Mailing-List: contact dev-help@activemq.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@activemq.apache.org
MIME-Version: 1.0
In-Reply-To: <CAH6wpnpcrUD1=gn1xx_XuuxBMuJ8SOZmnRF=qqaefSWwfFedtA@mail.gmail.com>
References: <D666AB6F-3337-40FA-B1A6-F567281FC972@me.com> <CAA9ykMJng9xiRXWVwmtx-jBCqHboE8+0zhQ76EN+simjjAkN5w@mail.gmail.com>
 <8D68B0BD-7EC9-4F96-AFE2-9EAC9DA5949C@me.com> <CACHnxzz1_JfLgE0J1QQ2RB_DcPFNzH4roJ_a-8tssSfq82rYwA@mail.gmail.com>
 <CAA9ykM+TrLNTJPyy8YENtv8w2zAW-fVP4b8UMTBmNkBXRsu0Zw@mail.gmail.com>
 <E885CFFA-C533-4BD6-943C-552B02678A4D@me.com> <FC929691-1BB8-4070-8045-E2DDFD4C8FFF@me.com>
 <CAFitrpTtp9yiQFJD5TgZ2=cPEEwrqTJ9yJJScvgeYPEqOJM2+g@mail.gmail.com> <CAH6wpnpcrUD1=gn1xx_XuuxBMuJ8SOZmnRF=qqaefSWwfFedtA@mail.gmail.com>
From: Justin Bertram <jbertram@apache.org>
Date: Fri, 27 Oct 2017 10:09:02 -0500
Message-ID: <CAF+kE=SJmoRKJ0QwWhRmQuk1hJHwPV5fat7ixXnEZysE6+-XQg@mail.gmail.com>
Subject: Re: Kafka ServiceConnector
To: dev@activemq.apache.org
Content-Type: multipart/alternative; boundary="001a113ce1864f04f8055c88ab72"
archived-at: Fri, 27 Oct 2017 15:09:08 -0000

--001a113ce1864f04f8055c88ab72
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

In my opinion integrations like this are best kept outside the broker.  If
they use the service connector hooks then there won't be any performance
impact whether or not they are in the broker code-base.

Also, I'm not sure I understand the comparison made between this
integration and the core bridge.  The core bridge is a necessary bit of
functionality to enable clustering.  Clustering is a core broker feature.
Integration with Kafka (while nice) is not a core broker feature.

As long as they aren't in the Artemis code-base then I don't think it's a
big deal where the integrations live as long as there some up-to-date
documentation about where to find them.


Justin

On Fri, Oct 27, 2017 at 5:05 AM, Martyn Taylor <mtaylor@redhat.com> wrote:

