I started working a bit on adding support for both spark-jobserver and
Zeppelin, but it was more just on-the-side work while waiting for other
things to compile.  I ran into some issues with each project though and
haven't had the bandwidth to investigate them further quite yet.  For
spark-jobserver, I did actually get it working with Bigtop (RPM
only--haven't done the DEB config), but I was getting test failures, so I
had to disable them temporarily for the Bigtop build.  For Zeppelin, I was
getting some build time failure that I wasn't sure about.

~ Jonathan




On 3/20/15, 2:50 PM, "Roman Shaposhnik" <[email protected]> wrote:

>I would love to see Zeppelin in Bigtop but personally
>I'm extremely short on cycles (you'll see at ApacheCON
>what keeps me busy).
>
>Thanks,
>Roman.
>
>On Fri, Mar 20, 2015 at 2:46 PM, Konstantin Boudnik <[email protected]>
>wrote:
>> Yup! It'd be great if we can start trying to get Zeppelin on board.
>>They have
>> just sent in their SGA, so I believe the first release might be coming
>>soon
>> too ;) Although, for the experiments we can just probably work off their
>> github bits.
>>
>> On Fri, Mar 20, 2015 at 04:34PM, RJ Nowling wrote:
>>> I didn't realize that you and Roman were involved with Zeppelin until
>>>I saw
>>> your emails on the Zeppelin dev list today.
>>>
>>> I see a lot of big data community interest in more user-friendly
>>>interfaces
>>> and visualization software -- following what the Python community has
>>>done
>>> with iPython notebook and friends.  Unfortunately, many of the popular
>>> solutions are proprietary.  I would love to push on open source
>>>efforts and
>>> open that area up.
>>>
>>> Zeppelin could make a great addition to BigTop.  It would provide
>>>BigTop
>>> with a competitive feature while also raising the profile of Zeppelin.
>>>
>>> Zeppelin also needs more testing and cleaning up.  I've had trouble
>>>getting
>>> it working (see the Zeppelin user mailing list).  Good opportunity to
>>>get
>>> in there and start hacking on the code.
>>>
>>> From Red Hat's perspective, packaging Zeppelin will require some work.
>>> Fedora, et al. require that each JAR is provided by a single RPM.
>>> Likewise, we'd have to work on splitting out the Node.js dependencies
>>>so
>>> they can depend on Fedora-available JARs.  I realize that BigTop's
>>> packaging isn't so strict, but I think packaging in BigTop is a good
>>>first
>>> step, and we can work with the Zeppelin community to work out any
>>>kinks in
>>> their build system related to packaging.
>>>
>>> I see a win-win situation.  I also see an opportunity for BigTop to
>>>once
>>> again prove that it's on the cutting edge.
>>>
>>>
>>> On Fri, Mar 20, 2015 at 11:53 AM, Konstantin Boudnik <[email protected]>
>>>wrote:
>>>
>>> > Thanks RJ! Actually, Zeppelin is quite interesting. If you look at
>>>their
>>> > proposal here https://wiki.apache.org/incubator/ZeppelinProposal
>>> > it says under Initial Goals
>>> >
>>> > ".... adding Zeppelin distribution to Apache Bigtop"
>>> >
>>> > So, I guess it's mutual interest! Would be nice to hear your
>>>thoughts on
>>> > that!
>>> > Thanks,
>>> >   Cos
>>> >
>>> > On Fri, Mar 20, 2015 at 07:49AM, RJ Nowling wrote:
>>> > > Hi all,
>>> > >
>>> > > I went to Spark Summit East 2015.  Many of the talks emphasized
>>>notebooks
>>> > > and visualization products for Spark.  My Red Hat team found a few
>>>open
>>> > > source alternatives, which I wanted to share for others who might
>>>be
>>> > > interested:
>>> > >
>>> > > * Apache Zeppelin (incubating) -
>>>https://zeppelin.incubator.apache.org/
>>> > > * Jove Jupyter - https://github.com/jove-sh/jove-jupyter-frontend
>>> > > * Spark Notebook - https://github.com/andypetrella/spark-notebook
>>> > >
>>> > > We're currently trying to get Apache Zeppelin working.  I'll
>>>report more
>>> > as
>>> > > I learn more.
>>> > >
>>> > > RJ
>>> >

Reply via email to