Release tag for 0.10.1

[Nick Dimiduk-2]

Hi Devs,

It seems no release tag was pushed to 0.10.1. I presume this was an
oversight. Is there some place I can look to see from which sha the 0.10.1
release was built? Are the RC vote threads the only cannon in this matter?

Thanks,
Nick

[Nick Dimiduk-2]

An only-slightly related question: Is Flink using Hadoop version specific
features in some way? IIRC, the basic APIs should be compatible back as far
as 2.2. I'm surprised to see builds of flink explicitly against many hadoop
versions, but 2.5.x is excluded.

-n

On Fri, Jan 8, 2016 at 9:45 AM, Nick Dimiduk <[hidden email]> wrote:

> Hi Devs,
>
> It seems no release tag was pushed to 0.10.1. I presume this was an
> oversight. Is there some place I can look to see from which sha the 0.10.1
> release was built? Are the RC vote threads the only cannon in this matter?
>
> Thanks,
> Nick
>

[Stephan Ewen]

Hi Nick!

We have not pushed a release tag, but have a frozen release-0.10.1-RC1
branch (https://github.com/apache/flink/tree/release-0.10.1-rc1)
A tag would be great, agree!

Flink does in its core not depend on Hadoop. The parts that reference
Hadoop (HDFS filesystem, YARN, MapReduce function/format compatibility) are
not using any Hadoop version specific code (with the exception of some YARN
functions, which are reflectively invoked). So it should work across
versions nicely. The main friction we saw were version clashes of
transitive dependencies.

The Flink CI builds include building Flink with Hadoop 2.5.0, see here:
https://github.com/apache/flink/blob/master/.travis.yml

Greetings,
Stephan



On Fri, Jan 8, 2016 at 6:54 PM, Nick Dimiduk <[hidden email]> wrote:

> An only-slightly related question: Is Flink using Hadoop version specific
> features in some way? IIRC, the basic APIs should be compatible back as far
> as 2.2. I'm surprised to see builds of flink explicitly against many hadoop
> versions, but 2.5.x is excluded.
>
> -n
>
> On Fri, Jan 8, 2016 at 9:45 AM, Nick Dimiduk <[hidden email]> wrote:
>
> > Hi Devs,
> >
> > It seems no release tag was pushed to 0.10.1. I presume this was an
> > oversight. Is there some place I can look to see from which sha the
> 0.10.1
> > release was built? Are the RC vote threads the only cannon in this
> matter?
> >
> > Thanks,
> > Nick
> >
>

[Nick Dimiduk]

Yes, a tag would be very good practice, IMHO. Those of us who need to run
release + patches appreciate the release hygiene :)

If all builds are created equal re: Hadoop versions, I recommend against
publishing Hadoop-specific tarballs on the downloads page; it left me quite
confused, as I'm sure it would other users.

On Friday, January 8, 2016, Stephan Ewen <[hidden email]> wrote:

> Hi Nick!
>
> We have not pushed a release tag, but have a frozen release-0.10.1-RC1
> branch (https://github.com/apache/flink/tree/release-0.10.1-rc1)
> A tag would be great, agree!
>
> Flink does in its core not depend on Hadoop. The parts that reference
> Hadoop (HDFS filesystem, YARN, MapReduce function/format compatibility) are
> not using any Hadoop version specific code (with the exception of some YARN
> functions, which are reflectively invoked). So it should work across
> versions nicely. The main friction we saw were version clashes of
> transitive dependencies.
>
> The Flink CI builds include building Flink with Hadoop 2.5.0, see here:
> https://github.com/apache/flink/blob/master/.travis.yml
>
> Greetings,
> Stephan
>
>
>
> On Fri, Jan 8, 2016 at 6:54 PM, Nick Dimiduk <[hidden email]
> <javascript:;>> wrote:
>
> > An only-slightly related question: Is Flink using Hadoop version specific
> > features in some way? IIRC, the basic APIs should be compatible back as
> far
> > as 2.2. I'm surprised to see builds of flink explicitly against many
> hadoop
> > versions, but 2.5.x is excluded.
> >
> > -n
> >
> > On Fri, Jan 8, 2016 at 9:45 AM, Nick Dimiduk <[hidden email]
> <javascript:;>> wrote:
> >
> > > Hi Devs,
> > >
> > > It seems no release tag was pushed to 0.10.1. I presume this was an
> > > oversight. Is there some place I can look to see from which sha the
> > 0.10.1
> > > release was built? Are the RC vote threads the only cannon in this
> > matter?
> > >
> > > Thanks,
> > > Nick
> > >
> >
>

[Stephan Ewen]

All build are built equal against Hadoop, but a specific Hadoop version is
still part of the Flink lib folder when downloaded.

Cross Hadoop-version compatibility is not good, i.e., when Flink has Hadoop
2.4 in the classpath, it does not work with a 2.6 YARN installation. That
is why we pre-build multiple versions.

