You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@bigtop.apache.org by Roman Shaposhnik <ro...@shaposhnik.org> on 2014/06/27 00:34:59 UTC

Thoughts on Bigtop 0.8.0

Hi!

based on Andrew P. suggestion I bumped the
version of JDK6 to the latest update (45) and
it seems to have cured the HBase problem.

As a side note: for a deprecated, unsupported
platform such as JDK6 there's no reason to use
anything but the latest version anyway.

But even with that, we still do have quite a few
failures in various components (Hive, Pig, etc.).

How do we attack it? Thoughts?

Thanks,
Roman.

Re: Thoughts on Bigtop 0.8.0

Posted by Konstantin Boudnik <co...@apache.org>.
That's my feeling as well - looks like Hive is till building on an older JDK
version.

On Thu, Jun 26, 2014 at 03:58PM, Mark Grover wrote:
> Thanks!
> 
> I am happy to help out. I looked at the Hive issue and it looks like we may
> be hitting http://bugs.java.com/bugdatabase/view_bug.do?bug_id=6302954.
> 
> So, if we were to upgrade that to that JDK for Hive, I have a feeling
> things would get better.
> 
> And, I am already on Crunch's case.
> 
> 
> On Thu, Jun 26, 2014 at 3:34 PM, Roman Shaposhnik <ro...@shaposhnik.org>
> wrote:
> 
> > Hi!
> >
> > based on Andrew P. suggestion I bumped the
> > version of JDK6 to the latest update (45) and
> > it seems to have cured the HBase problem.
> >
> > As a side note: for a deprecated, unsupported
> > platform such as JDK6 there's no reason to use
> > anything but the latest version anyway.
> >
> > But even with that, we still do have quite a few
> > failures in various components (Hive, Pig, etc.).
> >
> > How do we attack it? Thoughts?
> >
> > Thanks,
> > Roman.
> >

Re: Thoughts on Bigtop 0.8.0

Posted by Mark Grover <ma...@apache.org>.
Thanks!

I am happy to help out. I looked at the Hive issue and it looks like we may
be hitting http://bugs.java.com/bugdatabase/view_bug.do?bug_id=6302954.

So, if we were to upgrade that to that JDK for Hive, I have a feeling
things would get better.

And, I am already on Crunch's case.


On Thu, Jun 26, 2014 at 3:34 PM, Roman Shaposhnik <ro...@shaposhnik.org>
wrote:

> Hi!
>
> based on Andrew P. suggestion I bumped the
> version of JDK6 to the latest update (45) and
> it seems to have cured the HBase problem.
>
> As a side note: for a deprecated, unsupported
> platform such as JDK6 there's no reason to use
> anything but the latest version anyway.
>
> But even with that, we still do have quite a few
> failures in various components (Hive, Pig, etc.).
>
> How do we attack it? Thoughts?
>
> Thanks,
> Roman.
>

Re: Thoughts on Bigtop 0.8.0

Posted by Roman Shaposhnik <ro...@shaposhnik.org>.
Seems like we're all in agreement then:
   https://issues.apache.org/jira/browse/BIGTOP-1358

Thanks,
Roman.

On Tue, Jul 1, 2014 at 10:39 AM, Mark Grover <ma...@apache.org> wrote:
> +1 given the problems described.
>
>
> On Mon, Jun 30, 2014 at 10:13 PM, jay vyas <ja...@gmail.com>
> wrote:
>
>> +1 for jdk 7 !
>>
>>
>> On Mon, Jun 30, 2014 at 10:34 PM, Konstantin Boudnik <co...@apache.org>
>> wrote:
>>
>> > Yup, +1
>> >
>> > On Sun, Jun 29, 2014 at 05:19PM, Roman Shaposhnik wrote:
>> > > On Thu, Jun 26, 2014 at 7:03 PM, Andrew Purtell <ap...@apache.org>
>> > wrote:
>> > > > With a bit of hacking on the build we could cross-compile to avoid
>> the
>> > type
>> > > > related build failures that are creeping in as projects don't try
>> > compiling
>> > > > with JDK 6 any longer.
>> > > >
>> > > >     $ javac -target 1.6 -bootclasspath /path/to/jdk/1.6/lib/rt.jar
>> > > >
>> > > > Nasty to require a JDK 6 and 7 for building, though.
>> > >
>> > > At this point, I'd suggest biting the bullet and declaring 0.8.0
>> > > JDK7+ only. Even for build. Three of the dozen or so projects
>> > > already refuse to be built by anything but:
>> > >    * Hue 3.6 (can be hacked to stay with JDK6, but who will do it?)
>> > >    * Solr 4.9.0 (ditto)
>> > >    * Giraph 1.1.0 (not sure whether JDK7 specific functionality
>> remains)
>> > >
>> > > So, unless somebody volunteers to solve all three of the above,
>> > > I think 0.8.0 is about to move to JDK7.
>> > >
>> > > Scream within a week or forever...you know.
>> > >
>> > > Thanks,
>> > > Roman.
>> >
>>
>>
>>
>> --
>> jay vyas
>>

