[Question] How tunnel handle connections

[adamzhoul]

after reading: https://openyurt.io/blog/Build-side-efficient-collaborative-network

I have some questions:

if we have n tunnel-agent pods and m tunnel-server pods
does each tunnel-agent has m connections and each tunnel-server has n connections?
isn't this too complicated for agents and too many connections for agents and servers?

has the problem the article comes up been fixed?

what if a new agent hasn't connected to all server, how can we make sure apiServer traffic can be forward to that agent?
we put svcIP in host configmap, right ?

[YRXING]

If we have n tunnel-agent and m tunnel-server, then there will be m*n grpc connections. Deploying multiple tunnel servers can solve the single point problem .
if a new agent hasn't connected to all server, the traffic may not reach the tunnel agent. This is indeed a problem at the start of a new agent.
the api-server traffic will be forward tunnel-server firstly( through two ways: DNAT/DNS ), then it's the proxy server's responsibility to select the right tunnel agent (parse the http header's host filed) and forward to it.

[rambohe-ch]

after reading: https://openyurt.io/blog/Build-side-efficient-collaborative-network

I have some questions:

if we have n tunnel-agent pods and m tunnel-server pods does each tunnel-agent has m connections and each tunnel-server has n connections? isn't this too complicated for agents and too many connections for agents and servers?

has the problem the article comes up been fixed?

what if a new agent hasn't connected to all server, how can we make sure apiServer traffic can be forward to that agent?
we put svcIP in host configmap, right ?

@adamzhoul we will try to solve this problem in new edge network proposal, you can reference here: #637

[adamzhoul]

that's great