> Looking at the Kafka Connect example.  I see that the integration
> components are all maintained outside of the Kafka project.  They're list=
ed
> here: https://cwiki.apache.org/confluence/display/KAFKA/Ecosystem
>
> How about we do something similar and create a wiki/web page linking the
> integration projects.  I realise our website isn't the most accessible at=
m
> but going forward this should be improved.
>
> On Thu, Oct 26, 2017 at 5:42 PM, Robbie Gemmell <robbie.gemmell@gmail.com=
>
> wrote:
>
> > The related PR is https://github.com/apache/activemq-artemis/pull/1607
> > for anyone not following along thus far, or later reference.
> >
> > My main view there was that such an extension would be better
> > maintained seperate from the main broker repository+distribution for a
> > variety of reasons. After that I didn't have a strong view on where it
> > lives, but I would say that I don't actually see any issue at all with
> > it being outwith the ActiveMQ project. To me thats just a key thing
> > enabled by such plugin points existing, and not uncommon with other
> > projects/components.
> >
> > Robbie
> >
> > On 26 October 2017 at 16:46, Michael Andr=C3=A9 Pearce
> > <michael.andre.pearce@me.com> wrote:
> > > As noted on the PR part of the PR discussion is should Service
> > Connectors exist with the broker code.
> > >
> > > Service Connectors / Extensions are ActiveMQ Artemis specific.
> > >
> > > I think having them within the project space is good so we can grow a=
n
> > eco system of some,  people can just use will aid in having better brok=
er
> > adoption as tools people look for would be available if contributed.
> > >
> > > As stated above it seems the general consensus on the PR is though
> liked
> > is the broker code base best for them to live?
> > >
> > > If we don=E2=80=99t wish them to exist with broker code so they can h=
ave their
> > own release cycle, and don=E2=80=99t murky the water of the broker, I w=
ould like
> to
> > discuss if we could make a sub project for these to live, and possibly
> > propose a sub project be setup called =E2=80=9CActiveMQ Artemis Integra=
tions=E2=80=9D ?
> > >
> > > If setup I would also propose maybe moving the spring integration the=
re
> > also.
> > >
> > > Cheers
> > > Mike
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > > Sent from my iPhone
> > >
> > >> On 6 Oct 2017, at 12:20, Michael Andr=C3=A9 Pearce <
> > michael.andre.pearce@me.com> wrote:
> > >>
> > >> Yes and no, it would be same as arguing that core bridge between
> > artemis clusters could also be separated components.
> > >>
> > >> The idea is to effectively have similar to a core bridge but to kafk=
a.
> > >>
> > >> Some key parts why are (and probably similar to why core bridge is
> > implemented within the broker)
> > >>
> > >> Latency
> > >> Extra processes and hops
> > >> Out the box in broker support
> > >>
> > >> Cheers
> > >> Mike
> > >>
> > >>
> > >> Sent from my iPhone
> > >>
> > >>> On 6 Oct 2017, at 12:14, Andy Taylor <andy.tayls67@gmail.com> wrote=
:
> > >>>
> > >>> I think https://github.com/ppatierno/kafka-connect-amqp is an
> example
> > of
> > >>> this
> > >>>
> > >>> On 6 October 2017 at 12:07, Christopher Shannon <
> > >>> christopher.l.shannon@gmail.com> wrote:
> > >>>
> > >>>> Why not use Kafka's Connect framework?  Seems like building a sour=
ce
> > >>>> connector here would make the most sense.  This is not to say I'm
> > against
> > >>>> the idea I'm just wondering the benefits of doing it from the
> Artemis
> > side
> > >>>> of things when Kafka already has a framework that can be used to
> > >>>> import/export data.
> > >>>>
> > >>>> On Fri, Oct 6, 2017 at 6:16 AM, Michael Andr=C3=A9 Pearce <
> > >>>> michael.andre.pearce@me.com> wrote:
> > >>>>
> > >>>>> Hi Andy,
> > >>>>>
> > >>>>> As you may be aware or not we run both activemq and kafka each ha=
s
> a
> > >>>>> specific plus point to it, use the right tool for the right job.
> > >>>>>
> > >>>>> We use activemq as our feature rich, fully jms compliant and
> > performant
> > >>>>> broker.
> > >>>>>
> > >>>>> And use kafka more for post processing and data streaming, with
> > >>>> historical
> > >>>>> feed also.
> > >>>>>
> > >>>>> Idea here is that producers and consumers bind into our activemq
> > where
> > >>>> jms
> > >>>>> api is wanted or needed, lower per message latencies etc. but the=
n
> > we can
> > >>>>> bridge messages from a core queue into kafka so that the data can
> > flow to
> > >>>>> it as fast as possible and consumed from there also.
> > >>>>>
> > >>>>> Idea is to try make a hybrid taking best of both.
> > >>>>>
> > >>>>> Cheers
> > >>>>> Mike
> > >>>>>
> > >>>>> Sent from my iPhone
> > >>>>>
> > >>>>>> On 6 Oct 2017, at 09:37, Andy Taylor <andy.tayls67@gmail.com>
> > wrote:
> > >>>>>>
> > >>>>>> Whats the use case here Michael, sounds like a cool idea and
> > something
> > >>>>> that
> > >>>>>> has been mentioned by different people but I am yet to understan=
d
> > how
> > >>>>> this
> > >>>>>> could be used.
> > >>>>>>
> > >>>>>> On 6 October 2017 at 07:48, Michael Andr=C3=A9 Pearce <
> > >>>>>> michael.andre.pearce@me.com> wrote:
> > >>>>>>
> > >>>>>>> Hi All,
> > >>>>>>>
> > >>>>>>> I am looking to contribute an artemis ServiceConnector that wou=
ld
> > >>>>> leverage
> > >>>>>>> the existing service connector interface to bridge from a queue
> > into a
> > >>>>>>> kafka topic similar to bridge between artemis clusters.
> > >>>>>>>
> > >>>>>>> Idea is to make it configurable how to transform and encode the
> > >>>> message
> > >>>>> in
> > >>>>>>> kafka but as amqp is becoming very well supported by artemis
> > provide
> > >>>> out
> > >>>>>>> the box that the message be encoded in amqp byte protocol.
> > >>>>>>>
> > >>>>>>> It something I=E2=80=99ve been working on privately for a bit, =
but since
> > >>>>> becoming
> > >>>>>>> a committer here is really like to share it.
> > >>>>>>>
> > >>>>>>> I hope if I get a positive response here to get a branch ready
> > soon,
> > >>>> but
> > >>>>>>> feedback if this is welcomed would be great.
> > >>>>>>>
> > >>>>>>> Cheers
> > >>>>>>> Mike
> > >>>>>>>
> > >>>>>>>
> > >>>>>>> Sent from my iPhone
> > >>>>>
> > >>>>
> >
>

--001a113ce1864f04f8055c88ab72--