Re: Thoughts on Bigtop 0.8.0

Posted by Mark Grover <ma...@apache.org>.
+1 given the problems described.


On Mon, Jun 30, 2014 at 10:13 PM, jay vyas <ja...@gmail.com>
wrote:

> +1 for jdk 7 !
>
>
> On Mon, Jun 30, 2014 at 10:34 PM, Konstantin Boudnik <co...@apache.org>
> wrote:
>
> > Yup, +1
> >
> > On Sun, Jun 29, 2014 at 05:19PM, Roman Shaposhnik wrote:
> > > On Thu, Jun 26, 2014 at 7:03 PM, Andrew Purtell <ap...@apache.org>
> > wrote:
> > > > With a bit of hacking on the build we could cross-compile to avoid
> the
> > type
> > > > related build failures that are creeping in as projects don't try
> > compiling
> > > > with JDK 6 any longer.
> > > >
> > > >     $ javac -target 1.6 -bootclasspath /path/to/jdk/1.6/lib/rt.jar
> > > >
> > > > Nasty to require a JDK 6 and 7 for building, though.
> > >
> > > At this point, I'd suggest biting the bullet and declaring 0.8.0
> > > JDK7+ only. Even for build. Three of the dozen or so projects
> > > already refuse to be built by anything but:
> > >    * Hue 3.6 (can be hacked to stay with JDK6, but who will do it?)
> > >    * Solr 4.9.0 (ditto)
> > >    * Giraph 1.1.0 (not sure whether JDK7 specific functionality
> remains)
> > >
> > > So, unless somebody volunteers to solve all three of the above,
> > > I think 0.8.0 is about to move to JDK7.
> > >
> > > Scream within a week or forever...you know.
> > >
> > > Thanks,
> > > Roman.
> >
>
>
>
> --
> jay vyas
>

Re: Thoughts on Bigtop 0.8.0

Posted by jay vyas <ja...@gmail.com>.
+1 for jdk 7 !


On Mon, Jun 30, 2014 at 10:34 PM, Konstantin Boudnik <co...@apache.org> wrote:

> Yup, +1
>
> On Sun, Jun 29, 2014 at 05:19PM, Roman Shaposhnik wrote:
> > On Thu, Jun 26, 2014 at 7:03 PM, Andrew Purtell <ap...@apache.org>
> wrote:
> > > With a bit of hacking on the build we could cross-compile to avoid the
> type
> > > related build failures that are creeping in as projects don't try
> compiling
> > > with JDK 6 any longer.
> > >
> > >     $ javac -target 1.6 -bootclasspath /path/to/jdk/1.6/lib/rt.jar
> > >
> > > Nasty to require a JDK 6 and 7 for building, though.
> >
> > At this point, I'd suggest biting the bullet and declaring 0.8.0
> > JDK7+ only. Even for build. Three of the dozen or so projects
> > already refuse to be built by anything but:
> >    * Hue 3.6 (can be hacked to stay with JDK6, but who will do it?)
> >    * Solr 4.9.0 (ditto)
> >    * Giraph 1.1.0 (not sure whether JDK7 specific functionality remains)
> >
> > So, unless somebody volunteers to solve all three of the above,
> > I think 0.8.0 is about to move to JDK7.
> >
> > Scream within a week or forever...you know.
> >
> > Thanks,
> > Roman.
>



