[kolla-ansible] Single-node deployment with 2026.1 and RockyLinux10 breaks due to nova_libvirt error: "Waiting for nova-compute services to register themselves"
Hi all, I am just trying another single-node deployment with kolla-ansible 2026.1 on RockyLinux 10.1. Still getting the hang of kolla-ansible and OpenStack... Unfortunately the deployment breaks due to the nova compute never registering. The nova_compute container is shown as "healthy", but frequently restarts. And has lots of errors in the logs:
2026-05-17 19:26:36.425 2 ERROR nova.service 2026-05-17 19:26:36.429 2 INFO nova.virt.libvirt.driver [None req-a5a83d7f-19a7-4fec-abc6-c6df7111869b - - - - - -] Connection event '0' reason 'Failed to connect to libvirt: authentication failed: authentication failed' 2026-05-17 19:26:36.934 2 WARNING nova.virt.libvirt.driver [None req-a5a83d7f-19a7-4fec-abc6-c6df7111869b - - - - - -] Cannot update service status on host "openstack-rocky02" since it is not registered.: nova.exception_Remote.ComputeHostNotFound_Remote : Compute host openstack-rocky02 could not be found. 2026-05-17 19:26:36.935 2 WARNING nova.virt.libvirt.volume.mount [None req-a5a83d7f-19a7-4fec-abc6-c6df7111869b - - - - - -] host_down called, but we don't think host is up
The nova_libvirt container spits out errors every 3 minutes:
2026-05-17 17:23:36.084+0000: 151648: info : libvirt version: 11.5.0, package: 4.8.el10_1 (Rocky Linux Build System <releng@rockylinux.org>, 2026-03-17-19:29:48, ) 2026-05-17 17:23:36.084+0000: 151648: info : hostname: openstack-rocky02 2026-05-17 17:23:36.084+0000: 151648: error : virNetSASLSessionServerStep:599 : authentication failed: Failed to start SASL negotiation: -20 (SASL(-13): user not found: unable to canonify user and get auxprops) 2026-05-17 17:23:36.084+0000: 151648: error : remoteDispatchAuthSaslStep:4008 : authentication failed: authentication failed 2026-05-17 17:23:36.085+0000: 150292: error : virNetSocketReadWire:1782 : End of file while reading data: Input/output error
I looked through the kolla-ansible bug reports but found no match. I also tried searching the web, but also no results (at least in the kolla-ansible context). Any idea where to start digging (or even how to fix this)? Kind Regards, Johannes
Hi, I faced the same issue and didn't want to dig too deep for a test environment, so I disabled sasl: $ grep sasl /etc/kolla/globals.yml libvirt_enable_sasl: "no" This lets the compute nodes register, but of course does not fix the underlying issue. If I have the time I will investigate what I need to do in my lab environment to enable sasl. Regards, Eugen Zitat von Johannes Kastl <mail@ojkastl.de>:
Hi all,
I am just trying another single-node deployment with kolla-ansible 2026.1 on RockyLinux 10.1. Still getting the hang of kolla-ansible and OpenStack...
