If I understand this correctly, it will take the max(2, nproc/4) -- So upstream CI, we have 8 vCPUs for nodes, and based on my math that would put us at 2 workers, similar to what we do.. but it seems nothing tunes rpc_workers (however I guess that doesn't matter in an OVN situation?)
> If OVN ML2 plugin is used without any additional agents, neutron requires no worker for RPC message processing. Set both rpc_workers and rpc_state_report_workers to 0, to disable RPC workers.
Which means that doesn't matter.. and upstream CI is using 2 API workers and I assume not hitting this.. I don't know. We do run in Kubernetes/containerized, so not sure if that has any impact..
Hmm, upstream CI does this:
https:/ /github. com/openstack/ devstack/ blob/master/ lib/neutron# L444-L448 /github. com/openstack/ devstack/ blob/master/ stackrc# L778-L782
https:/
If I understand this correctly, it will take the max(2, nproc/4) -- So upstream CI, we have 8 vCPUs for nodes, and based on my math that would put us at 2 workers, similar to what we do.. but it seems nothing tunes rpc_workers (however I guess that doesn't matter in an OVN situation?)
https:/ /docs.openstack .org/neutron/ latest/ admin/config- wsgi.html
However, the docs do say...
> If OVN ML2 plugin is used without any additional agents, neutron requires no worker for RPC message processing. Set both rpc_workers and rpc_state_ report_ workers to 0, to disable RPC workers.
Which means that doesn't matter.. and upstream CI is using 2 API workers and I assume not hitting this.. I don't know. We do run in Kubernetes/ containerized, so not sure if that has any impact..