I am trying to setup OVN on a single (non-clustered) host running IncusOS.
The main purpose is to use isolated project-specific networks (features.networks=true) for wich OVN networks seems to be the only option.
I adapted the Incus OVN tutorial to IncusOS using this comment from @stgraber :
Now for OVN on IncusOS, the way this works is that IncusOS runs both OVS and the OVN controller (ovn-host). The control plane (ovn-central) should be run in containers or VMs, possibly on the same cluster, just so long as they’re not themselves running on an OVN network.
Specifically:
- Created an
onv-centralcontainer:
incus launch images:ubuntu/24.04 ovn-central
- Installed the
ovn-centralpackage:
incus exec ovn-central -- apt update
incus exec ovn-central -- apt install ovn-central
- Ensured that southbound and northbound databases listen to connections an all interfaces:
incus exec ovn-central -- vi /etc/default/ovn-central
and pasted:
OVN_CTL_OPTS="--db-nb-addr=0.0.0.0 --db-nb-create-insecure-remote=yes --db-sb-addr=0.0.0.0 --db-sb-create-insecure-remote=yes"
- Configured the OVN service of IncusOS where
<ovn-ceintral-IP>is the IP address of theovn-centralcontainer:
incus admin os service edit ovn
and pasted:
config:
database: tcp:<ovn-ceintral-IP>:6642
enabled: true
tunnel_address: ""
tunnel_protocol: geneve
state: {}
- Configured the communication between Incus and OVN northbound database:
incus config set network.ovn.northbound_connection=tcp:<ovn-ceintral-IP>:6641
The rest is standard:
incus network create ovntest --type=ovn network=UPLINK
incus launch images:debian/13 c1 --network ovntest
The OVN network ovntest is created fine and the container c1 gets network connectivity.
The problem: when the IncusOS host reboots, the shutdown process hangs for 15 minutes. The output of incus admin os debug log does not show many details during this time:
[2026/08/25 15:39:57 WEST] systemd-networkd: incusbr0: Link DOWN
[2026/08/25 15:39:57 WEST] dnsmasq: error binding DHCP socket to device incusbr0
[2026/08/25 15:39:57 WEST] dnsmasq-dhcp: router advertisement on fd42:f806:c351:2734::,
[2026/08/25 15:39:57 WEST] systemd: var-lib-incus-storage\x2dpools-local-containers-c1.mount: Deactivated successfully.
[2026/08/25 15:49:26 WEST] systemd: incus-startup.service: Stopping timed out. Terminating.
[2026/08/25 15:49:26 WEST] systemd: incus-startup.service: Control process exited, code=killed, status=15/TERM
[2026/08/25 15:49:26 WEST] systemd: incus-startup.service: Failed with result 'timeout'.
[2026/08/25 15:49:26 WEST] systemd: Stopped incus-startup.service - Incus - Startup check.
[2026/08/25 15:49:26 WEST] kernel: kauditd_printk_skb: 62 callbacks suppressed
[2026/08/25 15:49:26 WEST] kernel: audit: type=1131 audit(1787669366.635:720): pid=1 uid=0 auid=4294967295 ses=4294967295 subj=unconfined msg='unit=incus-startup comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=? res=failed'
[2026/08/25 15:49:26 WEST] incusd: time="2026-08-25T14:49:26Z" level=warning msg="Ignoring signal, shutdown already in progress" signal=terminated
[2026/08/25 15:49:26 WEST] systemd: Stopping incus.service - Incus - Daemon...
[2026/08/25 15:54:56 WEST] systemd: incus.service: State 'stop-sigterm' timed out. Killing.
[2026/08/25 15:54:56 WEST] systemd: incus.service: Killing process 1454 (incusd) with signal SIGKILL.
[2026/08/25 15:54:56 WEST] systemd: incus.service: Main process exited, code=killed, status=9/KILL
[2026/08/25 15:54:56 WEST] systemd: incus.service: Failed with result 'timeout'.
[2026/08/25 15:54:56 WEST] systemd: incus.service: Unit process 1611 (dnsmasq) remains running after unit stopped.
[2026/08/25 15:54:56 WEST] systemd: Stopped incus.service - Incus - Daemon.
[2026/08/25 15:54:56 WEST] systemd: incus.service: Consumed 9min 410ms CPU time, 1.3G memory peak.
The symptoms are very similar to those described in this post.
The system hangs even if I delete container c1 (but not the ovntest network) and stop the container ovn-central manually before reboot.
My impression is that the ovn services run by IncusOS (ovn-controller) require ovn databases run by ovn-central to be reachable during the shutdown process. This works for the clustered OVN setup where ovn-controller of other cluster members should answer the requests, but for a standalone OVN setup, the OVN databases becomes unreachable once the ovn-central container stops.
Is there any solution to this problem?
