Re: [netconf] Multiple servers and endpoint

[Kent Watsen <[e-mail]>]

> On Sep 3, 2019, at 7:55 AM, Balázs Lengyel <[e-mail]> wrote:
> 
> Hello Kent,
> In some client models you have multiple remote servers defined and multiple endpoints for each server. I would like to clarify the desired behavior for these. 
> E.g. netconf-client-grouping {
> the end-point description states: 
>  
>                  "A user-ordered list of endpoints that the NETCONF
>                   client will attempt to connect to in the specified
>                   sequence.  Defining more than one enables
>                   high-availability."
> As I understand this means the client will try to connect to the first endpoint. If that fails it tries the next one. Correct?


Correct.  That is the intention behind it being user-ordered.


>  For servers the description states: 
> "List of NETCONF servers the NETCONF client is to
>               initiate connections to in parallel."
> What does parallel men here? 
> Will it try to connect all, and abort all connection attempts after the first succeeds?
> Try to send the same information to multiple servers?
> Try to do some load-sharing between multiple servers?


The intention is that these are, e.g., distinct applications the client wishes to establish a connection with. For instance, a firewall might want to connect to a cloud-based deep-inspection engine as well as a cloud-based spam-analyzer.  For each application then, there may be a user-ordered list of "endpoints" for high-availability.

The word "parallel" may be overstated, as it doesn't matter if the implementation actually connects in parallel or not, especially in context of each connection being long-lived (periodic or persistent), the sequence of the initial connections loses its meaning.  Perhaps the text should instead say "maintain parallel or simultaneous connections with"?

Thanks,
Kent