hadoop-yarn-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ayush Saxena <ayush...@gmail.com>
Subject Re: [DISCUSS] which release lines should we still consider actively maintained?
Date Mon, 24 May 2021 15:46:15 GMT
+1, to mark 3.1.x EOL.
Apache Hive does depends on 3.1.0 as of now, but due to guave upgrade on branch-3.1, the attempt
to migrate to latest 3.1.x didn’t work for me atleast couple of months back. So, mostly
3.3.1 would be the only option replacing 3.1.0 there or at worst 3.3.2 in a couple of months.


-Ayush

> On 24-May-2021, at 8:43 PM, Arpit Agarwal <aagarwal@cloudera.com.invalid> wrote:
> 
> +1 to EOL 3.1.x at least.
> 
> 
>> On May 23, 2021, at 9:51 PM, Wei-Chiu Chuang <weichiu@cloudera.com.INVALID>
wrote:
>> 
>> Sean,
>> 
>> For reasons I don't understand, I never received emails from your new
>> address in the mailing list. Only Akira's response.
>> 
>> I was just able to start a thread like this.
>> 
>> I am +1 to EOL 3.1.5.
>> Reason? Spark is already on Hadoop 3.2. Hive and Tez are actively working
>> to support Hadoop 3.3. HBase supports Hadoop 3.3 already. They are the most
>> common Hadoop applications so I think a 3.1 isn't that necessarily
>> important.
>> 
>> With Hadoop 3.3.1, we have a number of improvements to support a better
>> HDFS upgrade experience, so upgrading from Hadoop 3.1 should be relatively
>> easy. Application upgrade takes some effort though (commons-lang ->
>> commons-lang3 migration for example)
>> I've been maintaining the HDFS code in branch-3.1, so from a
>> HDFS perspective the branch is always in a ready to release state.
>> 
>> The Hadoop 3.1 line is more than 3 years old. Maintaining this branch is
>> getting trickier. I am +100 to reduce the number of actively maintained
>> release line. IMO, 2 Hadoop 3 lines + 1 Hadoop 2 line is a good idea.
>> 
>> 
>> 
>> For Hadoop 3.3 line: If no one beats me, I plan to make a 3.3.2 in 2-3
>> months. And another one in another 2-3 months.
>> The Hadoop 3.3.1 has nearly 700 commits not in 3.3.0. It is very difficult
>> to make/validate a maint release with such a big divergence in the code.
>> 
>> 
>>> On Mon, May 24, 2021 at 12:06 PM Akira Ajisaka <aajisaka@apache.org <mailto:aajisaka@apache.org>>
wrote:
>>> 
>>> Hi Sean,
>>> 
>>> Thank you for starting the discussion.
>>> 
>>> I think branch-2.10, branch-3.1, branch-3.2, branch-3.3, and trunk
>>> (3.4.x) are actively maintained.
>>> 
>>> The next releases will be:
>>> - 3.4.0
>>> - 3.3.1 (Thanks, Wei-Chiu!)
>>> - 3.2.3
>>> - 3.1.5
>>> - 2.10.2
>>> 
>>>> Are there folks willing to go through being release managers to get more
>>> of these release lines on a steady cadence?
>>> 
>>> Now I'm interested in becoming a release manager of 3.1.5.
>>> 
>>>> If I were to take up maintenance release for one of them which should it
>>> be?
>>> 
>>> 3.2.3 or 2.10.2 seems to be a good choice.
>>> 
>>>> Should we declare to our downstream users that some of these lines
>>> aren’t going to get more releases?
>>> 
>>> Now I think we don't need to declare that. I believe 3.3.1, 3.2.3,
>>> 3.1.5, and 2.10.2 will be released in the near future.
>>> There are some earlier discussions of 3.1.x EoL, so 3.1.5 may be a
>>> final release of the 3.1.x release line.
>>> 
>>>> Is there downstream facing documentation somewhere that I missed for
>>> setting expectations about our release cadence and actively maintained
>>> branches?
>>> 
>>> As you commented, the confluence wiki pages for Hadoop releases were
>>> out of date. Updated [1].
>>> 
>>>> Do we have a backlog of work written up that could make the release
>>> process easier for our release managers?
>>> 