Best,
Stephan


On Sat, Jan 9, 2016 at 3:19 AM, Nick Dimiduk <[hidden email]> wrote:

> Yes, a tag would be very good practice, IMHO. Those of us who need to run
> release + patches appreciate the release hygiene :)
>
> If all builds are created equal re: Hadoop versions, I recommend against
> publishing Hadoop-specific tarballs on the downloads page; it left me quite
> confused, as I'm sure it would other users.
>
> On Friday, January 8, 2016, Stephan Ewen <[hidden email]> wrote:
>
> > Hi Nick!
> >
> > We have not pushed a release tag, but have a frozen release-0.10.1-RC1
> > branch (https://github.com/apache/flink/tree/release-0.10.1-rc1)
> > A tag would be great, agree!
> >
> > Flink does in its core not depend on Hadoop. The parts that reference
> > Hadoop (HDFS filesystem, YARN, MapReduce function/format compatibility)
> are
> > not using any Hadoop version specific code (with the exception of some
> YARN
> > functions, which are reflectively invoked). So it should work across
> > versions nicely. The main friction we saw were version clashes of
> > transitive dependencies.
> >
> > The Flink CI builds include building Flink with Hadoop 2.5.0, see here:
> > https://github.com/apache/flink/blob/master/.travis.yml
> >
> > Greetings,
> > Stephan
> >
> >
> >
> > On Fri, Jan 8, 2016 at 6:54 PM, Nick Dimiduk <[hidden email]
> > <javascript:;>> wrote:
> >
> > > An only-slightly related question: Is Flink using Hadoop version
> specific
> > > features in some way? IIRC, the basic APIs should be compatible back as
> > far
> > > as 2.2. I'm surprised to see builds of flink explicitly against many
> > hadoop
> > > versions, but 2.5.x is excluded.
> > >
> > > -n
> > >
> > > On Fri, Jan 8, 2016 at 9:45 AM, Nick Dimiduk <[hidden email]
> > <javascript:;>> wrote:
> > >
> > > > Hi Devs,
> > > >
> > > > It seems no release tag was pushed to 0.10.1. I presume this was an
> > > > oversight. Is there some place I can look to see from which sha the
> > > 0.10.1
> > > > release was built? Are the RC vote threads the only cannon in this
> > > matter?
> > > >
> > > > Thanks,
> > > > Nick
> > > >
> > >
> >
>

[mxm]

Hi Nick,

That was an oversight when the release was created. As Stephan
mentioned, we have a policy that the corresponding final release
branch is read-only. Creating the tag is just a formality but of
course important. I've pushed a 'release-0.10.1' release tag. The
corresponding hash is 2e9b231.

Cheers,
Max

On Mon, Jan 11, 2016 at 12:47 AM, Stephan Ewen <[hidden email]> wrote:

> All build are built equal against Hadoop, but a specific Hadoop version is
> still part of the Flink lib folder when downloaded.
>
> Cross Hadoop-version compatibility is not good, i.e., when Flink has Hadoop
> 2.4 in the classpath, it does not work with a 2.6 YARN installation. That
> is why we pre-build multiple versions.
>
> Best,
> Stephan
>
>
> On Sat, Jan 9, 2016 at 3:19 AM, Nick Dimiduk <[hidden email]> wrote:
>
>> Yes, a tag would be very good practice, IMHO. Those of us who need to run
>> release + patches appreciate the release hygiene :)
>>
>> If all builds are created equal re: Hadoop versions, I recommend against
>> publishing Hadoop-specific tarballs on the downloads page; it left me quite
>> confused, as I'm sure it would other users.
>>
>> On Friday, January 8, 2016, Stephan Ewen <[hidden email]> wrote:
>>
>> > Hi Nick!
>> >
>> > We have not pushed a release tag, but have a frozen release-0.10.1-RC1
>> > branch (https://github.com/apache/flink/tree/release-0.10.1-rc1)
>> > A tag would be great, agree!
>> >
>> > Flink does in its core not depend on Hadoop. The parts that reference
>> > Hadoop (HDFS filesystem, YARN, MapReduce function/format compatibility)
>> are
>> > not using any Hadoop version specific code (with the exception of some
>> YARN
>> > functions, which are reflectively invoked). So it should work across
>> > versions nicely. The main friction we saw were version clashes of
>> > transitive dependencies.
>> >
>> > The Flink CI builds include building Flink with Hadoop 2.5.0, see here:
>> > https://github.com/apache/flink/blob/master/.travis.yml
>> >
>> > Greetings,
>> > Stephan
>> >
>> >
>> >
>> > On Fri, Jan 8, 2016 at 6:54 PM, Nick Dimiduk <[hidden email]
>> > <javascript:;>> wrote:
>> >
>> > > An only-slightly related question: Is Flink using Hadoop version
>> specific
>> > > features in some way? IIRC, the basic APIs should be compatible back as
>> > far
>> > > as 2.2. I'm surprised to see builds of flink explicitly against many
>> > hadoop
>> > > versions, but 2.5.x is excluded.
>> > >
>> > > -n
>> > >
>> > > On Fri, Jan 8, 2016 at 9:45 AM, Nick Dimiduk <[hidden email]
>> > <javascript:;>> wrote:
>> > >
>> > > > Hi Devs,
>> > > >
>> > > > It seems no release tag was pushed to 0.10.1. I presume this was an
>> > > > oversight. Is there some place I can look to see from which sha the
>> > > 0.10.1
>> > > > release was built? Are the RC vote threads the only cannon in this
>> > > matter?
>> > > >
>> > > > Thanks,
>> > > > Nick
>> > > >
>> > >
>> >
>>

