Hi Mateusz,

Please see below

On 11/4/2015 11:13 AM, mp wrote:
Hi Dmitriy,

That is great news! I will definitely try out the new version as soon as it's released.

Is there any chance to include a fix for https://issues.apache.org/jira/browse/IGNITE-1823 in the upcoming version (see the message from Artem earlier in this thread)?
It depends on whether anyone in the community will pick it up or not.

In any case I still wouldn't recommend you to use the SHARED deployment mode if multiple developers actively testing something in a cluster. One day you may face with the issue when one developer's class is overriden by a new version of the same class from the other developer.

Would it be possible to achieve the following in the new version (this is a typical scenario in my setup): a cluster of server nodes on powerful machines, two analysts/developers are actively testing some prototype algorithm and are using the same class *with different code* and running their computations on the cluster *at the same time*. Each computation uses different code for the same class (the code exists only on the developer's client node)?

Are you going to work with any cache in a compute task body?
- If you aren't going then you can freely enable PRIVATE or ISOLATE mode (to exclude a possible case when one developer's class is overridden by a new class's version from the other developer) and start using Compute Engine; - If you are going than try to give a shot to SHARED mode but use Compute Engine instead of IgniteQueue. IGNITE-1823 is reproduced with IgniteQueue but probably this will work with Compute Engine.

Dmitriy is clarifying regarding the way the new version, he mentioned below, on the dev list. Should reply you soon.

--
Denis

Thanks a lot for sharing the outlook for the upcoming release!

Cheers,
-Mateusz

On Tue, Nov 3, 2015 at 8:13 PM, Dmitriy Setrakyan <[email protected] <mailto:[email protected]>> wrote:

    Hi Mateusz,

    In the upcoming Ignite 1.5 release, we will not be storing
    user-classes on the server side at all, so the need for deployment
    of enqueued objects goes away.

    With that in mind, we will be removing the requirement for caches
    to work only with SHARED and CONTINUOUS deployment modes, so you
    will be able to use PRIVATE or ISOLATED deployment modes to deploy
    your computations.

    We are currently actively working on finalizing the 1.5 release,
    so it should be coming real soon.

    D.

    On Tue, Nov 3, 2015 at 3:05 AM, mp <[email protected]
    <mailto:[email protected]>> wrote:

        Hi Denis,

        Thanks a lot for the clarifications.

        So, is there any way in Ignite to achieve my goal: different
        versions of the same class running on the cluster (they can be
        deployed at the same time from different client nodes)?
        If yes, I can think of a way in which I can achieve my goals
        without using the distributed queue.

        Cheers,
        -Mateusz


        On Tue, Nov 3, 2015 at 11:53 AM, Denis Magda
        <[email protected] <mailto:[email protected]>> wrote:

            Hi Mateusz,

            Unfortunately I don't see any way to specify userVersion
            programmatically.
            In any case CONTINIOUS and SHARED mode is not suitable for
            your task because
            the developers may want to have different version of the
            same class. Please
            refer to DeploymentMode documentation for more info on
            different mode types.

            Do you really need to use IgniteQueue in your tests
            because, yes, neither
            PRIVATE nor ISOLATED modes are supported for user caches.

            --
            Denis



            --
            View this message in context:
            
http://apache-ignite-users.70518.x6.nabble.com/Distributed-queue-problem-with-peerClassLoading-enabled-tp1762p1820.html
            Sent from the Apache Ignite Users mailing list archive at
            Nabble.com.





Reply via email to