IGNITE-1098: IgniteCache.invoke() works incorrectly in async mode.

[Vladimir Ozerov]

Igniters,

I noted strange behavior of IgniteCache.invoke() in async mode. If
exception is thrown from entry processor, sometimes it lands in future as
expected, but sometimes it is thrown right-away as if it was a synchronous
call. Behavior depends on topology and may be key affinity. Test case is
provided in the ticket: https://issues.apache.org/jira/browse/IGNITE-1098

As I understand, IgniteCache.invoke() must never throw processor exceptions
according to our "async" contract (as well as ny other async-supported
operations).

Can anyone throw a glance on the problem?

Vladimir.

[Atri Sharma]

I sure can, having dealt with async problems earlier.

Can you advice how to deal with this,please?

On Mon, Jul 6, 2015 at 8:43 PM, Vladimir Ozerov <[hidden email]>
wrote:

> Igniters,
>
> I noted strange behavior of IgniteCache.invoke() in async mode. If
> exception is thrown from entry processor, sometimes it lands in future as
> expected, but sometimes it is thrown right-away as if it was a synchronous
> call. Behavior depends on topology and may be key affinity. Test case is
> provided in the ticket: https://issues.apache.org/jira/browse/IGNITE-1098
>
> As I understand, IgniteCache.invoke() must never throw processor exceptions
> according to our "async" contract (as well as ny other async-supported
> operations).
>
> Can anyone throw a glance on the problem?
>
> Vladimir.
>

--
Regards,

Atri
*l'apprenant*

[Vladimir Ozerov]

Atri,

Look at IgniteCacheProxy.invoke() - this is where all things happen. Steps
to reproduce are in the ticket.

On Mon, Jul 6, 2015 at 6:14 PM, Atri Sharma <[hidden email]> wrote:

> I sure can, having dealt with async problems earlier.
>
> Can you advice how to deal with this,please?
>
> On Mon, Jul 6, 2015 at 8:43 PM, Vladimir Ozerov <[hidden email]>
> wrote:
>
> > Igniters,
> >
> > I noted strange behavior of IgniteCache.invoke() in async mode. If
> > exception is thrown from entry processor, sometimes it lands in future as
> > expected, but sometimes it is thrown right-away as if it was a
> synchronous
> > call. Behavior depends on topology and may be key affinity. Test case is
> > provided in the ticket:
> https://issues.apache.org/jira/browse/IGNITE-1098
> >
> > As I understand, IgniteCache.invoke() must never throw processor
> exceptions
> > according to our "async" contract (as well as ny other async-supported
> > operations).
> >
> > Can anyone throw a glance on the problem?
> >
> > Vladimir.
> >
>
>
>
> --
> Regards,
>
> Atri
> *l'apprenant*
>

[Atri Sharma]

Taking a look now.

On Mon, Jul 6, 2015 at 9:09 PM, Vladimir Ozerov <[hidden email]>
wrote:

> Atri,
>
> Look at IgniteCacheProxy.invoke() - this is where all things happen. Steps
> to reproduce are in the ticket.
>
> On Mon, Jul 6, 2015 at 6:14 PM, Atri Sharma <[hidden email]> wrote:
>
> > I sure can, having dealt with async problems earlier.
> >
> > Can you advice how to deal with this,please?
> >
> > On Mon, Jul 6, 2015 at 8:43 PM, Vladimir Ozerov <[hidden email]>
> > wrote:
> >
> > > Igniters,
> > >
> > > I noted strange behavior of IgniteCache.invoke() in async mode. If
> > > exception is thrown from entry processor, sometimes it lands in future
> as
> > > expected, but sometimes it is thrown right-away as if it was a
> > synchronous
> > > call. Behavior depends on topology and may be key affinity. Test case
> is
> > > provided in the ticket:
> > https://issues.apache.org/jira/browse/IGNITE-1098
> > >
> > > As I understand, IgniteCache.invoke() must never throw processor
> > exceptions
> > > according to our "async" contract (as well as ny other async-supported
> > > operations).
> > >
> > > Can anyone throw a glance on the problem?
> > >
> > > Vladimir.
> > >
> >
> >
> >
> > --
> > Regards,
> >
> > Atri
> > *l'apprenant*
> >
>

--
Regards,

Atri
*l'apprenant*