Hi,
I’m trying to determine whether this is expected behavior or a regression. My HAOS VM (KVM/libvirt on Archlinux host) used to shut down gracefully in about 20 seconds (virsh shutdown). It now consistently takes about 1m45s .
To investigate, I restored my recent VM backups one by one to check and compare old and new behaviour. Result is that:
Backup from 2026-05-30 : shutdown in ~20 seconds.
Core: 2026.5.4
Supervisor: 2026.05.1
HAOS: 17.3
Docker: 29.3.1
Backup from 2026-06-20 : shutdown in ~1m45s.
Core: 2026.6.4
Supervisor: 2026.06.2
HAOS: 18.0
Docker: 29.5.3
I also repeated the test with the VM completely disconnected from the network to rule out MQTT/cloud/integration delays. The timings were unchanged.
During shutdown, the serial console repeatedly shows:
A stop job is running for HAOS supervisor
while overlay mounts continue to be unmounted until the VM eventually powers off.
I noticed that Supervisor 2026.06 introduced changes related to graceful shutdown, so I’m wondering whether this could be expected behavior.
Has anyone observed a similar increase in shutdown time after upgrading around HAOS 18 / Supervisor 2026.06?
terrypen
July 12, 2026, 8:18pm
2
I am seeing a similar issue, without doing all the testing! It is taking a long time to stop a job. It eventually continues the reboot, but seems to hang for a while.
kero77
July 13, 2026, 7:21am
3
Thank you for your answer. At least, I’m not the only one.
I also posted on github’s HAOS issue tracker to see what happens:
opened 06:13AM - 13 Jul 26 UTC
Hello,
I'm not sure whether this is a regression or an expected consequence of … recent shutdown changes, but I wanted to report my observations because I’m trying to understand whether the problem is specific to my setup.
### Environment
- Home Assistant OS running as a KVM/QEMU virtual machine (libvirt) on an Archlinux host
- qcow2 disk image
- Graceful shutdown initiated with: `virsh shutdown haos`
Shutdown time is measured from the moment `virsh shutdown haos` is issued until `virsh domstate` reports `shut off`.
### Investigation
I noticed that my VM, which historically shut down in about **20 seconds**, now consistently requires about **1 minute 46 seconds**.
To identify when the behavior changed, I restored historical backups of the VM and measured the shutdown time for each restored image. The results were:
* **Backup from 2026-05-30**: shutdown in ~20 seconds.
* Core: 2026.5.4
* Supervisor: 2026.05.1
* HAOS: 17.3
* Docker: 29.3.1
* **Backup from 2026-06-20**: shutdown in ~1m45s.
* Core: 2026.6.4
* Supervisor: 2026.06.2
* HAOS: 18.0
* Docker: 29.5.3
Therefore, the change appeared somewhere between these two snapshots. Note that the same slow shutdown behaviour is still present on my current HAOS 18.1 installation (with core: 2026.7.1 and supervisor: 2026.06.2).
### Additional observations
To eliminate external influences, I repeated the tests with the VM completely disconnected from the network (virtual NIC link disabled before boot). The measured shutdown times remained basically unchanged (slightly shorter, about 3-4 seconds). This suggests that the additional delay is not caused by MQTT, cloud services, integrations or other network activity.
### Shutdown sequence
The shutdown completes successfully and the VM is never forcibly powered off. Only the duration has changed significantly.
During the longer shutdown, the serial console repeatedly displays:
`A stop job is running for HAOS supervisor`
At the same time, Docker overlay mounts continue to be unmounted progressively until the shutdown completes. The system therefore appears to be making progress rather than hanging completely.
### Why I'm reporting this
While investigating, I noticed that both HAOS 18.0 and Supervisor 2026.06 introduced changes related to the shutdown process.
I'm *not* claiming this is necessarily a bug.
My question is simply:
- Is a graceful shutdown time of around **1m45** now expected for HAOS 18?
- Or does this suggest a regression or an interaction specific to my installation?
I also asked on the Home Assistant Community forum and received at least one independent report from another user experiencing a similar delay, although without the same level of investigation:
https://community.home-assistant.io/t/shutdown-time-increased-since-recent-haos-supervisor-update/1016801
I'd be happy to perform additional tests if they could help identify the cause.
### 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?
18.1
### Did the problem occur after upgrading the Operating System?
Yes
### Hardware details
HAOS runs as a KVM/libvirt VM (qcow2 file) on an Archlinux system, hosted on an Asus NUC 14 Pro (Intel Core Ultra 5 125H).
### Steps to reproduce the issue
1. Run `virsh shutdown haos` on the libvirt host.
2. Poll the state with `virsh domstate haos`.
3. Measure the time until the state becomes `shut off`.
### Anything in the Supervisor logs that might be useful for us?
```txt
No obvious errors were found in the Supervisor logs. The shutdown completes successfully. The issue is not a failure but a significant increase in graceful shutdown duration. I can provide the complete Supervisor logs if requested.
```
### Anything in the Host logs that might be useful for us?
```txt
No obvious errors were found in the host logs. I can provide the full logs if requested.
```
### System information
_No response_
### Additional information
_No response_