You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by Enis Söztutar <en...@apache.org> on 2014/03/01 01:55:26 UTC
Hadoop1 support in 0.98/1.0
Hi,
The matrix in http://hbase.apache.org/book/configuration.html shows that
0.98 DOES NOT support hadoop-1.
I though we kept the support in 0.98. We have the build profile and jenkins
build, etc. Did we decide to drop support. Maybe I am misremembering.
I'll update that chart otherwise.
While we are at it, should 1.0 support hadoop1 or not. I think it would be
good to keep support for h1 in 1.0, and drop it in 2.0 line. But it would
mean if we do not drop support, we have to keep the support through 1.1,
1.2, etc.
Enis
Re: Hadoop1 support in 0.98/1.0
Posted by Enis Söztutar <en...@gmail.com>.
Thanks Biju.
I just sent out a heads-up announcement to user@.
Enis
On Wed, Mar 5, 2014 at 6:00 PM, Biju G.S Nair <bi...@acm.org> wrote:
> If we haven't done already, as some body suggested earlier, it would be
> good to communicate this to the "user" mailing list as well. This will help
> anyone using HBase to plan to move to hadoop 2.
>
>
>
> Thanks,
> Biju
> Tel#: 978-707-5066
>
>
> On Wed, Mar 5, 2014 at 8:16 PM, Enis Söztutar <en...@gmail.com> wrote:
>
> > Ok, the consensus seems to be to drop the support. I am all in favor of
> > less overhead, but was initially concerned about leaving some of the
> users
> > behind. If we are ok with that, lets pull the trigger.
> >
> > Opened an issue for doc and tasks :
> > https://issues.apache.org/jira/browse/HBASE-10690. Marked as critical
> for
> > 0.99.
> >
> > Enis
> >
> >
> > On Wed, Mar 5, 2014 at 5:06 PM, Ted Yu <yu...@gmail.com> wrote:
> >
> > > bq. Maybe a new hadoop-compat module for 2.3.0 +?
> > >
> > > +1
> > >
> > >
> > > On Wed, Mar 5, 2014 at 5:02 PM, Andrew Purtell <ap...@apache.org>
> > > wrote:
> > >
> > > > On Thu, Mar 6, 2014 at 3:39 AM, Stack <st...@duboce.net> wrote:
> > > >
> > > > > But if we instead allow that our versioning currently is of-kilter
> --
> > > > Lars
> > > > > Hofhansl has argued off-line that 0.96.0 should have been 1.0
> > > > >
> > > >
> > > > Related, is Hadoop 2.3.0 more like 3.0 ?
> > > >
> > > > If, for example, we wanted to pin the HFiles of IN_MEMORY tables in
> > > HDFS's
> > > > centralized cache, as HDFS-4949 suggests, then how different will our
> > > view
> > > > of the HDFS interfaces be between 2.0.0 and 2.3.0+? Would the (I'm
> > > > guessing) necessary reflection be a perf issue? Maybe a new
> > hadoop-compat
> > > > module for 2.3.0 +? Or drop support for Hadoop < 2.3.0 ?
> > > >
> > > > Point is - after we finish this discussion about Hadoop 1 or not
> (seems
> > > > not), then what to do about the different flavors of Hadoop 2.
> > > >
> > > > --
> > > > Best regards,
> > > >
> > > > - Andy
> > > >
> > > > Problems worthy of attack prove their worth by hitting back. - Piet
> > Hein
> > > > (via Tom White)
> > > >
> > >
> >
>
Re: Hadoop1 support in 0.98/1.0
Posted by "Biju G.S Nair" <bi...@acm.org>.
If we haven't done already, as some body suggested earlier, it would be
good to communicate this to the "user" mailing list as well. This will help
anyone using HBase to plan to move to hadoop 2.
Thanks,
Biju
Tel#: 978-707-5066
On Wed, Mar 5, 2014 at 8:16 PM, Enis Söztutar <en...@gmail.com> wrote:
> Ok, the consensus seems to be to drop the support. I am all in favor of
> less overhead, but was initially concerned about leaving some of the users
> behind. If we are ok with that, lets pull the trigger.
