Helm chart or Kubernetes manifests for Hive Metastore

[Shohei Okumiya <ok...@apache.org>]

Hi Hive community members,

Each Hive release ships HiveServer2 and standalone HMS to DockerHub
today. This excellent work has significantly improved the experience
of locally testing Hive.

As a next step, should we officially provide k8s manifests of Hive
Metastore? I suspect many environments use the most common container
orchestration tool to run Docker images. The official setup enables
these users to test the latest HMS in their development environments
and ultimately in production environments, which demonstrates that HMS
is cloud-native. As HMS is almost stateless, it is likely low-hanging
fruit.

Any opinions?

Regards,
Okumin

Re: Helm chart or Kubernetes manifests for Hive Metastore

[Shohei Okumiya <ok...@apache.org>]

As the first step, I created two tickets and started working on the
first one. I believe we will obtain the test capabilities and can
develop/release images more confidently.

- HIVE-29261: Clean up GitHub Actions for Docker Release
- HIVE-29260: Add smoke tests for Docker images

Regards,
Okumin

On Thu, Oct 9, 2025 at 5:09 PM Shohei Okumiya <ok...@apache.org> wrote:
>
> Hi Stamatis,
>
> Thanks for the reaction and feedback. I agree that testing them is an
> essential process. I will survey how we can reasonably test Docker
> images first.
>
> Regards,
> Okumin
>
> On Wed, Oct 8, 2025 at 4:45 PM Stamatis Zampetakis <za...@gmail.com> wrote:
> >
> > Hey Okumin,
> >
> > Any kind of improvement that facilitates the deployment of Hive is
> > definitely nice to have. The tricky part is that if we claim that
> > something is officially supported and made to work in a certain way we
> > should have some tests around it. The Dockerfiles that we have are
> > super helpful yet we have zero tests covering this aspect and finding
> > problems relies exclusively on people manually checking them and doing
> > arbitrary checks during the release. All in all, if the new manifests
> > come along with some automated testing to ensure that the deployment
> > is functional then this would be a very valuable contribution.
> >
> > Best,
> > Stamatis
> >
> > On Mon, Oct 6, 2025 at 7:11 AM Shohei Okumiya <ok...@apache.org> wrote:
> > >
> > > Hi Hive community members,
> > >
> > > Each Hive release ships HiveServer2 and standalone HMS to DockerHub
> > > today. This excellent work has significantly improved the experience
> > > of locally testing Hive.
> > >
> > > As a next step, should we officially provide k8s manifests of Hive
> > > Metastore? I suspect many environments use the most common container
> > > orchestration tool to run Docker images. The official setup enables
> > > these users to test the latest HMS in their development environments
> > > and ultimately in production environments, which demonstrates that HMS
> > > is cloud-native. As HMS is almost stateless, it is likely low-hanging
> > > fruit.
> > >
> > > Any opinions?
> > >
> > > Regards,
> > > Okumin

Re: Helm chart or Kubernetes manifests for Hive Metastore

[Shohei Okumiya <ok...@apache.org>]

Hi Stamatis,

Thanks for the reaction and feedback. I agree that testing them is an
essential process. I will survey how we can reasonably test Docker
images first.

Regards,
Okumin

On Wed, Oct 8, 2025 at 4:45 PM Stamatis Zampetakis <za...@gmail.com> wrote:
>
> Hey Okumin,
>
> Any kind of improvement that facilitates the deployment of Hive is
> definitely nice to have. The tricky part is that if we claim that
> something is officially supported and made to work in a certain way we
> should have some tests around it. The Dockerfiles that we have are
> super helpful yet we have zero tests covering this aspect and finding
> problems relies exclusively on people manually checking them and doing
> arbitrary checks during the release. All in all, if the new manifests
> come along with some automated testing to ensure that the deployment
> is functional then this would be a very valuable contribution.
>
> Best,
> Stamatis
>
> On Mon, Oct 6, 2025 at 7:11 AM Shohei Okumiya <ok...@apache.org> wrote:
> >
> > Hi Hive community members,
> >
> > Each Hive release ships HiveServer2 and standalone HMS to DockerHub
> > today. This excellent work has significantly improved the experience
> > of locally testing Hive.
> >
> > As a next step, should we officially provide k8s manifests of Hive
> > Metastore? I suspect many environments use the most common container
> > orchestration tool to run Docker images. The official setup enables
> > these users to test the latest HMS in their development environments
> > and ultimately in production environments, which demonstrates that HMS
> > is cloud-native. As HMS is almost stateless, it is likely low-hanging
> > fruit.
> >
> > Any opinions?
> >
> > Regards,
> > Okumin

Re: Helm chart or Kubernetes manifests for Hive Metastore

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

Hey Okumin,

Any kind of improvement that facilitates the deployment of Hive is
definitely nice to have. The tricky part is that if we claim that
something is officially supported and made to work in a certain way we
should have some tests around it. The Dockerfiles that we have are
super helpful yet we have zero tests covering this aspect and finding
problems relies exclusively on people manually checking them and doing
arbitrary checks during the release. All in all, if the new manifests
come along with some automated testing to ensure that the deployment
is functional then this would be a very valuable contribution.

Best,
Stamatis

On Mon, Oct 6, 2025 at 7:11 AM Shohei Okumiya <ok...@apache.org> wrote:
>
> Hi Hive community members,
>
> Each Hive release ships HiveServer2 and standalone HMS to DockerHub
> today. This excellent work has significantly improved the experience
> of locally testing Hive.
>
> As a next step, should we officially provide k8s manifests of Hive
> Metastore? I suspect many environments use the most common container
> orchestration tool to run Docker images. The official setup enables
> these users to test the latest HMS in their development environments
> and ultimately in production environments, which demonstrates that HMS
> is cloud-native. As HMS is almost stateless, it is likely low-hanging
> fruit.
>
> Any opinions?
>
> Regards,
> Okumin