf3731d4 (disabled), veth4623118 (disabled), vethec3950d (disabled), vethce4c7ef (disabled), veth1e53068 (disabled), veth1afdad4 (disabled), veth4ed9ef2 (disabled)
ipv4_addresses	lo (127.0.0.1/8), enp0s3 (192.168.1.10/24), docker0 (172.30.232.1/23), hassio (172.30.32.1/23), vethd7d7598 (), vetha3bda91 (), veth2705565 (), veth891784e (), veth541da13 (), vethf3731d4 (), veth4623118 (), vethec3950d (), vethce4c7ef (), veth1e53068 (), veth1afdad4 (), veth4ed9ef2 ()
ipv6_addresses	lo (::1/128), enp0s3 (fddf:3940:83b2::dc7/128, 2001:4060:c00b:dd30::dc7/128, 2001:4060:c00b:dd30:5696:ac91:d59:be47/64, fddf:3940:83b2:0:18d5:ab4b:45b9:2e08/64, fe80::aa8a:2217:6ead:feb2/64), docker0 (fe80::d0e9:1fff:fe7a:36e/64), hassio (fe80::875:2bff:feb2:466f/64), vethd7d7598 (fe80::cc6f:caff:fe80:f940/64), vetha3bda91 (fe80::387a:2fff:fe50:e303/64), veth2705565 (fe80::c4c3:92ff:fe32:6c19/64), veth891784e (fe80::c093:78ff:fe54:f2fd/64), veth541da13 (fe80::c70:7aff:fee8:ef76/64), vethf3731d4 (fe80::888e:99ff:fe96:c291/64), veth4623118 (fe80::4c02:4ff:fe40:cdd1/64), vethec3950d (fe80::fce1:2dff:fe5b:8c81/64), vethce4c7ef (fe80::c4fa:27ff:fe64:1317/64), veth1e53068 (fe80::8411:7aff:fe31:34cf/64), veth1afdad4 (fe80::5491:9eff:fe97:b8f9/64), veth4ed9ef2 (fe80::148b:15ff:fe44:9c17/64)
announce_addresses	192.168.1.10, fddf:3940:83b2::dc7, 2001:4060:c00b:dd30::dc7, 2001:4060:c00b:dd30:5696:ac91:d59:be47, fddf:3940:83b2:0:18d5:ab4b:45b9:2e08, fe80::aa8a:2217:6ead:feb2

Recorder

oldest_recorder_run	8 March 2025 at 08:37
current_recorder_run	19 March 2025 at 19:45
estimated_db_size	251.07 MiB
database_engine	sqlite
database_version	3.48.0

Additional information

Switching Back to 6GB RAM does solve the issue...

But WTF happend here?

Increasing the RAM again to 20GB again shows 90% Full.

[Impact123]

It might be related to ballooning. total only showing 11G also hints towards that. Can you share free -h from your TrueNAS host while this is happening as well?
I'm curious why you allocate so much to HAOS.

[WladyTee]

My personal experience on Intel NUC (No WM, only HAOS) with 4GB.
Since OS 15, I'm experiencing multiple self rebooting, database corruption and slow operations with several integration timeouts.

It seems there are some concurrent tasks that make the system crash.
I'm investigating now, in the beginning I believed into a hardware issue but I have to check and restore the previous backup.

I hope it helps.

[sairon]

It might be related to ballooning.

That might be a good lead. I saw something similar in Proxmox when "Minimum memory" was different than the "Memory" configured, when the VM started to eat more memory, it went haywire when the allocation changed. With fixed allocation, the problem went away.

[haldi4803]

I'm curious why you allocate so much to HAOS.

i was running Whisper with the tiny-int8 Model and it couldnt grasp a single sentence correct. thats why i thought use a bigger model. but then RAM was not enough.

Ballooning might be a good argument!
I set a minimum of 4gb and increased only the maximum.

[haldi4803]

Yeah.
Would you look at that?
I've set RAM fixed to 12 GiB and it works!

takes off pirate hat let's change the title accordingly...