Yes. I think the error information needs to be exposed. Especially when it
comes to long-running operations where recreating an issue would likely to
be a bit of a hassle.

/Danushka


On Thu, May 15, 2014 at 1:12 AM, Miller, Mark <[email protected]> wrote:

> Remember, I am on the naïve side so I am not saying what should be in
> terms of implementation.
> Certainly, passing to the Gateway for storage in the Gateway database is a
> good solution for the Gateway.
>
> I was wondering if the Airavata operators might need access to the same
> information, for debugging purposes.
>
> -----Original Message-----
> From: Marlon Pierce [mailto:[email protected]]
> Sent: Wednesday, May 14, 2014 12:33 PM
> To: [email protected]
> Subject: Re: Validation failures and storing the failures
>
> I was assuming the gateway would figure out what to do with the errors
> after step 4.  This obviously needs more thought on how this would work.
>
> Marlon
>
> On 5/14/14 3:27 PM, Miller, Mark wrote:
> > I feel as long as they are stored somewhere, that should be fine.
> > The question is, if Airavata doesn't have access to the gateway
> database, will they be hamstrung by the inability to retrieve failure
> messages that might be informative?
> > Maybe everything will be in the logs anyhow?
> >
> > Mark
> >
> > -----Original Message-----
> > From: Marlon Pierce [mailto:[email protected]]
> > Sent: Wednesday, May 14, 2014 12:23 PM
> > To: [email protected]
> > Subject: Re: Validation failures and storing the failures
> >
> > I don't think these need to be stored, just sent back to the gateway.
> > If I recall correctly, steps are
> >
> > 1. Gateway invokes "execute" method of API server.
> >
> > 2. API Server runs Orchestrator's "validate()".
> >
> > 3. Orch returns result of validation.
> >
> > 4. API server returns results to the gateway.
> >
> > 5. If validation passed, API Server then runs the Orchestrator's
> "launch()".
> >
> > So step #4 should enough.
> >
> > Marlon
> >
> >
> > On 5/14/14 12:17 PM, Lahiru Gunathilake wrote:
> >> Hi All,
> >>
> >> We recently added validation logics to orchestrator, so gateway
> >> developers can add/configure their validators to be invoked. If they
> >> failed they can wrap an error message in their validation logic and
> return that.
> >> Orchestrator needs to take these error messages and store them to a
> >> persistent storage so that if the validation failed these errors can
> >> be showed to the gateway user to make it correct. Unless we give a
> >> proper error message to the end user there is no way to make it correct.
> >>
> >> Currently in the orchestrator implementation it does not store the
> >> error but we have a proper result object which keeps the validation
> >> state and a message.
> >>
> >> Where should I save this information in our data model ? This is an
> >> information specific to each invocation and there could be multiple
> >> error messages came from multiple validators because we do not return
> >> immediately after the first validation failure but rather we invoke
> >> all the validators and collect all the errors (currently we just have
> >> access to them and simply logging those errors in server side). If
> >> there is no placeholder for these validation errors can we add
> >> something to the data model to store them and retrieve them from the
> client code ?
> >>
> >> I have added FAQ type of a document for Orchestrator[1], please
> >> provide your feedback on this.
> >>
> >> [1]
> >> https://docs.google.com/document/d/1FOc0X6HCMZ9E-fTnZQ8aGo7tK20C8lvKl
> >> y
> >> cNouswICI/edit
> >>
> >> Regards
> >> Lahiru
> >>
>
>

Reply via email to