bernardodemarco commented on PR #7333:
URL: https://github.com/apache/cloudstack/pull/7333#issuecomment-4120305762

   Hello, @matheusfontes and @weizhouapache
   
   > Most VM settings are intended to support user VMs only.
   > I'm not aware of any recent changes that extend this support to Virtual 
Routers or Internal LB.
   
   Yes. There was the PR #10736 that renamed the `user_vm_details` table in the 
`cloud` database to `vm_instance_details`, aiming to better separate 
responsibilities and enable support for VM details across different VM types, 
including system VMs. However, support for system VM details has not yet been 
implemented.
   
   ---
   
   > We had to implement this change using the 
agent.hooks.libvirt_vm_xml_transformer.script hook on KVM hosts to enable 
multiqueue for Virtual Routers.
   
   @matheusfontes, this is a great workaround. However, it is important to note 
that it does not work for VPC VRs. During deployment, the Management Server 
sends a `StartCommand` to the Agent, which deploys the instance and configures 
only the NIC for the link-local/control network (CIDR `169.254.0.0/16`).
   
   After the VR is successfully deployed, the NICs associated with public and 
guest networks are configured separately through `PlugNicCommand`. As a result, 
when the Agent hook is triggered, the VR domain XML contains only a single NIC, 
making it impossible to configure the remaining interfaces.
   
   ---
   
   > can you try with some key/value settings of the SYSTEM template ?
   > We tested using key/value settings on the SYSTEM template, but it didn’t 
have any effect on Virtual Routers during deployment.
   
   Yes. Currently, template settings are not inherited by system VMs.
   
   > While this could be a good alternative if supported, it would still not 
allow us to adjust the number of queues based on the number of vCPUs.
   > This type of configuration would fit better at the System Offering level, 
where CPU characteristics are already defined.
   
   Yes, while that approach would be valid, maintaining consistency with the 
existing settings model for user VMs would be a better solution IMHO. This 
would also allow to adjust the number of queues based on the CPU 
characteristics of the VMs.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to