>
> Opened an issue for doc and tasks :
> https://issues.apache.org/jira/browse/HBASE-10690. Marked as critical for
> 0.99.
>
> Enis
>
>
> On Wed, Mar 5, 2014 at 5:06 PM, Ted Yu <yu...@gmail.com> wrote:
>
> > bq. Maybe a new hadoop-compat module for 2.3.0 +?
> >
> > +1
> >
> >
> > On Wed, Mar 5, 2014 at 5:02 PM, Andrew Purtell <ap...@apache.org>
> > wrote:
> >
> > > On Thu, Mar 6, 2014 at 3:39 AM, Stack <st...@duboce.net> wrote:
> > >
> > > > But if we instead allow that our versioning currently is of-kilter --
> > > Lars
> > > > Hofhansl has argued off-line that 0.96.0 should have been 1.0
> > > >
> > >
> > > Related, is Hadoop 2.3.0 more like 3.0 ?
> > >
> > > If, for example, we wanted to pin the HFiles of IN_MEMORY tables in
> > HDFS's
> > > centralized cache, as HDFS-4949 suggests, then how different will our
> > view
> > > of the HDFS interfaces be between 2.0.0 and 2.3.0+? Would the (I'm
> > > guessing) necessary reflection be a perf issue? Maybe a new
> hadoop-compat
> > > module for 2.3.0 +? Or drop support for Hadoop < 2.3.0 ?
> > >
> > > Point is - after we finish this discussion about Hadoop 1 or not (seems
> > > not), then what to do about the different flavors of Hadoop 2.
> > >
> > > --
> > > Best regards,
> > >
> > > - Andy
> > >
> > > Problems worthy of attack prove their worth by hitting back. - Piet
> Hein
> > > (via Tom White)
> > >
> >
>
Re: Hadoop1 support in 0.98/1.0
Posted by Enis Söztutar <en...@gmail.com>.
+1.
On Wed, Mar 19, 2014 at 5:48 PM, Ted Yu <yu...@gmail.com> wrote:
> Now that hadoop-1 support is dropped from trunk,
> should HBase-TRUNK-on-Hadoop-1.1 build be disabled ?
>
> Cheers
>
>
> On Wed, Mar 5, 2014 at 5:16 PM, Enis Söztutar <en...@gmail.com> wrote:
>
> > Ok, the consensus seems to be to drop the support. I am all in favor of
> > less overhead, but was initially concerned about leaving some of the
> users
> > behind. If we are ok with that, lets pull the trigger.
> >
> > Opened an issue for doc and tasks :
> > https://issues.apache.org/jira/browse/HBASE-10690. Marked as critical
> for
> > 0.99.
> >
> > Enis
> >
> >
> > On Wed, Mar 5, 2014 at 5:06 PM, Ted Yu <yu...@gmail.com> wrote:
> >
> > > bq. Maybe a new hadoop-compat module for 2.3.0 +?
> > >
> > > +1
> > >
> > >
> > > On Wed, Mar 5, 2014 at 5:02 PM, Andrew Purtell <ap...@apache.org>
> > > wrote:
> > >
> > > > On Thu, Mar 6, 2014 at 3:39 AM, Stack <st...@duboce.net> wrote:
> > > >
> > > > > But if we instead allow that our versioning currently is of-kilter
> --
> > > > Lars
> > > > > Hofhansl has argued off-line that 0.96.0 should have been 1.0
> > > > >
> > > >
> > > > Related, is Hadoop 2.3.0 more like 3.0 ?
> > > >
> > > > If, for example, we wanted to pin the HFiles of IN_MEMORY tables in
> > > HDFS's
> > > > centralized cache, as HDFS-4949 suggests, then how different will our
> > > view
> > > > of the HDFS interfaces be between 2.0.0 and 2.3.0+? Would the (I'm
> > > > guessing) necessary reflection be a perf issue? Maybe a new
> > hadoop-compat
> > > > module for 2.3.0 +? Or drop support for Hadoop < 2.3.0 ?
