synapse-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From indika kumara <indika.k...@gmail.com>
Subject Re: Issue with switch on/off LoadBalance and FailOver endpoints from Jconsole
Date Fri, 04 Feb 2011 04:00:46 GMT
More thoughts on the use of indirect endpoints. Just my personal view ...

I think it is a rare usecase that a user wants to use the same endpoint
instance in two different groups. It is like calling the same service from
different endpoint groups.

What he really wants is to improve the readability, maintainability and
reuse of the synapse.xml. It is the same as a Java method (or any other)
with a lot of codes. We usually refactor such a method by groping code into
a set of methods. If we want to reuse a method (any other) across classes,
then we parameterized the method (Here I am referring to the capabilty of
passing parameters or arguments). The capability of parameterization for a
language construct is powerful and allow a maximum reuse.

Defining an endpoint in-lined with all details such as error handing, etc is
equal to a make the code inside a method large. It is possible to become
many information redundant except the endpoint URI. A current indirect
endpoint can only support improving readability and maintainability. But, it
can not support true reuse (avoiding redundant information, reuse across
endpoints). A parameterized endpoint can support both. For instance, an
endpoint definition with a URI which is given as a parameter and an indirect
endpoint refer such as endpoint with the name of the EP and the URI as a
parameter.

~ Indika

On Thu, Feb 3, 2011 at 10:59 PM, indika kumara <indika.kuma@gmail.com>wrote:

> + 1
>
> BTW, why do we use indirect endpoints?  if it is for reuse then what are
> the reusable elements - an endpoint definition, an runtime endpoint instance
> or both?  what kind of reuse an end user expects?
>
> If the child endpoint of ST_Quote_LB1 get inactive due to failures, should
> that cause to stop ST_Quote_FO1 working? if the the runtime instance is
> reused, then I believe It would happen. But, if we expect to reuse only the
> endpoint definition, it should not happen. And for the latter case, the
> switchoff of ST_Quote_LB1 (including its children) would not affect any
> other endpoint. So, if we expected the reuse of endpoint definition, then,
> the current implementation of the indirect endpoint is wrong.
>
> ~ Indika
>
> On Thu, Feb 3, 2011 at 12:11 PM, Miyuru Wanninayaka <miyurudw@gmail.com>wrote:
>
>> Hi all,
>>
>> As current implementation of LoadBalance and FailOver endpoints, when we
>> turned off LB/FO endpoint from jconsole, what really happens is all child
>> endpoints of FO/LB endpoints gets switched off. I think this behavior is
>> wrong. Since FO/LB endpoint has a context, it's possible to keep status in
>> that context and handle in send() method of LB/OF endpoints without turning
>> off child endpoints.
>>
>> Also in shared indirect endpoint situations like this:
>>
>>     <endpoint name="ST_Quote_LB1">
>>         <loadbalance
>> algorithm="org.apache.synapse.endpoints.algorithms.RoundRobin">
>>             <endpoint key="*SimpleStockQuoteService_EP*"/>
>>         </loadbalance>
>>     </endpoint>
>>     <endpoint name="ST_Quote_FO1">
>>         <failover>
>>             <endpoint key="*SimpleStockQuoteService_EP*"/>
>>         </failover>
>>     </endpoint>
>>     <endpoint name="*SimpleStockQuoteService_EP*">
>>         <address uri="
>> http://localhost:9000/services/SimpleStockQuoteService"/>
>>     </endpoint>
>>
>> turning off ST_Quote_LB1 will cause ST_Quote_FO1 to stop working.
>>
>> Any ideas?
>>
>> --
>> Miyuru Wanninayaka
>> Software Engineer - WSO2 Inc <http://wso2.com/>.
>>
>
>

Mime
View raw message