Unfortunately the deployment breaks due to the nova compute never registering. The nova_compute container is shown as "healthy", but frequently restarts. And has lots of errors in the logs:
2026-05-17 19:26:36.425 2 ERROR nova.service 2026-05-17 19:26:36.429 2 INFO nova.virt.libvirt.driver [None req-a5a83d7f-19a7-4fec-abc6-c6df7111869b - - - - - -] Connection event '0' reason 'Failed to connect to libvirt: authentication failed: authentication failed' 2026-05-17 19:26:36.934 2 WARNING nova.virt.libvirt.driver [None req-a5a83d7f-19a7-4fec-abc6-c6df7111869b - - - - - -] Cannot update service status on host "openstack-rocky02" since it is not registered.: nova.exception_Remote.ComputeHostNotFound_Remote : Compute host openstack-rocky02 could not be found. 2026-05-17 19:26:36.935 2 WARNING nova.virt.libvirt.volume.mount [None req-a5a83d7f-19a7-4fec-abc6-c6df7111869b - - - - - -] host_down called, but we don't think host is up
The nova_libvirt container spits out errors every 3 minutes:
2026-05-17 17:23:36.084+0000: 151648: info : libvirt version: 11.5.0, package: 4.8.el10_1 (Rocky Linux Build System <releng@rockylinux.org>, 2026-03-17-19:29:48, ) 2026-05-17 17:23:36.084+0000: 151648: info : hostname: openstack-rocky02 2026-05-17 17:23:36.084+0000: 151648: error : virNetSASLSessionServerStep:599 : authentication failed: Failed to start SASL negotiation: -20 (SASL(-13): user not found: unable to canonify user and get auxprops) 2026-05-17 17:23:36.084+0000: 151648: error : remoteDispatchAuthSaslStep:4008 : authentication failed: authentication failed 2026-05-17 17:23:36.085+0000: 150292: error : virNetSocketReadWire:1782 : End of file while reading data: Input/output error
I looked through the kolla-ansible bug reports but found no match. I also tried searching the web, but also no results (at least in the kolla-ansible context).
Any idea where to start digging (or even how to fix this)?
Kind Regards, Johannes
Hi, On 17.05.26 20:01 Eugen Block wrote:
I faced the same issue and didn't want to dig too deep for a test environment, so I disabled sasl:
$ grep sasl /etc/kolla/globals.yml libvirt_enable_sasl: "no"
Hmm, this does not exist in my globals.yml, but I'll set it and try again. Apparently this is set as default in two of the roles underneath:
venv_kolla-ansible_GIT_2026.1/share/kolla-ansible/ansible/roles/nova-cell/defaults/main.yml:libvirt_enable_sasl: true venv_kolla-ansible_GIT_2026.1/share/kolla-ansible/ansible/roles/masakari/defaults/main.yml:libvirt_enable_sasl: true
Kind Regards, Johannes
I've seen this error before:
virNetSASLSessionServerStep:599 : authentication failed: Failed to start SASL negotiation: -20 (SASL(-13): user not found: unable to canonify user and get auxprops)
It can be that your sasl user is "username@nodename" but your node's hostname is "username@nodename.fqdn" which then causes the "user not found" error. You can confirm this is the case by getting into your libvirt container on your host: `docker exec -it nova_libvirt bash` and running: `hostname -f` And then checking the sasl database `sasldblistusers2 -f /etc/libvirt/passwd.db` If the output of the sasl db check username, and the hostname output by hostname -f does not match exactly, then you will have your user problems there. There is a gerrit issue about this somewhere; The root cause (if this is the fault for you) will be the configuration of your /etc/hosts file. The entry for your /etc/hosts in your compute node needs to have the short-form hostname first, and then the long form hostname (fqdn) Hopefully that helps. Kind Regards, Joel McLean Cyber Security and Product Development Manager Australia’s First Tier IV Data Centre https://www.micron21.com/ 1300 769 972 03 9751 7618 0407 888 429 joel.mclean@micron21.com -----Original Message----- From: Johannes Kastl <mail@ojkastl.de> Sent: Monday, 18 May 2026 4:50 AM To: openstack-discuss@lists.openstack.org Subject: Re: [kolla-ansible] Single-node deployment with 2026.1 and RockyLinux10 breaks due to nova_libvirt error: "Waiting for nova-compute services to register themselves" Hi, On 17.05.26 20:01 Eugen Block wrote:
I faced the same issue and didn't want to dig too deep for a test environment, so I disabled sasl:
$ grep sasl /etc/kolla/globals.yml libvirt_enable_sasl: "no"
Hmm, this does not exist in my globals.yml, but I'll set it and try again. Apparently this is set as default in two of the roles underneath:
venv_kolla-ansible_GIT_2026.1/share/kolla-ansible/ansible/roles/nova-c ell/defaults/main.yml:libvirt_enable_sasl: true venv_kolla-ansible_GIT_2026.1/share/kolla-ansible/ansible/roles/masaka ri/defaults/main.yml:libvirt_enable_sasl: true
Kind Regards, Johannes
Oh that was very helpful! I had a different issue (failed to bind port) during setup with the compute nodes, the ovn chassis also contained the fqdn instead of shortname. So after fixing /etc/hosts the port binding was successfull, but I didn't make the connection to libvirt. I just reconfigured with sasl enabled and now it works. Thanks a lot! Zitat von Joel McLean <joel.mclean@micron21.com>:
I've seen this error before:
virNetSASLSessionServerStep:599 : authentication failed: Failed to start SASL negotiation: -20 (SASL(-13): user not found: unable to canonify user and get auxprops)
It can be that your sasl user is "username@nodename" but your node's hostname is "username@nodename.fqdn" which then causes the "user not found" error.