> > > >
> > > > Point is - after we finish this discussion about Hadoop 1 or not
> (seems
> > > > not), then what to do about the different flavors of Hadoop 2.
> > > >
> > > > --
> > > > Best regards,
> > > >
> > > > - Andy
> > > >
> > > > Problems worthy of attack prove their worth by hitting back. - Piet
> > Hein
> > > > (via Tom White)
> > > >
> > >
> >
>
Re: Hadoop1 support in 0.98/1.0
Posted by Ted Yu <yu...@gmail.com>.
Now that hadoop-1 support is dropped from trunk,
should HBase-TRUNK-on-Hadoop-1.1 build be disabled ?
Cheers
On Wed, Mar 5, 2014 at 5:16 PM, Enis Söztutar <en...@gmail.com> wrote:
> Ok, the consensus seems to be to drop the support. I am all in favor of
> less overhead, but was initially concerned about leaving some of the users
> behind. If we are ok with that, lets pull the trigger.
>
> Opened an issue for doc and tasks :
> https://issues.apache.org/jira/browse/HBASE-10690. Marked as critical for
> 0.99.
>
> Enis
>
>
> On Wed, Mar 5, 2014 at 5:06 PM, Ted Yu <yu...@gmail.com> wrote:
>
> > bq. Maybe a new hadoop-compat module for 2.3.0 +?
> >
> > +1
> >
> >
> > On Wed, Mar 5, 2014 at 5:02 PM, Andrew Purtell <ap...@apache.org>
> > wrote:
> >
> > > On Thu, Mar 6, 2014 at 3:39 AM, Stack <st...@duboce.net> wrote:
> > >
> > > > But if we instead allow that our versioning currently is of-kilter --
> > > Lars
> > > > Hofhansl has argued off-line that 0.96.0 should have been 1.0
> > > >
> > >
> > > Related, is Hadoop 2.3.0 more like 3.0 ?
> > >
> > > If, for example, we wanted to pin the HFiles of IN_MEMORY tables in
> > HDFS's
> > > centralized cache, as HDFS-4949 suggests, then how different will our
> > view
> > > of the HDFS interfaces be between 2.0.0 and 2.3.0+? Would the (I'm
> > > guessing) necessary reflection be a perf issue? Maybe a new
> hadoop-compat
> > > module for 2.3.0 +? Or drop support for Hadoop < 2.3.0 ?
> > >
> > > Point is - after we finish this discussion about Hadoop 1 or not (seems
> > > not), then what to do about the different flavors of Hadoop 2.
> > >
> > > --
> > > Best regards,
> > >
> > > - Andy
> > >
> > > Problems worthy of attack prove their worth by hitting back. - Piet
> Hein
> > > (via Tom White)
> > >
> >
>
Re: Hadoop1 support in 0.98/1.0
Posted by Enis Söztutar <en...@gmail.com>.
Ok, the consensus seems to be to drop the support. I am all in favor of
less overhead, but was initially concerned about leaving some of the users
behind. If we are ok with that, lets pull the trigger.
Opened an issue for doc and tasks :
https://issues.apache.org/jira/browse/HBASE-10690. Marked as critical for
0.99.
Enis
On Wed, Mar 5, 2014 at 5:06 PM, Ted Yu <yu...@gmail.com> wrote:
> bq. Maybe a new hadoop-compat module for 2.3.0 +?
>
> +1
>
>
> On Wed, Mar 5, 2014 at 5:02 PM, Andrew Purtell <ap...@apache.org>
> wrote:
>
> > On Thu, Mar 6, 2014 at 3:39 AM, Stack <st...@duboce.net> wrote:
> >
> > > But if we instead allow that our versioning currently is of-kilter --
> > Lars
> > > Hofhansl has argued off-line that 0.96.0 should have been 1.0
> > >
> >
> > Related, is Hadoop 2.3.0 more like 3.0 ?