-- 
jay vyas

Re: Thoughts on Bigtop 0.8.0

Posted by Konstantin Boudnik <co...@apache.org>.
Yup, +1

On Sun, Jun 29, 2014 at 05:19PM, Roman Shaposhnik wrote:
> On Thu, Jun 26, 2014 at 7:03 PM, Andrew Purtell <ap...@apache.org> wrote:
> > With a bit of hacking on the build we could cross-compile to avoid the type
> > related build failures that are creeping in as projects don't try compiling
> > with JDK 6 any longer.
> >
> >     $ javac -target 1.6 -bootclasspath /path/to/jdk/1.6/lib/rt.jar
> >
> > Nasty to require a JDK 6 and 7 for building, though.
> 
> At this point, I'd suggest biting the bullet and declaring 0.8.0
> JDK7+ only. Even for build. Three of the dozen or so projects
> already refuse to be built by anything but:
>    * Hue 3.6 (can be hacked to stay with JDK6, but who will do it?)
>    * Solr 4.9.0 (ditto)
>    * Giraph 1.1.0 (not sure whether JDK7 specific functionality remains)
> 
> So, unless somebody volunteers to solve all three of the above,
> I think 0.8.0 is about to move to JDK7.
> 
> Scream within a week or forever...you know.
> 
> Thanks,
> Roman.

Re: Thoughts on Bigtop 0.8.0

Posted by Andrew Purtell <ap...@apache.org>.
On Sun, Jun 29, 2014 at 5:19 PM, Roman Shaposhnik <ro...@shaposhnik.org>
wrote:

> At this point, I'd suggest biting the bullet and declaring 0.8.0
> JDK7+ only. Even for build.
>

​+1 here​



-- 
Best regards,

   - Andy

Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)

Re: Thoughts on Bigtop 0.8.0

Posted by Roman Shaposhnik <ro...@shaposhnik.org>.
On Thu, Jun 26, 2014 at 7:03 PM, Andrew Purtell <ap...@apache.org> wrote:
> With a bit of hacking on the build we could cross-compile to avoid the type
> related build failures that are creeping in as projects don't try compiling
> with JDK 6 any longer.
>
>     $ javac -target 1.6 -bootclasspath /path/to/jdk/1.6/lib/rt.jar
>
> Nasty to require a JDK 6 and 7 for building, though.

At this point, I'd suggest biting the bullet and declaring 0.8.0
JDK7+ only. Even for build. Three of the dozen or so projects
already refuse to be built by anything but:
   * Hue 3.6 (can be hacked to stay with JDK6, but who will do it?)
   * Solr 4.9.0 (ditto)
   * Giraph 1.1.0 (not sure whether JDK7 specific functionality remains)

So, unless somebody volunteers to solve all three of the above,
I think 0.8.0 is about to move to JDK7.

Scream within a week or forever...you know.

Thanks,
Roman.

Re: Thoughts on Bigtop 0.8.0

Posted by Andrew Purtell <ap...@apache.org>.
With a bit of hacking on the build we could cross-compile to avoid the type
related build failures that are creeping in as projects don't try compiling
with JDK 6 any longer.

    $ javac -target 1.6 -bootclasspath /path/to/jdk/1.6/lib/rt.jar

Nasty to require a JDK 6 and 7 for building, though.


On Thu, Jun 26, 2014 at 6:44 PM, Peter Linnell <pl...@apache.org> wrote:

