Open powerds opened 7 years ago
v1k0d3n
commented
7 years ago @powerds thanks for checking out openstack-helm! so this feature is in fact the direction of openstack-helm, and we are working on this currently. we will have a couple backend options for cinder (as an example) to show developers exactly how we plan on addressing this feature request going forward (and before we enter openstack proper).
powerds
commented
7 years ago @v1k0d3n thanks. but it looks there's some difference between cinder multi backend function (#300) and neutron agent plugin function on openstack-helm. because in cinder case, there's no change of k8s resources. the cinder configuration acts in cinder-volume deployment so it's enough to fix cinder.conf. but in neutron case, ovs-agent, ovs-db, ovs-vswitchd daemonsets should be not deployed when we set up to use linuxbridge-agent. it means neutron-helm chart should control what *-agent daemonsets will be deployed or not. do we need new issue for neutron?
Is this a bug report or feature request? (choose one): feature request
Kubernetes Version (output of
kubectl version):Helm Client and Tiller Versions (output of
helm version):Development or Deployment Environment?:
Release Tag or Master:
Expected Behavior:
What Actually Happened:
How to Reproduce the Issue (as minimally as possible):
Any Additional Comments:
openstack's modules are implemented to select plugin backend system or agents. for example, nova can select hypervisors like kvm, vmware, xen and so on, neutron can select agents like linuxbridge-agent, opencswitch-agent and so on. So it needs to design to choose plugins in helm charts. in neutron-helm package case, it needs to configure to disable openvswitch-agent to do not deploy openvswitch-agent daemonset, and enable another agent like linuxbridge-agent daemonset and config map. almost all openstack's modules have same concept to use plugins, so we need to design all charts can use.