> >
> > If, for example, we wanted to pin the HFiles of IN_MEMORY tables in
> HDFS's
> > centralized cache, as HDFS-4949 suggests, then how different will our
> view
> > of the HDFS interfaces be between 2.0.0 and 2.3.0+? Would the (I'm
> > guessing) necessary reflection be a perf issue? Maybe a new hadoop-compat
> > module for 2.3.0 +? Or drop support for Hadoop < 2.3.0 ?
> >
> > Point is - after we finish this discussion about Hadoop 1 or not (seems
> > not), then what to do about the different flavors of Hadoop 2.
> >
> > --
> > Best regards,
> >
> > - Andy
> >
> > Problems worthy of attack prove their worth by hitting back. - Piet Hein
> > (via Tom White)
> >
>
Re: Hadoop1 support in 0.98/1.0
Posted by Ted Yu <yu...@gmail.com>.
bq. Maybe a new hadoop-compat module for 2.3.0 +?
+1
On Wed, Mar 5, 2014 at 5:02 PM, Andrew Purtell <ap...@apache.org> wrote:
> On Thu, Mar 6, 2014 at 3:39 AM, Stack <st...@duboce.net> wrote:
>
> > But if we instead allow that our versioning currently is of-kilter --
> Lars
> > Hofhansl has argued off-line that 0.96.0 should have been 1.0
> >
>
> Related, is Hadoop 2.3.0 more like 3.0 ?
>
> If, for example, we wanted to pin the HFiles of IN_MEMORY tables in HDFS's
> centralized cache, as HDFS-4949 suggests, then how different will our view
> of the HDFS interfaces be between 2.0.0 and 2.3.0+? Would the (I'm
> guessing) necessary reflection be a perf issue? Maybe a new hadoop-compat
> module for 2.3.0 +? Or drop support for Hadoop < 2.3.0 ?
>
> Point is - after we finish this discussion about Hadoop 1 or not (seems
> not), then what to do about the different flavors of Hadoop 2.
>
> --
> Best regards,
>
> - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein
> (via Tom White)
>
Re: Hadoop1 support in 0.98/1.0
Posted by Andrew Purtell <ap...@apache.org>.
On Thu, Mar 6, 2014 at 3:39 AM, Stack <st...@duboce.net> wrote:
> But if we instead allow that our versioning currently is of-kilter -- Lars
> Hofhansl has argued off-line that 0.96.0 should have been 1.0
>
Related, is Hadoop 2.3.0 more like 3.0 ?
If, for example, we wanted to pin the HFiles of IN_MEMORY tables in HDFS's
centralized cache, as HDFS-4949 suggests, then how different will our view
of the HDFS interfaces be between 2.0.0 and 2.3.0+? Would the (I'm
guessing) necessary reflection be a perf issue? Maybe a new hadoop-compat
module for 2.3.0 +? Or drop support for Hadoop < 2.3.0 ?
Point is - after we finish this discussion about Hadoop 1 or not (seems
not), then what to do about the different flavors of Hadoop 2.
--
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)
Re: Hadoop1 support in 0.98/1.0
Posted by Andrew Purtell <ap...@apache.org>.
On Thu, Mar 6, 2014 at 7:43 AM, Stack <st...@duboce.net> wrote:
> Suggest leaving 0.98 as is shipping hadoop1 w/ an X in its hadoop1 support
> column that we will point people at should they ask questions about 0.98 on
> hadoop1 and then, in 1.0, purge hadoop1 (give it XXX in the hadoop grid).
>
+1
--
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)
Re: Hadoop1 support in 0.98/1.0
Posted by Stack <st...@duboce.net>.
On Wed, Mar 5, 2014 at 3:29 PM, Enis Söztutar <en...@gmail.com> wrote:
...
>
> > Currently it is X. Do we want to go back up to NT?
> >
>
> I thought we did not explicitly decide to be that case (see my question in
> the first email in the thread). It is X possibly because we overlooked. We
> are still releasing hadoop-1 artifacts and build and run tests with
> hadoop1.
>
>
Pardon me.
Suggest leaving 0.98 as is shipping hadoop1 w/ an X in its hadoop1 support
column that we will point people at should they ask questions about 0.98 on
hadoop1 and then, in 1.0, purge hadoop1 (give it XXX in the hadoop grid).
