branch-1 policy for builds against Hadoop 3.0.0

[Andrew Purtell <ap...@apache.org>]

Does anyone view compilation problems against Hadoop 3.0.0 in branch-1 as a
release blocker? See HBASE-19421.

In my view
​, use of Hadoop 3 with branch-1 HBase is not something we are going to
recommend for production use for some time, and Hadoop 3.0.0 is not even GA
yet. (I know it may be close.) Therefore, builds that work with it are nice
to have, but not essential. Compilation problems found for e.g. 1.4.0 can
be fixed for subsequent release 1.4.1.

​Please speak up if you disagree.  Let's discuss this before we have a RC
vote where it may come up.

--
Best regards,
Andrew

Words like orphans lost among the crosstalk, meaning torn from truth's
decrepit hands
   - A23, Crosstalk

Re: branch-1 policy for builds against Hadoop 3.0.0

[Andrew Purtell <ap...@apache.org>]

Thanks Josh.

I've fixed the build so it at least compiles against Hadoop 3.0.0, and I
think that is good enough. We do not need to claim anything better than
that for branch-1.4 or branch-1 at this time.

On Mon, Dec 4, 2017 at 3:30 PM, Josh Elser <el...@apache.org> wrote:

> +1 not release blockers
>
> This stems from this line of thinking:
>
> * Hadoop does not dictate HBase's release schedule (except in extenuating
> circumstances)
> * HBase can't ensure compatibility against something that does not yet
> exist
>
>
> On 12/4/17 3:43 PM, Andrew Purtell wrote:
>
>> Does anyone view compilation problems against Hadoop 3.0.0 in branch-1 as
>> a
>> release blocker? See HBASE-19421.
>>
>> In my view
>> ​, use of Hadoop 3 with branch-1 HBase is not something we are going to
>> recommend for production use for some time, and Hadoop 3.0.0 is not even
>> GA
>> yet. (I know it may be close.) Therefore, builds that work with it are
>> nice
>> to have, but not essential. Compilation problems found for e.g. 1.4.0 can
>> be fixed for subsequent release 1.4.1.
>>
>> ​Please speak up if you disagree.  Let's discuss this before we have a RC
>> vote where it may come up.
>>
>> --
>> Best regards,
>> Andrew
>>
>> Words like orphans lost among the crosstalk, meaning torn from truth's
>> decrepit hands
>>     - A23, Crosstalk
>>
>>


-- 
Best regards,
Andrew

Words like orphans lost among the crosstalk, meaning torn from truth's
decrepit hands
   - A23, Crosstalk

Re: branch-1 policy for builds against Hadoop 3.0.0

[Nick Dimiduk <nd...@gmail.com>]

On Mon, Dec 4, 2017 at 3:30 PM, Josh Elser <el...@apache.org> wrote:

> +1 not release blockers
>
> This stems from this line of thinking:
>
> * Hadoop does not dictate HBase's release schedule (except in extenuating
> circumstances)
> * HBase can't ensure compatibility against something that does not yet
> exist


Agreed.

On 12/4/17 3:43 PM, Andrew Purtell wrote:
>
>> Does anyone view compilation problems against Hadoop 3.0.0 in branch-1 as
>> a
>> release blocker? See HBASE-19421.
>>
>> In my view
>> ​, use of Hadoop 3 with branch-1 HBase is not something we are going to
>> recommend for production use for some time, and Hadoop 3.0.0 is not even
>> GA
>> yet. (I know it may be close.) Therefore, builds that work with it are
>> nice
>> to have, but not essential. Compilation problems found for e.g. 1.4.0 can
>> be fixed for subsequent release 1.4.1.
>>
>> ​Please speak up if you disagree.  Let's discuss this before we have a RC
>> vote where it may come up.
>>
>> --
>> Best regards,
>> Andrew
>>
>> Words like orphans lost among the crosstalk, meaning torn from truth's
>> decrepit hands
>>     - A23, Crosstalk
>>
>>

Re: branch-1 policy for builds against Hadoop 3.0.0

[Josh Elser <el...@apache.org>]

+1 not release blockers

This stems from this line of thinking:

* Hadoop does not dictate HBase's release schedule (except in 
extenuating circumstances)
* HBase can't ensure compatibility against something that does not yet exist

On 12/4/17 3:43 PM, Andrew Purtell wrote:
> Does anyone view compilation problems against Hadoop 3.0.0 in branch-1 as a
> release blocker? See HBASE-19421.
> 
> In my view
> ​, use of Hadoop 3 with branch-1 HBase is not something we are going to
> recommend for production use for some time, and Hadoop 3.0.0 is not even GA
> yet. (I know it may be close.) Therefore, builds that work with it are nice
> to have, but not essential. Compilation problems found for e.g. 1.4.0 can
> be fixed for subsequent release 1.4.1.
> 
> ​Please speak up if you disagree.  Let's discuss this before we have a RC
> vote where it may come up.
> 
> --
> Best regards,
> Andrew
> 
> Words like orphans lost among the crosstalk, meaning torn from truth's
> decrepit hands
>     - A23, Crosstalk
>