[nova][openstack-ansible] "No Valid Host Found" when running on VirtualBox
Hello, I'm trying to run a test/learning cluster on VirtualBox VMs deployed with Openstack-Ansible prior to deploying on real hardware. The playbooks succeed, and the basic functional checks all look good. However, I am unable to deploy an instance due to 'No Valid Host Found'. I have seen that this might be because VirtualBox does not support KVM. But I am also perplexed - I am able to deploy instances on an AIO running in VirtualBox on the same physical host. If there is an easy fix for this please point me in the right direction. If I need to destroy and re-clone my VMs I can try that as well. In an effort to learn, though, I'd really like to learn how to adjust the current deployment. Thanks. -Dave -- Dave Hall Binghamton University kdhall@binghamton.edu
Hey, One question - have you set nova_virt_type to qemu in overrides? As otherwise you indeed need to have nested virtualization enabled. чт, 9 июн. 2022 г., 20:21 Dave Hall <kdhall@binghamton.edu>:
Hello,
I'm trying to run a test/learning cluster on VirtualBox VMs deployed with Openstack-Ansible prior to deploying on real hardware.
The playbooks succeed, and the basic functional checks all look good. However, I am unable to deploy an instance due to 'No Valid Host Found'. I have seen that this might be because VirtualBox does not support KVM. But I am also perplexed - I am able to deploy instances on an AIO running in VirtualBox on the same physical host.
If there is an easy fix for this please point me in the right direction. If I need to destroy and re-clone my VMs I can try that as well. In an effort to learn, though, I'd really like to learn how to adjust the current deployment.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
Dmitry, I do have 'nested VT-x/AMD-V' turned on on all of my VirtualBox VMs, but not 'PAE/NX'. Also, I have installed the VBox Guest Additions, but I disabled vboxadd-service.service due to clock synchronization issues. Lastly, I added nova_virt_type: qemu to /etc/openstack_deploy/user_variables.yml and re-ran the playbooks. If re-running the playbooks won't flip this setting, is there some other approach, like destroying the LXC containers for nova and then re-run the playbooks? Add another compute node? Side note: I also added # rsyslog server log_hosts: log1: ip: 172.29.236.14 to openstack_user_config.yml. After doing so, 'openstack-ansible setup-hosts' hangs during a task related to rsyslog, but the other two playbooks complete successfully. Thanks. -Dave -- Dave Hall Binghamton University kdhall@binghamton.edu On Thu, Jun 9, 2022 at 3:18 PM Dmitriy Rabotyagov <noonedeadpunk@gmail.com> wrote:
Hey,
One question - have you set nova_virt_type to qemu in overrides? As otherwise you indeed need to have nested virtualization enabled.
чт, 9 июн. 2022 г., 20:21 Dave Hall <kdhall@binghamton.edu>:
Hello,
I'm trying to run a test/learning cluster on VirtualBox VMs deployed with Openstack-Ansible prior to deploying on real hardware.
The playbooks succeed, and the basic functional checks all look good. However, I am unable to deploy an instance due to 'No Valid Host Found'. I have seen that this might be because VirtualBox does not support KVM. But I am also perplexed - I am able to deploy instances on an AIO running in VirtualBox on the same physical host.
If there is an easy fix for this please point me in the right direction. If I need to destroy and re-clone my VMs I can try that as well. In an effort to learn, though, I'd really like to learn how to adjust the current deployment.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
The only playbook needs rerunning is os-nova-install.yml. And it should target "compute" VM where nova-compute service runs. There's no reason to destroy containers to apply setting, but yes, we have playbooks for destroying/recreating containers . We actually don't use rsyslog as of today. We should likely fix documentation to reflect that. All loges are stored in journald. What does `openstack compute service list` return as a result? Is there any nova-compute in the list and it's alive? if yes, you should check logs for nova-scheduler in nova_api container and placement in it's container. Eventually logs are exposed to host also but can't recall from top of my head the command to check from host efficiently. пт, 10 июн. 2022 г., 19:04 Dave Hall <kdhall@binghamton.edu>:
Dmitry,
I do have 'nested VT-x/AMD-V' turned on on all of my VirtualBox VMs, but not 'PAE/NX'. Also, I have installed the VBox Guest Additions, but I disabled vboxadd-service.service due to clock synchronization issues.