St.Ack
Re: Hadoop1 support in 0.98/1.0
Posted by Enis Söztutar <en...@gmail.com>.
>
> Just to say that there is no NS, its X:
> http://hbase.apache.org/book.html#hadoop
Yep.
> Currently it is X. Do we want to go back up to NT?
>
I thought we did not explicitly decide to be that case (see my question in
the first email in the thread). It is X possibly because we overlooked. We
are still releasing hadoop-1 artifacts and build and run tests with
hadoop1.
>
>
>
> > It seems that the consensus is to either
> > go with NS or NT for 1.0. I would like to completely commit and drop
> > support completely (not releasing the artifacts, possibly breaking the
> > build with hadoop1) if we are going that route. Again it feels like not
> > enough time has passed, but if 1.0 comes with hadoop-1 support, we have
> to
> > stick with it in all of the 1.x series.
> >
>
> +1 on all-out dropping it.
>
> St.Ack
>
Re: Hadoop1 support in 0.98/1.0
Posted by Stack <st...@duboce.net>.
On Wed, Mar 5, 2014 at 3:11 PM, Enis Söztutar <en...@gmail.com> wrote:
> In the matrix, we have NS = Not supported, and NT = Not tested.
>
>
Just to say that there is no NS, its X:
http://hbase.apache.org/book.html#hadoop
> We can go 0.98 with hadoop 1 = NT.
Currently it is X. Do we want to go back up to NT?
> It seems that the consensus is to either
> go with NS or NT for 1.0. I would like to completely commit and drop
> support completely (not releasing the artifacts, possibly breaking the
> build with hadoop1) if we are going that route. Again it feels like not
> enough time has passed, but if 1.0 comes with hadoop-1 support, we have to
> stick with it in all of the 1.x series.
>
+1 on all-out dropping it.
St.Ack
Re: Hadoop1 support in 0.98/1.0
Posted by Enis Söztutar <en...@gmail.com>.
In the matrix, we have NS = Not supported, and NT = Not tested.
We can go 0.98 with hadoop 1 = NT. It seems that the consensus is to either
go with NS or NT for 1.0. I would like to completely commit and drop
support completely (not releasing the artifacts, possibly breaking the
build with hadoop1) if we are going that route. Again it feels like not
enough time has passed, but if 1.0 comes with hadoop-1 support, we have to
stick with it in all of the 1.x series.
Enis
On Wed, Mar 5, 2014 at 12:50 PM, Stack <st...@duboce.net> wrote:
> On Wed, Mar 5, 2014 at 11:51 AM, lars hofhansl <la...@apache.org> wrote:
>
> > I suppose the next question is: Do we intentionally drop it, or just not
> > spend any additional time on it? (I.e. some new features might not work
> > with Hadoop-1, etc).
> > We could keep the -hadoop1 test suites running and fix failures, but not
> > do anything beyond that.
> >
>
>
> We could.
>
> I was going to suggest "And change its state in the hadoop support grid
> from supported to not-supported" but it looks like (Andrew?) someone
> already did the honors for 0.98 on:
> http://hbase.apache.org/book.html#hadoop
>
> St.Ack
>
Re: Hadoop1 support in 0.98/1.0
Posted by Stack <st...@duboce.net>.
On Wed, Mar 5, 2014 at 11:51 AM, lars hofhansl <la...@apache.org> wrote:
> I suppose the next question is: Do we intentionally drop it, or just not
> spend any additional time on it? (I.e. some new features might not work
> with Hadoop-1, etc).
> We could keep the -hadoop1 test suites running and fix failures, but not
> do anything beyond that.
>
We could.
I was going to suggest "And change its state in the hadoop support grid
from supported to not-supported" but it looks like (Andrew?) someone
already did the honors for 0.98 on: http://hbase.apache.org/book.html#hadoop
St.Ack
Re: Hadoop1 support in 0.98/1.0
Posted by lars hofhansl <la...@apache.org>.