>>> The release process is documented and maintained:
>>> https://cwiki.apache.org/confluence/display/HADOOP2/HowToRelease
>>> Also, there are some backlogs [1], [2].
>>> 
>>> [1]:
>>> https://cwiki.apache.org/confluence/display/HADOOP/Hadoop+Active+Release+Lines
>>> [2]: https://cwiki.apache.org/confluence/display/HADOOP/Roadmap
>>> 
>>> Thanks,
>>> Akira
>>> 
>>> On Fri, May 21, 2021 at 7:12 AM Sean Busbey <sbusbey@apple.com.invalid>
>>> wrote:
>>>> 
>>>> 
>>>> Hi folks!
>>>> 
>>>> Which release lines do we as a community still consider actively
>>> maintained?
>>>> 
>>>> I found an earlier discussion[1] where we had consensus to consider
>>> branches that don’t get maintenance releases on a regular basis end-of-life
>>> for practical purposes. The result of that discussion was written up in our
>>> wiki docs in the “EOL Release Branches” page, summarized here
>>>> 
>>>>> If no volunteer to do a maintenance release in a short to mid-term
>>> (like 3 months to 1 or 1.5 year).
>>>> 
>>>> Looking at release lines that are still on our download page[3]:
>>>> 
>>>> * Hadoop 2.10.z - last release 8 months ago
>>>> * Hadoop 3.1.z - last release 9.5 months ago
>>>> * Hadoop 3.2.z - last release 4.5 months ago
>>>> * Hadoop 3.3.z - last release 10 months ago
>>>> 
>>>> And then trunk holds 3.4 which hasn’t had a release since the branch-3.3
>>> fork ~14 months ago.
>>>> 
>>>> I can see that Wei-Chiu has been actively working on getting the 3.3.1
>>> release out[4] (thanks Wei-Chiu!) but I do not see anything similar for the
>>> other release lines.
>>>> 
>>>> We also have pages on the wiki for our project roadmap of release[5],
>>> but it seems out of date since it lists in progress releases that have
>>> happened or branches we have announced as end of life, i.e. 2.8.
>>>> 
>>>> We also have a group of pages (sorry, I’m not sure what the confluence
>>> jargon is for this) for “hadoop active release lines”[6] but this list has
>>> 2.8, 2.9, 3.0, 3.1, and 3.3. So several declared end of life lines and no
>>> 2.10 or 3.2 despite those being our release lines with the most recent
>>> releases.
>>>> 
>>>> Are there folks willing to go through being release managers to get more
>>> of these release lines on a steady cadence?
>>>> 
>>>> If I were to take up maintenance release for one of them which should it
>>> be?
>>>> 
>>>> Should we declare to our downstream users that some of these lines
>>> aren’t going to get more releases?
>>>> 
>>>> Is there downstream facing documentation somewhere that I missed for
>>> setting expectations about our release cadence and actively maintained
>>> branches?
>>>> 
>>>> Do we have a backlog of work written up that could make the release
>>> process easier for our release managers?
>>>> 
>>>> 
>>>> [1]: https://s.apache.org/7c8jt
>>>> [2]: https://s.apache.org/4no96
>>>> [3]: https://hadoop.apache.org/releases.html
>>>> [4]: https://s.apache.org/1bvwe
>>>> [5]: https://cwiki.apache.org/confluence/display/HADOOP/Roadmap
>>>> [6]:
>>> https://cwiki.apache.org/confluence/display/HADOOP/Hadoop+Active+Release+Lines
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: yarn-dev-unsubscribe@hadoop.apache.org
>>>> For additional commands, e-mail: yarn-dev-help@hadoop.apache.org
>>>> 
>>> 
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: yarn-dev-unsubscribe@hadoop.apache.org <mailto:yarn-dev-unsubscribe@hadoop.apache.org>
>>> For additional commands, e-mail: yarn-dev-help@hadoop.apache.org <mailto:yarn-dev-help@hadoop.apache.org>

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-dev-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-dev-help@hadoop.apache.org


Mime
View raw message