> I thought Bigtop was the openSUSE of Big Data, moving foward without
> the  of bleeding edge ;-)
>
> All joking aside, I am with Cos on this. Just the existence of Bigtop
> has pushed the larger ecosystem to think being their own little gardens
> and that includes dropping deprecated components.
>
> I can remember not so long ago we had to drag in Java 1.5 to build docs
> for some stuff.
>
> I am confortable with our cruising speed.
>
> Thanks,
> Peter
>
>
> On Thu, 26 Jun 2014 17:15:38 -0700
> Konstantin Boudnik <co...@apache.org> wrote:
>
> > On Thu, Jun 26, 2014 at 06:59PM, jay vyas wrote:
> > > Im not casting a vote :BUT thinking out loud:
> > >
> > > If bigtop is really the "fedora" of bigdata distros, then moving to
> > > Java 7 is the natural thing to do.
> >
> > I beg to differ: Bigtop is a Debian of bigdata ;) Agree with the rest
> > though. However, the question is about timing. I'd say: we can try to
> > give it a full build attempt using nothing but JDK7 and see how the
> > things are going. If something breaks - we can do 0.8.1. which will
> > only be focused on JDK7 upgrade. Thought?
> >
> > > Fedora integrates the trailblazing linux technology - without fear
> > > of consequences.... and I see BigTop as doing the exact same thing
> > > in the hadoop community.
> > >
> > > For me -- BigTop is a platform to innovate and do cool new stuff on
> > > -- so I say keep it that way and keep fearlessly moving forward :)
> > >
> > >
> > > On Thu, Jun 26, 2014 at 6:34 PM, Roman Shaposhnik
> > > <ro...@shaposhnik.org> wrote:
> > >
> > > > Hi!
> > > >
> > > > based on Andrew P. suggestion I bumped the
> > > > version of JDK6 to the latest update (45) and
> > > > it seems to have cured the HBase problem.
> > > >
> > > > As a side note: for a deprecated, unsupported
> > > > platform such as JDK6 there's no reason to use
> > > > anything but the latest version anyway.
> > > >
> > > > But even with that, we still do have quite a few
> > > > failures in various components (Hive, Pig, etc.).
> > > >
> > > > How do we attack it? Thoughts?
> > > >
> > > > Thanks,
> > > > Roman.
> > > >
> > >
> > >
> > >
> > > --
> > > jay vyas
>
>


-- 
Best regards,

   - Andy

Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)

Re: Thoughts on Bigtop 0.8.0

Posted by Peter Linnell <pl...@apache.org>.
I thought Bigtop was the openSUSE of Big Data, moving foward without
the  of bleeding edge ;-)

All joking aside, I am with Cos on this. Just the existence of Bigtop
has pushed the larger ecosystem to think being their own little gardens
and that includes dropping deprecated components.

I can remember not so long ago we had to drag in Java 1.5 to build docs
for some stuff.

I am confortable with our cruising speed.

Thanks,
Peter 


On Thu, 26 Jun 2014 17:15:38 -0700
Konstantin Boudnik <co...@apache.org> wrote:

> On Thu, Jun 26, 2014 at 06:59PM, jay vyas wrote:
> > Im not casting a vote :BUT thinking out loud:
> > 
> > If bigtop is really the "fedora" of bigdata distros, then moving to
> > Java 7 is the natural thing to do.
> 
> I beg to differ: Bigtop is a Debian of bigdata ;) Agree with the rest
> though. However, the question is about timing. I'd say: we can try to
> give it a full build attempt using nothing but JDK7 and see how the
> things are going. If something breaks - we can do 0.8.1. which will
> only be focused on JDK7 upgrade. Thought?
> 
> > Fedora integrates the trailblazing linux technology - without fear
> > of consequences.... and I see BigTop as doing the exact same thing
> > in the hadoop community.
> > 
> > For me -- BigTop is a platform to innovate and do cool new stuff on
> > -- so I say keep it that way and keep fearlessly moving forward :)
> > 
> > 
> > On Thu, Jun 26, 2014 at 6:34 PM, Roman Shaposhnik
> > <ro...@shaposhnik.org> wrote:
> > 
> > > Hi!
> > >
> > > based on Andrew P. suggestion I bumped the
> > > version of JDK6 to the latest update (45) and
> > > it seems to have cured the HBase problem.
> > >
> > > As a side note: for a deprecated, unsupported
> > > platform such as JDK6 there's no reason to use
> > > anything but the latest version anyway.
> > >
> > > But even with that, we still do have quite a few
> > > failures in various components (Hive, Pig, etc.).
> > >
> > > How do we attack it? Thoughts?
> > >
> > > Thanks,
> > > Roman.
> > >
> > 
> > 
> > 
> > -- 
> > jay vyas


Re: Thoughts on Bigtop 0.8.0

Posted by Konstantin Boudnik <co...@apache.org>.
On Thu, Jun 26, 2014 at 06:59PM, jay vyas wrote:
> Im not casting a vote :BUT thinking out loud:
> 
> If bigtop is really the "fedora" of bigdata distros, then moving to Java 7
> is the natural thing to do.