I suppose the next question is: Do we intentionally drop it, or just not spend any additional time on it? (I.e. some new features might not work with Hadoop-1, etc).
We could keep the -hadoop1 test suites running and fix failures, but not do anything beyond that.
-- Lars
________________________________
From: Stack <st...@duboce.net>
To: HBase Dev List <de...@hbase.apache.org>; lars hofhansl <la...@apache.org>
Sent: Wednesday, March 5, 2014 11:39 AM
Subject: Re: Hadoop1 support in 0.98/1.0
(Good discusssion)
Intellectually and if we go by the numbers, it makes 'sense' keeping
hadoop1 support in hbase1.
But if we instead allow that our versioning currently is of-kilter -- Lars
Hofhansl has argued off-line that 0.96.0 should have been 1.0 -- and if we
are realistic and ask the question, who of us here has the time and
interest supporting a hbase1 on hadoop1 (not I!, No one?), and as per Andy
above, h2 is just better all around, then IMO hadoop1 should be dropped in
1.0 (Is it too late to deprecate in 0.98?).
St.Ack
On Tue, Mar 4, 2014 at 5:45 PM, lars hofhansl <la...@apache.org> wrote:
> +1 on dropping Hadoop 1 support after 0.98.
> If we want to tie HBase performance featured a bit closer to Hadoop
> features it seems we have to do that.
>
>
> -- Lars
>
>
>
> ________________________________
> From: Andrew Purtell <ap...@apache.org>
> To: "dev@hbase.apache.org" <de...@hbase.apache.org>
> Sent: Tuesday, March 4, 2014 5:09 PM
> Subject: Re: Hadoop1 support in 0.98/1.0
>
>
> On Wed, Mar 5, 2014 at 4:57 AM, Enis Söztutar <en...@gmail.com> wrote:
>
>
> > Hadoop-2.2 which is the first GA release of Hadoop, was released in
> October
> > 2013. It is not enough time passed to drop support I feel.
> >
>
> This makes a lot of sense for 0.94. That is our (un)official (?)
> long-term-stable release.
>
> The 0.96 "singularity" was released around the same time as Hadoop 2.2,
> certainly our 0.98 was, and since the stability of the stack is the sum of
> its parts for those releases I'm not sure what Hadoop 1 offers besides the
> burden of legacy interfaces and lower performance.
>
> --
> Best regards,
>
> - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein
> (via Tom White)
>
Re: Hadoop1 support in 0.98/1.0
Posted by Stack <st...@duboce.net>.
(Good discusssion)
Intellectually and if we go by the numbers, it makes 'sense' keeping
hadoop1 support in hbase1.
But if we instead allow that our versioning currently is of-kilter -- Lars
Hofhansl has argued off-line that 0.96.0 should have been 1.0 -- and if we
are realistic and ask the question, who of us here has the time and
interest supporting a hbase1 on hadoop1 (not I!, No one?), and as per Andy
above, h2 is just better all around, then IMO hadoop1 should be dropped in
1.0 (Is it too late to deprecate in 0.98?).
St.Ack
On Tue, Mar 4, 2014 at 5:45 PM, lars hofhansl <la...@apache.org> wrote:
> +1 on dropping Hadoop 1 support after 0.98.
> If we want to tie HBase performance featured a bit closer to Hadoop
> features it seems we have to do that.
>
>
> -- Lars
>
>
>
> ________________________________
> From: Andrew Purtell <ap...@apache.org>
> To: "dev@hbase.apache.org" <de...@hbase.apache.org>
> Sent: Tuesday, March 4, 2014 5:09 PM
> Subject: Re: Hadoop1 support in 0.98/1.0
>
>
> On Wed, Mar 5, 2014 at 4:57 AM, Enis Söztutar <en...@gmail.com> wrote:
>
>
> > Hadoop-2.2 which is the first GA release of Hadoop, was released in
> October
> > 2013. It is not enough time passed to drop support I feel.
> >
>
> This makes a lot of sense for 0.94. That is our (un)official (?)
> long-term-stable release.
