Openstack Failed to launch Instances [Error: No Valid Host was found]

2
votes

I have tried to install openstack on Centos 7.
Below are the configurations:

  1. Neutron Node (VM with 2 Vcpu, 2 GB RAM , 3 NICs)
  2. Controller Node(VM with 2 Vcpu, 8 GB RAM, 1 NIC)
  3. Compute Node(Physical Machine 24 CPU, 64 GB RAM, 2 NIC)
    # egrep -c '(vmx|svm)' /proc/cpuinfo gives an output of 24

I have followed the online documentation: OPENSTACK INSTALLATION GUIDE FOR RED HAT ENTERPRISE LINUX 7, CENTOS 7, AND FEDORA 20 - JUNO (Install and configure network node) line by line and verified all configurations 3 times.

Now whenever I try to launch instances using the Horizon dashboard, I get this error No Valid Host was found!!

Nova-compute Log gives me this error

Nova-compute.log

2015-08-21 22:55:00.391 41235 TRACE nova.compute.manager [instance: d9bfa207->6e85-4ca8-a385-52d90818a49b] _("Unexpected vif_type=%s") % vif_type) 2015-08-21 22:55:00.391 41235 TRACE nova.compute.manager [instance: d9bfa207->6e85-4ca8-a385-52d90818a49b] NovaException: Unexpected vif_type=binding_failed

These are my neutron and compute node network interfaces http://pastebin.com/vJyCy5zc

Nova.conf file on the Compute node http://pastebin.com/RvYkUEfW

I have searched google, openstack forum and elsewhere and tried all their suggestions, but the issue remains unsolved. Here is a link to my openstack-forum question: Unable to launch instances (Unexpected vif_type=binding failed)

Also note that I do not have any connectivity issues to the AMQP server on the controller, as I get successfully connected message in my logs.

All my openstack services are up and running properly Has anyone faced a similar issue? I have been struggling with this issue, since last 2 weeks.
Any help will be deeply revered.

1
openstackopenstack-novaopenstack-neutronopenvswitch
The binding_failed message indicates a problem with your Neutron configuration. You should check in the agent log on your compute host (possibly /var/log/neutron/openvswitch-agent.log) and the server log (/var/log/neutron/server.log) on the controller, and update your question with any errors you find there. - larsks
please have a look at [pastebin.com/vJyCy5zc] for my n/w interfaces. Logs are at [pastebin.com/aspEAFkt]. - gaurav parashar
I got a workaround. I created another network with some other IP range(10.0.12.0/24) other than the already existing (203.0.113.0/24) N/w. Now when i use this 10 series N/w to launch instances, i see a successful message, but when i try to use the 203 series , it again gives me the same binding failed error. Can you sugggest what is causing this? And more importantly, how do i rectify this? Have a look at my interfaces as pasted in the above comment. 172 series is the management interface and 182 is the tunnel one. Thanks - gaurav parashar
Ok! some more info.....here is what i see in the subnet table under the neutron database. I see enable_dhcp set to 0 for the 203 series and 1 for the 10 series. This is perhaps the issue. But how do i fix it?? - gaurav parashar
Ok enabled_ dhcp for the 203 series network, still same issue - gaurav parashar

1 Answers

1
votes

Modify /etc/nova/nova.conf file and increase the allocation ratio . cpu_allocation_ratio and memory_allocation_ratio in compute node . Restart the nova and nova-scheduler service in compute node

systemctl restart openstack-nova-* systemctl restart openstack-nova-scheduler.service