InOut patterns and clustering

[Jamie Le Grange <ja...@googlemail.com>]

hi,

i've been looking at the loanbroker example in 2.0.2 because i want to
do something similar and it looks like the pattern used there is to
stash the initial inout exchange in a concurrenthashmap so that when
the result finally comes in, it can be matched up with the request. is
this going to work across a cluster i.e. are inout messages guaranteed
to come back to the same component that initiated an inout exchange?
if so, what happens if that engine in the cluster fails midway through
processing?

cheers,
j.

Re: InOut patterns and clustering

[Jamie Le Grange <ja...@googlemail.com>]

Hi Guillaume,

Thanks for the clarification. Since, in the loanbroker example, the
component is relying on getting the message back to the originating
component it may be advantageous to wait for it the engine to come
back up instead of sending it to another node in the cluster... unless
the component state is also being distributed (anyone know if
activespace is still under active development).

cheers,
j.

On 17/01/06, Guillaume Nodet <gn...@gmail.com> wrote:
> In a cluster, all the exchanges are routed back to the exact originator
> component.
> You are right that this may lead to problems when the node fails.  Some
> messages
> won't be processed until it comes up again.
>
> Could you please raise a jira for that ?
>
> Cheers,
> Guillaume Nodet
>
> Jamie Le Grange wrote:
>
> >hi,
> >
> >i've been looking at the loanbroker example in 2.0.2 because i want to
> >do something similar and it looks like the pattern used there is to
> >stash the initial inout exchange in a concurrenthashmap so that when
> >the result finally comes in, it can be matched up with the request. is
> >this going to work across a cluster i.e. are inout messages guaranteed
> >to come back to the same component that initiated an inout exchange?
> >if so, what happens if that engine in the cluster fails midway through
> >processing?
> >
> >cheers,
> >j.
> >
> >
> >
> >
>

Re: InOut patterns and clustering

[Guillaume Nodet <gn...@gmail.com>]

In a cluster, all the exchanges are routed back to the exact originator 
component.
You are right that this may lead to problems when the node fails.  Some 
messages
won't be processed until it comes up again.

Could you please raise a jira for that ?

Cheers,
Guillaume Nodet

Jamie Le Grange wrote:

>hi,
>
>i've been looking at the loanbroker example in 2.0.2 because i want to
>do something similar and it looks like the pattern used there is to
>stash the initial inout exchange in a concurrenthashmap so that when
>the result finally comes in, it can be matched up with the request. is
>this going to work across a cluster i.e. are inout messages guaranteed
>to come back to the same component that initiated an inout exchange?
>if so, what happens if that engine in the cluster fails midway through
>processing?
>
>cheers,
>j.
>
>
>  
>