Lastly, I added
nova_virt_type: qemu
to /etc/openstack_deploy/user_variables.yml and re-ran the playbooks. If re-running the playbooks won't flip this setting, is there some other approach, like destroying the LXC containers for nova and then re-run the playbooks? Add another compute node?
Side note: I also added
# rsyslog server log_hosts: log1: ip: 172.29.236.14
to openstack_user_config.yml. After doing so, 'openstack-ansible setup-hosts' hangs during a task related to rsyslog, but the other two playbooks complete successfully.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
On Thu, Jun 9, 2022 at 3:18 PM Dmitriy Rabotyagov <noonedeadpunk@gmail.com> wrote:
Hey,
One question - have you set nova_virt_type to qemu in overrides? As otherwise you indeed need to have nested virtualization enabled.
чт, 9 июн. 2022 г., 20:21 Dave Hall <kdhall@binghamton.edu>:
Hello,
I'm trying to run a test/learning cluster on VirtualBox VMs deployed with Openstack-Ansible prior to deploying on real hardware.
The playbooks succeed, and the basic functional checks all look good. However, I am unable to deploy an instance due to 'No Valid Host Found'. I have seen that this might be because VirtualBox does not support KVM. But I am also perplexed - I am able to deploy instances on an AIO running in VirtualBox on the same physical host.
If there is an easy fix for this please point me in the right direction. If I need to destroy and re-clone my VMs I can try that as well. In an effort to learn, though, I'd really like to learn how to adjust the current deployment.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
Dmitry, I reran os-nova-install.yml. I observed from the dashboard that my compute node is listed as running QEMU. `openstack compute service list` shows my compute node as up and enabled. I think I did find a way to get past 'No Valid Host'. But then the instance failed with a volume error. I think maybe I don't have all of the needed pieces in place - networks, ports, security groups, etc. Is there a tutorial that you could point me to? Thanks. -Dave -- Dave Hall Binghamton University kdhall@binghamton.edu On Fri, Jun 10, 2022 at 1:28 PM Dmitriy Rabotyagov <noonedeadpunk@gmail.com> wrote:
The only playbook needs rerunning is os-nova-install.yml. And it should target "compute" VM where nova-compute service runs. There's no reason to destroy containers to apply setting, but yes, we have playbooks for destroying/recreating containers .
We actually don't use rsyslog as of today. We should likely fix documentation to reflect that. All loges are stored in journald.
What does `openstack compute service list` return as a result? Is there any nova-compute in the list and it's alive?
if yes, you should check logs for nova-scheduler in nova_api container and placement in it's container. Eventually logs are exposed to host also but can't recall from top of my head the command to check from host efficiently.
пт, 10 июн. 2022 г., 19:04 Dave Hall <kdhall@binghamton.edu>:
Dmitry,
I do have 'nested VT-x/AMD-V' turned on on all of my VirtualBox VMs, but not 'PAE/NX'. Also, I have installed the VBox Guest Additions, but I disabled vboxadd-service.service due to clock synchronization issues.
Lastly, I added
nova_virt_type: qemu
to /etc/openstack_deploy/user_variables.yml and re-ran the playbooks. If re-running the playbooks won't flip this setting, is there some other approach, like destroying the LXC containers for nova and then re-run the playbooks? Add another compute node?
Side note: I also added
# rsyslog server log_hosts: log1: ip: 172.29.236.14
to openstack_user_config.yml. After doing so, 'openstack-ansible setup-hosts' hangs during a task related to rsyslog, but the other two playbooks complete successfully.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
On Thu, Jun 9, 2022 at 3:18 PM Dmitriy Rabotyagov < noonedeadpunk@gmail.com> wrote:
Hey,
One question - have you set nova_virt_type to qemu in overrides? As otherwise you indeed need to have nested virtualization enabled.