>
> The 0.96 "singularity" was released around the same time as Hadoop 2.2,
> certainly our 0.98 was, and since the stability of the stack is the sum of
> its parts for those releases I'm not sure what Hadoop 1 offers besides the
> burden of legacy interfaces and lower performance.
>
> --
> Best regards,
>
> - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein
> (via Tom White)
>
Re: Hadoop1 support in 0.98/1.0
Posted by lars hofhansl <la...@apache.org>.
+1 on dropping Hadoop 1 support after 0.98.
If we want to tie HBase performance featured a bit closer to Hadoop features it seems we have to do that.
-- Lars
________________________________
From: Andrew Purtell <ap...@apache.org>
To: "dev@hbase.apache.org" <de...@hbase.apache.org>
Sent: Tuesday, March 4, 2014 5:09 PM
Subject: Re: Hadoop1 support in 0.98/1.0
On Wed, Mar 5, 2014 at 4:57 AM, Enis Söztutar <en...@gmail.com> wrote:
> Hadoop-2.2 which is the first GA release of Hadoop, was released in October
> 2013. It is not enough time passed to drop support I feel.
>
This makes a lot of sense for 0.94. That is our (un)official (?)
long-term-stable release.
The 0.96 "singularity" was released around the same time as Hadoop 2.2,
certainly our 0.98 was, and since the stability of the stack is the sum of
its parts for those releases I'm not sure what Hadoop 1 offers besides the
burden of legacy interfaces and lower performance.
--
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)
Re: Hadoop1 support in 0.98/1.0
Posted by Andrew Purtell <ap...@apache.org>.
On Wed, Mar 5, 2014 at 4:57 AM, Enis Söztutar <en...@gmail.com> wrote:
> Hadoop-2.2 which is the first GA release of Hadoop, was released in October
> 2013. It is not enough time passed to drop support I feel.
>
This makes a lot of sense for 0.94. That is our (un)official (?)
long-term-stable release.
The 0.96 "singularity" was released around the same time as Hadoop 2.2,
certainly our 0.98 was, and since the stability of the stack is the sum of
its parts for those releases I'm not sure what Hadoop 1 offers besides the
burden of legacy interfaces and lower performance.
--
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)
Re: Hadoop1 support in 0.98/1.0
Posted by Enis Söztutar <en...@gmail.com>.
I would love to drop hadoop-1 support as soon as possible, but I don't
think we should do it in HBase-1.0.
Hadoop-2.2 which is the first GA release of Hadoop, was released in October
2013. It is not enough time passed to drop support I feel.
Enis
On Tue, Mar 4, 2014 at 5:07 AM, Jonathan Hsieh <jo...@cloudera.com> wrote:
> I'd be ok with saying hadoop2.0 as a min for hbase 1.0. If we can support
> fewer versions we have fewer compat cases to maintain and can clean up code
> sooner.
>
> 0.96 defaults to 1.x
> 0.98 defaults to 2.x
> trunk defaults to 2.x
>
> The major distros (CDH, HWX, Intel?) have been on hadoop2 so we'd be
> keeping support for other users. Can we do a quick survey on user@ to see
> if we should support it?
>
> Jon.
>
>
> On Fri, Feb 28, 2014 at 5:28 PM, Ted Yu <yu...@gmail.com> wrote:
>
> > bq. I'll update that chart
> >
> > Please do so.
> >
> > bq. should 1.0 support hadoop1
> >
> > hadoop1 support should be kept.
> > This would allow users whose hbase deployment only occupies a portion of
> > the whole hadoop-1 cluster flexibility of upgrading hbase only.
> >
> > Cheers
> >
> >
> > On Fri, Feb 28, 2014 at 4:55 PM, Enis Söztutar <en...@apache.org> wrote:
> >
> > > Hi,
> > >
> > > The matrix in http://hbase.apache.org/book/configuration.html shows
> that
> > > 0.98 DOES NOT support hadoop-1.
> > >
> > > I though we kept the support in 0.98. We have the build profile and
> > jenkins
> > > build, etc. Did we decide to drop support. Maybe I am misremembering.
> > >
> > > I'll update that chart otherwise.