[Nick Dimiduk]

Thanks Max. I'm accustomed to projects advertising a release with a fixed
ref such as a sha or tag, not a branch. Much obliged.

-n

On Friday, January 15, 2016, Maximilian Michels <[hidden email]> wrote:

> Hi Nick,
>
> That was an oversight when the release was created. As Stephan
> mentioned, we have a policy that the corresponding final release
> branch is read-only. Creating the tag is just a formality but of
> course important. I've pushed a 'release-0.10.1' release tag. The
> corresponding hash is 2e9b231.
>
> Cheers,
> Max
>
> On Mon, Jan 11, 2016 at 12:47 AM, Stephan Ewen <[hidden email]
> <javascript:;>> wrote:
> > All build are built equal against Hadoop, but a specific Hadoop version
> is
> > still part of the Flink lib folder when downloaded.
> >
> > Cross Hadoop-version compatibility is not good, i.e., when Flink has
> Hadoop
> > 2.4 in the classpath, it does not work with a 2.6 YARN installation. That
> > is why we pre-build multiple versions.
> >
> > Best,
> > Stephan
> >
> >
> > On Sat, Jan 9, 2016 at 3:19 AM, Nick Dimiduk <[hidden email]
> <javascript:;>> wrote:
> >
> >> Yes, a tag would be very good practice, IMHO. Those of us who need to
> run
> >> release + patches appreciate the release hygiene :)
> >>
> >> If all builds are created equal re: Hadoop versions, I recommend against
> >> publishing Hadoop-specific tarballs on the downloads page; it left me
> quite
> >> confused, as I'm sure it would other users.
> >>
> >> On Friday, January 8, 2016, Stephan Ewen <[hidden email]
> <javascript:;>> wrote:
> >>
> >> > Hi Nick!
> >> >
> >> > We have not pushed a release tag, but have a frozen release-0.10.1-RC1
> >> > branch (https://github.com/apache/flink/tree/release-0.10.1-rc1)
> >> > A tag would be great, agree!
> >> >
> >> > Flink does in its core not depend on Hadoop. The parts that reference
> >> > Hadoop (HDFS filesystem, YARN, MapReduce function/format
> compatibility)
> >> are
> >> > not using any Hadoop version specific code (with the exception of some
> >> YARN
> >> > functions, which are reflectively invoked). So it should work across
> >> > versions nicely. The main friction we saw were version clashes of
> >> > transitive dependencies.
> >> >
> >> > The Flink CI builds include building Flink with Hadoop 2.5.0, see
> here:
> >> > https://github.com/apache/flink/blob/master/.travis.yml
> >> >
> >> > Greetings,
> >> > Stephan
> >> >
> >> >
> >> >
> >> > On Fri, Jan 8, 2016 at 6:54 PM, Nick Dimiduk <[hidden email]
> <javascript:;>
> >> > <javascript:;>> wrote:
> >> >
> >> > > An only-slightly related question: Is Flink using Hadoop version
> >> specific
> >> > > features in some way? IIRC, the basic APIs should be compatible
> back as
> >> > far
> >> > > as 2.2. I'm surprised to see builds of flink explicitly against many
> >> > hadoop
> >> > > versions, but 2.5.x is excluded.
> >> > >
> >> > > -n
> >> > >
> >> > > On Fri, Jan 8, 2016 at 9:45 AM, Nick Dimiduk <[hidden email]
> <javascript:;>
> >> > <javascript:;>> wrote:
> >> > >
> >> > > > Hi Devs,
> >> > > >
> >> > > > It seems no release tag was pushed to 0.10.1. I presume this was
> an
> >> > > > oversight. Is there some place I can look to see from which sha
> the
> >> > > 0.10.1
> >> > > > release was built? Are the RC vote threads the only cannon in this
> >> > > matter?
> >> > > >
> >> > > > Thanks,
> >> > > > Nick
> >> > > >
> >> > >
> >> >
> >>
>