чт, 9 июн. 2022 г., 20:21 Dave Hall <kdhall@binghamton.edu>:
Hello,
I'm trying to run a test/learning cluster on VirtualBox VMs deployed with Openstack-Ansible prior to deploying on real hardware.
The playbooks succeed, and the basic functional checks all look good. However, I am unable to deploy an instance due to 'No Valid Host Found'. I have seen that this might be because VirtualBox does not support KVM. But I am also perplexed - I am able to deploy instances on an AIO running in VirtualBox on the same physical host.
If there is an easy fix for this please point me in the right direction. If I need to destroy and re-clone my VMs I can try that as well. In an effort to learn, though, I'd really like to learn how to adjust the current deployment.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
What's the volume error? What does 'openstack volume show <uuid>' show for status? For networking good set of useful commands you can find here https://docs.openstack.org/neutron/yoga/admin/archives/use.html But all really depends on setup, as openstack is very flexible so it all depends. вс, 12 июн. 2022 г., 00:28 Dave Hall <kdhall@binghamton.edu>:
Dmitry,
I reran os-nova-install.yml. I observed from the dashboard that my compute node is listed as running QEMU.
`openstack compute service list` shows my compute node as up and enabled.
I think I did find a way to get past 'No Valid Host'. But then the instance failed with a volume error.
I think maybe I don't have all of the needed pieces in place - networks, ports, security groups, etc. Is there a tutorial that you could point me to?
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
On Fri, Jun 10, 2022 at 1:28 PM Dmitriy Rabotyagov < noonedeadpunk@gmail.com> wrote:
The only playbook needs rerunning is os-nova-install.yml. And it should target "compute" VM where nova-compute service runs. There's no reason to destroy containers to apply setting, but yes, we have playbooks for destroying/recreating containers .
We actually don't use rsyslog as of today. We should likely fix documentation to reflect that. All loges are stored in journald.
What does `openstack compute service list` return as a result? Is there any nova-compute in the list and it's alive?
if yes, you should check logs for nova-scheduler in nova_api container and placement in it's container. Eventually logs are exposed to host also but can't recall from top of my head the command to check from host efficiently.
пт, 10 июн. 2022 г., 19:04 Dave Hall <kdhall@binghamton.edu>:
Dmitry,
I do have 'nested VT-x/AMD-V' turned on on all of my VirtualBox VMs, but not 'PAE/NX'. Also, I have installed the VBox Guest Additions, but I disabled vboxadd-service.service due to clock synchronization issues.
Lastly, I added
nova_virt_type: qemu
to /etc/openstack_deploy/user_variables.yml and re-ran the playbooks. If re-running the playbooks won't flip this setting, is there some other approach, like destroying the LXC containers for nova and then re-run the playbooks? Add another compute node?
Side note: I also added
# rsyslog server log_hosts: log1: ip: 172.29.236.14
to openstack_user_config.yml. After doing so, 'openstack-ansible setup-hosts' hangs during a task related to rsyslog, but the other two playbooks complete successfully.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
On Thu, Jun 9, 2022 at 3:18 PM Dmitriy Rabotyagov < noonedeadpunk@gmail.com> wrote:
Hey,
One question - have you set nova_virt_type to qemu in overrides? As otherwise you indeed need to have nested virtualization enabled.
чт, 9 июн. 2022 г., 20:21 Dave Hall <kdhall@binghamton.edu>:
Hello,
I'm trying to run a test/learning cluster on VirtualBox VMs deployed with Openstack-Ansible prior to deploying on real hardware.
The playbooks succeed, and the basic functional checks all look good. However, I am unable to deploy an instance due to 'No Valid Host Found'. I have seen that this might be because VirtualBox does not support KVM. But I am also perplexed - I am able to deploy instances on an AIO running in VirtualBox on the same physical host.
If there is an easy fix for this please point me in the right direction. If I need to destroy and re-clone my VMs I can try that as well. In an effort to learn, though, I'd really like to learn how to adjust the current deployment.
Thanks.
-Dave
-- Dave Hall Binghamton University kdhall@binghamton.edu
participants (2)
-
Dave Hall
-
Dmitriy Rabotyagov