I beg to differ: Bigtop is a Debian of bigdata ;) Agree with the rest though.
However, the question is about timing. I'd say: we can try to give it a full
build attempt using nothing but JDK7 and see how the things are going. If
something breaks - we can do 0.8.1. which will only be focused on JDK7
upgrade. Thought?

> Fedora integrates the trailblazing linux technology - without fear of
> consequences.... and I see BigTop as doing the exact same thing in the
> hadoop community.
> 
> For me -- BigTop is a platform to innovate and do cool new stuff on -- so I
> say keep it that way and keep fearlessly moving forward :)
> 
> 
> On Thu, Jun 26, 2014 at 6:34 PM, Roman Shaposhnik <ro...@shaposhnik.org>
> wrote:
> 
> > Hi!
> >
> > based on Andrew P. suggestion I bumped the
> > version of JDK6 to the latest update (45) and
> > it seems to have cured the HBase problem.
> >
> > As a side note: for a deprecated, unsupported
> > platform such as JDK6 there's no reason to use
> > anything but the latest version anyway.
> >
> > But even with that, we still do have quite a few
> > failures in various components (Hive, Pig, etc.).
> >
> > How do we attack it? Thoughts?
> >
> > Thanks,
> > Roman.
> >
> 
> 
> 
> -- 
> jay vyas

Re: Thoughts on Bigtop 0.8.0

Posted by Mark Grover <gr...@gmail.com>.
I totally agree, I don't know if it makes sense to bring in the complexity
of switching out major java version for Bigtop 0.8 at this point.


On Thu, Jun 26, 2014 at 3:59 PM, jay vyas <ja...@gmail.com>
wrote:

> Im not casting a vote :BUT thinking out loud:
>
> If bigtop is really the "fedora" of bigdata distros, then moving to Java 7
> is the natural thing to do.
>
> Fedora integrates the trailblazing linux technology - without fear of
> consequences.... and I see BigTop as doing the exact same thing in the
> hadoop community.
>
> For me -- BigTop is a platform to innovate and do cool new stuff on -- so I
> say keep it that way and keep fearlessly moving forward :)
>
>
> On Thu, Jun 26, 2014 at 6:34 PM, Roman Shaposhnik <ro...@shaposhnik.org>
> wrote:
>
> > Hi!
> >
> > based on Andrew P. suggestion I bumped the
> > version of JDK6 to the latest update (45) and
> > it seems to have cured the HBase problem.
> >
> > As a side note: for a deprecated, unsupported
> > platform such as JDK6 there's no reason to use
> > anything but the latest version anyway.
> >
> > But even with that, we still do have quite a few
> > failures in various components (Hive, Pig, etc.).
> >
> > How do we attack it? Thoughts?
> >
> > Thanks,
> > Roman.
> >
>
>
>
> --
> jay vyas
>

Re: Thoughts on Bigtop 0.8.0

Posted by jay vyas <ja...@gmail.com>.
Im not casting a vote :BUT thinking out loud:

If bigtop is really the "fedora" of bigdata distros, then moving to Java 7
is the natural thing to do.

Fedora integrates the trailblazing linux technology - without fear of
consequences.... and I see BigTop as doing the exact same thing in the
hadoop community.

For me -- BigTop is a platform to innovate and do cool new stuff on -- so I
say keep it that way and keep fearlessly moving forward :)


On Thu, Jun 26, 2014 at 6:34 PM, Roman Shaposhnik <ro...@shaposhnik.org>
wrote:

> Hi!
>
> based on Andrew P. suggestion I bumped the
> version of JDK6 to the latest update (45) and
> it seems to have cured the HBase problem.
>
> As a side note: for a deprecated, unsupported
> platform such as JDK6 there's no reason to use
> anything but the latest version anyway.
>
> But even with that, we still do have quite a few
> failures in various components (Hive, Pig, etc.).
>
> How do we attack it? Thoughts?
>
> Thanks,
> Roman.
>



-- 
jay vyas