When running in fabric8 v2/OSE 3.1, which is how apiman will be
primarily productized, the Implementation endpoint for a particular
service will be the one that load balances between the pods.
We also support a pluggable API connector layer, so there are
opportunities to create custom connectors for other platforms, in cases
where a load balanced HTTP endpoint is not available.
-Eric
On 9/21/2015 2:16 PM, Charles Moulliard wrote:
That should be good that we discuss what we have done for Fabric
project to maybe find a common solution even if fabric uses behind the scene zookeeper to
replicate end points registered into local nodes and assigned to a group that our http
gateway will use to find one of the endpoints to load balance
Sent from my iPhone
> On 21 sept. 2015, at 20:07, Eric Wittmann <eric.wittmann(a)redhat.com> wrote:
>
> I see - I misunderstood your question. Your issue is that you want apiman to do the
load balancing for you.
>
> We do not currently have that capability. You would need to stick a load balancer in
between the API Gateway and your service. Then use the load-balanced endpoint as the
Implementation endpoint in apiman.
>
> We are discussing future support for basic routing capabilities in apiman, but no
decisions have yet been made.
>
> -Eric
>
>> On 9/21/2015 1:40 PM, Charles Moulliard wrote:
>> As the endoint registered within the API UI under implementation of a
>> service is tied to a "HTTP Server" running within a machine
>>
(
https://github.com/FuseByExample/rest-dsl-in-action/blob/master/images/ap...),
>> I don't see how elasticsearch could manage multiple HTTP Servers for one
>> Api endpoint
>>
>> Example: My REST service runs at this address
>>
http://localhost:8181/blog/article but also at
>>
http://localhost:8282/blog/article. As the API endpoint that we define
>> within the screen "implementation" of the Service Screen of the API UI
>> will accept one
>> backend-service implementation, how can we tell to Api that 2 nodes are
>> available for the same service ?
>>
>>> On 21/09/15 19:19, Eric Wittmann wrote:
>>> apiman should already support having multiple, load-balanced, active
>>> gateway nodes. We use elasticsearch to coordinate between the nodes.
>>> So the only requirement is that all apiman gateway nodes are
>>> configured to point at the same ES cluster.
>>>
>>> -Eric
>>>
>>>> On 9/21/2015 12:07 PM, Charles Moulliard wrote:
>>>> Hi,
>>>>
>>>> Is it planned that Apiman will support active-active scenario and will
>>>> like Fuse HTTP Gateway load babalance requests to one of the available
>>>> endpoints ?
>>>>
>>>> Regards,
>>>>
>>>> --
>>>> Charles Moulliard
>>>> Principal Solution Architect / JBoss Fuse Expert - Global Enablement
>>>> @redhat
>>>> cmoulliard(a)redhat.com | work: +31 205 65 12 84 | mobile: +32 473 604 014
>>>> MC-Square Business "Stockholm", Leonardo Da Vincilaan 19,
Diegem 1831 -
>>>> Belgium
>>>> twitter: @cmoulliard <
http://twitter.com/cmoulliard> | blog:
>>>> cmoulliard.github.io <
http://cmoulliard.github.io>
>>>> committer: apache camel, karaf, servicemix, hawtio, fabric8, drools,
>>>> jbpm, deltaspike
>>>>
>>>>
>>>> _______________________________________________
>>>> Apiman-dev mailing list
>>>> Apiman-dev(a)lists.jboss.org
>>>>
https://lists.jboss.org/mailman/listinfo/apiman-dev
>>