> > >
> > > While we are at it, should 1.0 support hadoop1 or not. I think it would
> > be
> > > good to keep support for h1 in 1.0, and drop it in 2.0 line. But it
> would
> > > mean if we do not drop support, we have to keep the support through
> 1.1,
> > > 1.2, etc.
> > >
> > > Enis
> > >
> >
>
>
>
> --
> // Jonathan Hsieh (shay)
> // HBase Tech Lead, Software Engineer, Cloudera
> // jon@cloudera.com // @jmhsieh
>
Re: Hadoop1 support in 0.98/1.0
Posted by Andrew Purtell <ap...@apache.org>.
On Tue, Mar 4, 2014 at 9:07 PM, Jonathan Hsieh <jo...@cloudera.com> wrote:
>
The major distros (CDH, HWX, Intel?) have been on hadoop2 so we'd b
e
keeping support for other users.
Yes, Intel's distro is Hadoop 2 based. There is a legacy product that uses
Hadoop 1 but, legacy.
> Can we do a quick survey on user@ to see if we should support it?
+1
--
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein
(via Tom White)
Re: Hadoop1 support in 0.98/1.0
Posted by Jonathan Hsieh <jo...@cloudera.com>.
I'd be ok with saying hadoop2.0 as a min for hbase 1.0. If we can support
fewer versions we have fewer compat cases to maintain and can clean up code
sooner.
0.96 defaults to 1.x
0.98 defaults to 2.x
trunk defaults to 2.x
The major distros (CDH, HWX, Intel?) have been on hadoop2 so we'd be
keeping support for other users. Can we do a quick survey on user@ to see
if we should support it?
Jon.
On Fri, Feb 28, 2014 at 5:28 PM, Ted Yu <yu...@gmail.com> wrote:
> bq. I'll update that chart
>
> Please do so.
>
> bq. should 1.0 support hadoop1
>
> hadoop1 support should be kept.
> This would allow users whose hbase deployment only occupies a portion of
> the whole hadoop-1 cluster flexibility of upgrading hbase only.
>
> Cheers
>
>
> On Fri, Feb 28, 2014 at 4:55 PM, Enis Söztutar <en...@apache.org> wrote:
>
> > Hi,
> >
> > The matrix in http://hbase.apache.org/book/configuration.html shows that
> > 0.98 DOES NOT support hadoop-1.
> >
> > I though we kept the support in 0.98. We have the build profile and
> jenkins
> > build, etc. Did we decide to drop support. Maybe I am misremembering.
> >
> > I'll update that chart otherwise.
> >
> > While we are at it, should 1.0 support hadoop1 or not. I think it would
> be
> > good to keep support for h1 in 1.0, and drop it in 2.0 line. But it would
> > mean if we do not drop support, we have to keep the support through 1.1,
> > 1.2, etc.
> >
> > Enis
> >
>
--
// Jonathan Hsieh (shay)
// HBase Tech Lead, Software Engineer, Cloudera
// jon@cloudera.com // @jmhsieh
Re: Hadoop1 support in 0.98/1.0
Posted by Ted Yu <yu...@gmail.com>.
bq. I'll update that chart
Please do so.
bq. should 1.0 support hadoop1
hadoop1 support should be kept.
This would allow users whose hbase deployment only occupies a portion of
the whole hadoop-1 cluster flexibility of upgrading hbase only.
Cheers
On Fri, Feb 28, 2014 at 4:55 PM, Enis Söztutar <en...@apache.org> wrote:
> Hi,
>
> The matrix in http://hbase.apache.org/book/configuration.html shows that
> 0.98 DOES NOT support hadoop-1.
>
> I though we kept the support in 0.98. We have the build profile and jenkins
> build, etc. Did we decide to drop support. Maybe I am misremembering.
>
> I'll update that chart otherwise.
>
> While we are at it, should 1.0 support hadoop1 or not. I think it would be
> good to keep support for h1 in 1.0, and drop it in 2.0 line. But it would
> mean if we do not drop support, we have to keep the support through 1.1,
> 1.2, etc.
>
> Enis
>