Re: [DISCUS] Plan the next Hive release

[Stamatis Zampetakis <za...@gmail.com>]

First of all, I would like to say a big thanks to those people who took the
time to share their opinion on this topic. I will also take the opportunity
to highlight that such discussions that directly affect the future of the
project are important thus I would expect much more participation.
Expressing agreement or disagreement with some proposal is what helps to
drive the project forward and create a healthy community. The PMC is
responsible for the oversight and the future of the project and people who
want to become part of it should not hesitate to share their opinion and
come up with their own proposals.

Indeed there is a lack of reviews cause committers have limited time and
changing priorities. I believe that the "commit and release from master"
proposal is gonna reduce the maintenance burden and thus liberate some time
from committers to focus on reviews and other important topics and
hopefully increase the release cadence.

There were some arguments about the stability of master compared to
branch-4.x which imply that there are some known regressions that are
blocking the release. If someone has a link or references to such patches
please share them here to see how we can prioritize the remaining work.

From my perspective a release does not have to be perfect, it just has to
be better than the previous one. We cannot wait to fix every major known
issue before releasing the new version because there are always new things
that pop up. If we fixed some stuff without causing regressions then this
is good enough to become a release and I am sure users will be grateful.

Best,
Stamatis


On Thu, Aug 1, 2024 at 6:35 AM Butao Zhang <bu...@163.com> wrote:

> Hi folks,
>
>              Hive-4.0 branch is a relative stable branch, so backport
> some fix and releasing a stable minor version based on Hive-4.0 is more
> reasonable to me. But as Stamatis said, `It's been already 4 months since
> the release of Hive 4.0.0`, I also think 4 months is a little long for
> waiting a minor release. The reason for not releasing the version for a
> long time, is mainly what Denys said  `many major bug fixes are simply not
> being reviewed on time.`
>
>             IMHO, we need to encourage more committers/contributors to
> take part in Hive project, including code review, doc/conf supplement,
> performance test and so on. Meanwhile, the minor release cycle should be
> less 3 months, and then we can attract more people to test the version and
> report issues as soon as possible, and i think the real users test are more
> important.
>
>
> Thanks,
> Butao Zhang
> ---- Replied Message ----
> From Zhihua Deng<de...@apache.org> <undefined>
> Date 8/1/2024 11:47
> To <de...@hive.apache.org> <de...@hive.apache.org>
> Subject Re: [DISCUS] Plan the next Hive release
> Hi All,
>
> Hive 4.0.1 Overview:
> Now there are six Hive issues that tagged with hive-4.0.1-must left
> un-merged to branch-4.0, most of them are resolved on the master. Regarding
> to TEZ-4557, a workaround in Hive is to add the missing jars to the
> hive.aux.jars.path, which is acceptable, from my perspective, I would like
> to skip including the TEZ-4557 in this release.
>
> I like the idea of cherry-picking only critical bugs and important
> improvements based on one major release, it can introduce less surprises,
> stable updates and learning cost savings. To reduce maintenance burden for
> the community, I think we can take some rules. For instance, Any release
> branch would be freeze after some time or some minors released(like
> the branch-4.x could be expired two years latter or after three minor
> releases).
>
> Regards,
> Zhihua
>
> On 2024/07/31 11:30:25 Denys Kuzmenko wrote:
>
> Hi Stamatis,
>
> `It's been already 4 months since the release of Hive 4.0.0`. That's true,
> Zhihua can give a more detailed view of things, however, the main reason is
> that it took a while to resolve the identified bugs.
> ATM the only blocking item is TEZ-4557, which might not be available until
> September, so we might decide to skip it or try to workaround in Hive if
> possible.
>
> Hive-4.0 branch was well-tested by the community and is in a much better
> shape than the master.
> For Hive-4.1 release plan was to migrate to JDK-17+ which might take more
> time.
>
> `Commit and release always from master.` I don't think that's a good idea
> as it won't solve the mentioned here reason for the change: increase
> release cadence
>
> I feel the current problem is the lack of support from committers, as many
> major bug fixes are simply not being reviewed on time.
>
> Regards,
> Denys
>
>
>

Re: [DISCUS] Plan the next Hive release

[Stamatis Zampetakis <za...@gmail.com>]

I went quickly over the list of know issues shared by Denys [1] but I
don't see something there that is a regression on master (affects
master but doesn't affect 4.0.0). In other words, it is good to solve
these issues ASAP but till this happens we can still cut a new
release.

If we run the TPC-DS workload we should do it against master rather
than branch-4.0.1. It's better to focus our efforts on changes that
have more impact.

The JDK-17 has not landed yet and it may still take some time till it
gets merged. If we cut a release from master now (or in the next few
days) we are not affected by the upgrade.

[1] https://issues.apache.org/jira/issues/?jql=labels%20%3D%20hive-4.1.0-must%20and%20resolution%20%3D%20Unresolved

On Mon, Sep 9, 2024 at 12:36 PM Denys Kuzmenko <dk...@apache.org> wrote:
>
> JDK-17 support should only be part of 4.1+ releases, giving customers time to upgrade, while still getting critical fixes in the 4.0 line.

Re: [DISCUS] Plan the next Hive release

[Denys Kuzmenko <dk...@apache.org>]

JDK-17 support should only be part of 4.1+ releases, giving customers time to upgrade, while still getting critical fixes in the 4.0 line.

Re: [DISCUS] Plan the next Hive release

[Denys Kuzmenko <dk...@apache.org>]

Just a few known issues in master:
https://issues.apache.org/jira/issues/?jql=labels%20%3D%20hive-4.1.0-must%20and%20resolution%20%3D%20Unresolved

To get a better view of things, we need to execute TPC-DS workload against branch-4.0.1 and master.