On Tue, Nov 25, 2008 at 2:01 PM, Chris Taylor <[EMAIL PROTECTED]> wrote:

> Brilliant!  Thanks, Alex.
>
> Incidentally, in analyzing this issue it would seem that the process store
> was holding multiple "Active" versions of the same process (so, multiple
> PIDs of the same Type in the STORE_PROCESS were Active at the same time).
>
> I wonder if this development environment, wherein we are using the
> DeploymentService to deploy Deployment Units (and thus versioned deployment
> directories), combined with a couple different deployments of the ODE
> runtime engine was causing these "logically" retired processes to hang
> around as Active.
>
> Going out on a limb here, but does it make sense during times of recovery
> that the engine might get confused about which process Type to route to if
> there were multiple Active Process IDs of it at the same time?
>

Yeah, in that case the result is undefined. There's basically no way for the
engine to guess which one is the real destination.

Matthieu


>
>
>
>
>
> ________________________________
> From: Alex Boisvert <[EMAIL PROTECTED]>
> To: [email protected]
> Sent: Tuesday, November 25, 2008 2:50:47 PM
> Subject: Re: Client calling retired process?
>
> On Tue, Nov 25, 2008 at 12:46 PM, Chris Taylor <[EMAIL PROTECTED]> wrote:
>
> > Oh, jeez.  Forget that last one, I obviously wasn't thinking about it
> > enough before I posted it.  We have deployed a couple of different
> versions
> > of the ODE runtime engine, and Websphere stomps on the previous each time
> we
> > do, so we lose the actual /processes/deployment folder for previous,
> retired
> > process versions in that scenario.
> >
> > Still, perhaps a good argument for an external, configurable processes
> > folder?
>
>
> I believe Ode uses the "ode-axis2.working.dir" property in
> ode-axis2.properties:
>
> e.g.
> ode-axis2.working.dir=${org.apache.ode.configDir}/../
>
> alex
>
>
>
>
>

Reply via email to