Re: [Regops] draft-wullink-restful-epp-00.txt

> Sure, but what happens with an existing EPP-session when one of the
> application servers suddenly becomes unavailable?

It depends where the state is held (gateway or application server).  If
the application server is stateless, than nothing happens to the EPP
session.  If the application server is stateful, then hopefully the
gateway will recognize the dropped application server connection and will
in turn close the client connection.

> Yes, but that implies more complexity on the gateways/loadbalancers.

Load balancing behind the gateways is just one option, but you are correct
that it adds complexity.

> Besides that, we noticed there are quite a lot of registrars using a
> 'login-command-logout' sequence for every request they wish to make.
> That is, in our view, less efficient than simply issuing one RESTful
> request each time.

We see this as well, but there are another set of registrars that keep
long running EPP sessions that get the benefit of stateful sessions.  A
best practice is to keep a pool of long running sessions for the best
performance.  



> We've encountered a number of scenario's where clients ran into their
> session limits unexpectedly. I remember one case where the client forgot
> to send a <logout>, but that is off course their problem. There are also
> scenario's where a flaky TCP connection or buggy client messes things
> up. Every so often this results in questions to our supportdesk.

You're going to most likely have some sort of server policy on load
(connections, bandwidth) in either event that can result in
troubleshooting and questions to customer support.  I'm not sure if
stateless versus stateful will help reduce the server policies and
subsequently the troubleshooting.



--
  
JG
 

 
James Gould
Principal Software Engineer
[email protected]
 
703-948-3271 (Office)
12061 Bluemont Way
Reston, VA 20190
VerisignInc.com







On 4/23/12 3:07 PM, "Marco Davids (SIDN)" <[email protected]> wrote:

>Op 23-04-12 18:26, Gould, James schreef:
>
>> Load balancing the connections distributes the load fine today.
>
>Sure, but what happens with an existing EPP-session when one of the
>application servers suddenly becomes unavailable?
>
>> If there is a need to load balance on a per command basis,
>> you can keep the state in the gateways
>
>Yes, but that implies more complexity on the gateways/loadbalancers.
>
>Besides that, we noticed there are quite a lot of registrars using a
>'login-command-logout' sequence for every request they wish to make.
>That is, in our view, less efficient than simply issuing one RESTful
>request each time.
>
>> 2. EPP sessions can wind up in a state where they are no longer linked
>>to
>> an active TCP session - Can someone describe this issue?
>
>We've encountered a number of scenario's where clients ran into their
>session limits unexpectedly. I remember one case where the client forgot
>to send a <logout>, but that is off course their problem. There are also
>scenario's where a flaky TCP connection or buggy client messes things
>up. Every so often this results in questions to our supportdesk.
>
>--
>Marco
>_______________________________________________
>provreg mailing list
>[email protected]
>https://www.ietf.org/mailman/listinfo/provreg

_______________________________________________
provreg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/provreg