You can confirm this is the case by getting into your libvirt container on your host: `docker exec -it nova_libvirt bash` and running: `hostname -f` And then checking the sasl database `sasldblistusers2 -f /etc/libvirt/passwd.db`
If the output of the sasl db check username, and the hostname output by hostname -f does not match exactly, then you will have your user problems there. There is a gerrit issue about this somewhere;
The root cause (if this is the fault for you) will be the configuration of your /etc/hosts file. The entry for your /etc/hosts in your compute node needs to have the short-form hostname first, and then the long form hostname (fqdn)
Hopefully that helps.
Kind Regards, Joel McLean Cyber Security and Product Development Manager Australia’s First Tier IV Data Centre https://www.micron21.com/
1300 769 972 03 9751 7618 0407 888 429 joel.mclean@micron21.com
-----Original Message----- From: Johannes Kastl <mail@ojkastl.de> Sent: Monday, 18 May 2026 4:50 AM To: openstack-discuss@lists.openstack.org Subject: Re: [kolla-ansible] Single-node deployment with 2026.1 and RockyLinux10 breaks due to nova_libvirt error: "Waiting for nova-compute services to register themselves"
Hi,
On 17.05.26 20:01 Eugen Block wrote:
I faced the same issue and didn't want to dig too deep for a test environment, so I disabled sasl:
$ grep sasl /etc/kolla/globals.yml libvirt_enable_sasl: "no"
Hmm, this does not exist in my globals.yml, but I'll set it and try again.
Apparently this is set as default in two of the roles underneath:
venv_kolla-ansible_GIT_2026.1/share/kolla-ansible/ansible/roles/nova-c ell/defaults/main.yml:libvirt_enable_sasl: true venv_kolla-ansible_GIT_2026.1/share/kolla-ansible/ansible/roles/masaka ri/defaults/main.yml:libvirt_enable_sasl: true
Kind Regards, Johannes
Hi Joel, On 18.05.26 01:32 Joel McLean wrote:
virNetSASLSessionServerStep:599 : authentication failed: Failed to start SASL negotiation: -20 (SASL(-13): user not found: unable to canonify user and get auxprops)
It can be that your sasl user is "username@nodename" but your node's hostname is "username@nodename.fqdn" which then causes the "user not found" error.
Ouch.
The root cause (if this is the fault for you) will be the configuration of your /etc/hosts file. The entry for your /etc/hosts in your compute node needs to have the short-form hostname first, and then the long form hostname (fqdn)
Which unfortunately is not in line with what I have in all of my host's /etc/hosts files. The manpage does not mention order, but all examples in the manpage have the FQDN first and the short name second. I set this up like hundreds of years ago, but rechecked the manpage and found this still being the only clue on the order of entries.
Hopefully that helps. Yes, it does. I tested this and with the short hostname first I do not get this error during deployment.
I also found that changing the /etc/hosts file on a machine already having a running kolla-ansible installation leads to the hypervisor not being found / being down after a reboot. :-) Kind Regards, Johannes
participants (3)
-
Eugen Block
-
Joel McLean
-
Johannes Kastl