Yun Zhou

> I'm OK with bumping the timeout, but would it be more prudent to run pprof in your setup and find out what is taking so long? @winsopc has kindly...

@trozet with the lines removed, nodeTracker initialization of the same zone took less than 2 seconds. but that is for experimental only, right? as it would fail to create template...

issue is no longer relevant because we now found several issues with service template enabled and downstream is now moving back to disable service template. Close this PR for now.

> > Is there a reason why we don't want to use LB Templates? > > There are several issues we found downstream. @cathy-zhou can you please raise issues for:...

> 1. Performance regression while using service templates. @hzhou8 do you have corresponding issue filed for the offloading performance issue when svc_template is enabled? if so, can you post the...

> Please also add this to https://github.com/ovn-org/ovn-kubernetes/blob/master/helm/ovn-kubernetes/values.yaml so helm can benefit from it. @flavio-fernandes please take a look

the way we support it downstream is to add a particular annotation in the net-attach-def to indicate on this particular network, spoofcheck is disabled.

> I think we should add a couple of e2e tests to assert the pods start as expected if we configure routes on them. At least for a default route...

close this MR as @trozet is suggesting a different approach (level-driven network manager controller)