openstack new compute node can’t migrate for

update /etc/hosts resolve

restart nova_compute, nova_libvirt container

error log:

Live migration Error : Failed to connect to remote libvirt URI qemu tcp://compute1-VirtualBox/system: Unable to resolve address service '16509': Name or service not known

resolve:

Hmm, that was not fixed because it is docker that manages /etc/hosts. It seems to be copying this on container restart indeed. Interesting. I run my deploy with DNS so never experienced this but this surely looks like an issue to be aware of when users report their problems. We might want to think about a possible workaround. Maybe start mounting /etc/hosts ro in all containers so it follows host's - we don't do docker networking so it does not matter that docker cannot add new entries this way.

link :

https://ask.openstack.org/en/question/79336/live-migration-error-failed-to-connect-to-remote-libvirt-uri-qemu-tcpcompute1-virtualboxsystem-unable-to-resolve-address-service-16509-name-or-service/

https://bugs.launchpad.net/kolla-ansible/+bug/1857245

openstack instance err:”Exeeded maximum number of retries”

0, the error instance log show:

image.png

1, check the /var/log/nova/nova-compute.log, the log show:

image.png

maybe the device cannot support the kvm or not have the modue

reference from : 

"https://ask.openstack.org/en/question/99882/libvirterror-invalid-argument-could-not-find-capabilities-for-domaintypekvm/"

"https://github.com/minishift/minishift/issues/1630"

2, install the qemu-kvm

image.png

3, restart the nova-compute container 

and relaunch a new instance, it still show the same err:"Exeeded maximum number of retries", but the nova-compute.log show difference, there is no permission access kvm character device

image.png

reference from: "https://serverfault.com/questions/1002043/libvirt-has-no-kvm-capabilities-even-though-qemu-kvm-works"

4, and the change the permission to 666 /dev/kvm

image.png

5, relaunch the instance and it works