KOLLA-OPENSTACK: Getting error "no valid hosts found" error when I try to create a VM in mutinode env

asked 2017-06-22 18:35:52 -0600

hai_senthil gravatar image

updated 2017-06-27 01:10:52 -0600

1.I have installed openstack using kolla-ansible multimode deployment following the steps given in the kolla-openstack documentation (one controller and one compute host) 2.Steps were using bootstrap-servers,kolla-build,kolla-prechecks,kolla-deploy 3.Everything got successfully deployed and openstack containers started running. 4. I was able to see the dashboard of the openstack and login inside 5. In the controller I added cirros image, created a flavor , neutron subnet work, now when I try to build this image in my compute host it says error as ”valid host not found” NOTE: The /etc/host/nova-compute/nova.conf contains

 [libvirt]
 connection_uri = qemu+tcp://10.125.154.15/system

I tried to edit the a the above as per a few suggestions posted on the similar issue. It may be because of dockerisation/. Not sure

Thanks in advance for pointers towards the right solution..

Error in Controller node:

No valid host was found. There are not enough hosts available."  "code": 500, "details": "  File \"/var/lib/kolla/venv/local/lib/python2.7/site-packages/nova/conductor/manager.py\", line 866, in schedule_and_build_instances |
|                                      |     request_specs[0].to_legacy_filter_properties_dict())

IN COMPUTE NODE: Error messaged noticed in the log file:

 tail -f  /var/lib/docker/volumes/kolla_logs/_data/nova/nova-compute.log

2017-06-22 11:41:43.803 7 WARNING nova.scheduler.client.report [req-2f358d19-60d3-4110-a502-1e4d570d70df - - - - -] Placement API service is not responding.^M
2017-06-22 11:41:43.807 7 WARNING nova.scheduler.client.report [req-2f358d19-60d3-4110-a502-1e4d570d70df - - - - -] Placement API service is not responding.^M

+++++ Update +++++

I just followed the instructions.. The version I have is 4.0.2 (10.125.154.35:4000/kolla/ubuntu-source-nova-scheduler:4.0.2) OCTATA branch Yes. Looks like this error message is just Warning. No other useful log info.

I guess qemu since teh nova config files has a reference to it ex: /etc/kolla/nova-api/nova.conf:connection_uri = qemu+tcp://10.125.154.35/system BUT Ihave not installed anything related to qemu since I assumed Kolla should take care of getting all the relevant packages with right versions.? I have not downloaded all the qemu related packages since it may not go well with kolls precheck..

Any further pointers? or ways to further debug? Thanks =Senthil

++++++++++++++++++++++UPDATE+++++++++++++++++++++++ Today I cleaned up all and re-deployed all again. When I tried to create VM and it failed as mentioned above. I could see error messages in nova compute log as given below... I appreciate any help to get the solution..

docker exec -u root -it 752331bfcd9b /bin/bash

And inside the container check for thi log cat /var/log/kolla/nova/nova-compute.log

2017-06-26 15:16:42.791 7 ERROR nova.compute.manager [req-7029f39e-936e-42b6-97cc-6bdbde025a80 - - - - -] No compute node record for host kollacompute1 2017-06-26 15:16:42.798 7 WARNING nova.compute.monitors [req-7029f39e-936e-42b6-97cc-6bdbde025a80 - - - - -] Excluding nova.compute.monitors.cpu monitor virt_driver. Not in the list of enabled monitors (CONF.compute_monitors). 2017-06-26 15:16:43.192 7 WARNING nova.compute.resource_tracker [req-7029f39e-936e-42b6-97cc-6bdbde025a80 - - - - -] No compute node record for kollacompute1:kollacompute1 2017-06-26 15:16:43.219 7 INFO nova.compute.resource_tracker [req-7029f39e-936e-42b6-97cc-6bdbde025a80 - - - - -] Compute_service record created ... (more)

edit retag flag offensive close merge delete

Comments

I don't know how this is configured with Kolla, but if the compute node doesn't get access to the placement service, Nova doesn't know that it has free capacity and therefore claims that there is no host suitable for running the instance.

Bernd Bausch gravatar imageBernd Bausch ( 2017-06-23 00:45:55 -0600 )edit

nova.conf on the compute node contains placement service connection details. They are either incorrect, or a firewall or other network problem prevents the compute node from connecting.

Bernd Bausch gravatar imageBernd Bausch ( 2017-06-23 00:47:47 -0600 )edit

On the other hand, the placement service warning message is normal when nova-compute starts up faster than nova-api on the controller. It may be a temporary situation.

Bernd Bausch gravatar imageBernd Bausch ( 2017-06-23 00:49:47 -0600 )edit

libvirt url is correct, is the way the nova_compute container is able to comunicate with nova_libvirt container. Placement warning is normal during startup, other case is that the error remains. What kolla version are you using, using kvm or qemu for virtualization?

Eduardo Gonzalez gravatar imageEduardo Gonzalez ( 2017-06-23 03:34:43 -0600 )edit

What I mean, are you using kvm for virtualization or qemu for nested virtualization? What is the output of egrep -c '(vmx|svm)' /proc/cpuinfo

If is less than 1 you need to configure nova to use qemu instead of kvm

[libvirt]
virt_type = qemu
cpu_mode = none
Eduardo Gonzalez gravatar imageEduardo Gonzalez ( 2017-06-23 07:20:06 -0600 )edit