You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by Ayush Saxena <ay...@gmail.com> on 2026/03/10 06:01:34 UTC
Re: [DISCUSS] Publish nightly or snapshot Docker images
I think we should use a single repository for now and rely on a proper
suffix such as 4.3.0-SNAPSHOT. That version can be updated as new snapshots
are produced.
Maintaining a separate repository would add some operational overhead, and
given that we don’t have a large number of volunteers maintaining things,
keeping everything in one place seems more practical. From a user
perspective as well, this approach is simpler—users would only need to
switch the snapshot version in their export HIVE_VERSION command to start
using the newer build.
If there are no strong objections to this approach, I suggest we proceed
with it for now. If later we feel there is a clear need for a separate
repository, we can always revisit and move in that direction.
-Ayush
On Wed, 18 Feb 2026 at 15:27, Shohei Okumiya <ok...@apache.org> wrote:
> Hi,
>
> I'm glad to see many positive reactions and information new to me. Thank
> you!
>
> We should probably discuss which repository to use for this new
> purpose. I believe we can reuse apache/hive as long as we use a proper
> tag name. It is also an option to use apache/hive-nightly or something
> like Apache Solr. I am currently biased toward having a separate
> repository, i.e., apache/hive-nightly, for two reasons. From Docker
> users' perspective, pushing all in a single repo will likely obscure
> stable tags such as 4.2.0. From the committers' perspective,
> separating a repo would make management tasks, such as expiring old
> images, easier and safer. It's not a strong opinion.
>
> Once it's decided, I can try to create a pull request (and an INFRA
> ticket if we need a new repo), and we can discuss other details, such
> as naming conventions or frequency, there.
>
> Best,
> Okumin
>
> On Wed, Feb 18, 2026 at 4:39 PM Ayush Saxena <ay...@gmail.com> wrote:
> >
> > +1.
> >
> > I believe there is a policy covering this (likely in the incubator
> documentation), though I can’t locate it at the moment. As far as I
> remember, publishing such images is allowed provided they are suffixed with
> -SNAPSHOT and do not use tags like latest.
> >
> > I’ll look for the exact reference and share it here once I find it.
> >
> > -Ayush
> >
> > On Tue, 17 Feb 2026 at 17:52, Attila Turoczy <at...@cloudera.com>
> wrote:
> >>
> >> +1
> >> do not want to repeat why it is important and valuable. Very cool idea!
> >> What is the next step to make it happen? :)
> >>
> >> -Attila
> >>
> >> On Sun, Feb 15, 2026 at 3:22 PM Denys Kuzmenko <dk...@apache.org>
> wrote:
> >>>
> >>> +1 for the initiative!
> >>>
> >>> I think having nightly/snapshot Docker images would be really useful
> for Hive developers. It allows quick testing of the latest master, easier
> development of container-related features, and keeps the Docker build
> pipeline actively tested.
>
Re: [DISCUSS] Publish nightly or snapshot Docker images
Posted by Shohei Okumiya <ok...@apache.org>.
Hi Hive developers,
We eventually started publishing apache/hive:nightly and apache/hive:
standalone-metastore-nightly! This is just a start. Please feel free
to propose any improvements you have.
https://hub.docker.com/r/apache/hive/tags
Note that they are intended only for development purposes. Don't share
it with end users. It's still convenient for us. I used it right away.
https://issues.apache.org/jira/browse/HIVE-29563
I appreciate Denys reviewing this and follow-up tickets!
Best,
Okumin
On Sun, Mar 22, 2026 at 6:38 PM Shohei Okumiya <ok...@apache.org> wrote:
>
> Thanks for your feedback! Let's start with the apache/hive repo.
>
> > Just one question, Would this overwrite the same hive snapshot image on every nightly run or would we have have multiple nightly images , one for every run? The second one will quickly pollute the hive docker hub with too many images.
>
> I prefer to overwrite it at this point because adding a unique postfix
> later should be easy enough. If some of us already have a good reason
> to have a unique tag, we can go with it, though.
>
> I created a pull request. Please feel free to give me further feedback here.
> https://github.com/apache/hive/pull/6387
>
> Best,
> Okumin
>
>
> On Tue, Mar 10, 2026 at 6:08 PM Denys Kuzmenko <dk...@apache.org> wrote:
> >
> > +1 for the `-SNAPSHOT` suffix and single repo
Re: [DISCUSS] Publish nightly or snapshot Docker images
Posted by Shohei Okumiya <ok...@apache.org>.
Thanks for your feedback! Let's start with the apache/hive repo.
> Just one question, Would this overwrite the same hive snapshot image on every nightly run or would we have have multiple nightly images , one for every run? The second one will quickly pollute the hive docker hub with too many images.
I prefer to overwrite it at this point because adding a unique postfix
later should be easy enough. If some of us already have a good reason
to have a unique tag, we can go with it, though.
I created a pull request. Please feel free to give me further feedback here.
https://github.com/apache/hive/pull/6387
Best,
Okumin
On Tue, Mar 10, 2026 at 6:08 PM Denys Kuzmenko <dk...@apache.org> wrote:
>
> +1 for the `-SNAPSHOT` suffix and single repo
Re: [DISCUSS] Publish nightly or snapshot Docker images
Posted by Denys Kuzmenko <dk...@apache.org>.
+1 for the `-SNAPSHOT` suffix and single repo
Re: [DISCUSS] Publish nightly or snapshot Docker images
Posted by Simhadri G <si...@apache.org>.
Yes, i think it is better to use the same docker hub repo.
Just one question, Would this overwrite the same hive snapshot image on
every nightly run or would we have have multiple nightly images , one for
every run? The second one will quickly pollute the hive docker hub with too
many images.
Thanks
Simhadri G
On Tue, Mar 10, 2026, 1:24 PM Stamatis Zampetakis <za...@gmail.com> wrote:
> I agree, we can use the existing repository and just add an appropriate
> suffix.
>
> Best,
> Stamatis
>
> On Tue, Mar 10, 2026 at 7:03 AM Ayush Saxena <ay...@gmail.com> wrote:
> >
> > I think we should use a single repository for now and rely on a proper
> suffix such as 4.3.0-SNAPSHOT. That version can be updated as new snapshots
> are produced.
> >
> > Maintaining a separate repository would add some operational overhead,
> and given that we don’t have a large number of volunteers maintaining
> things, keeping everything in one place seems more practical. From a user
> perspective as well, this approach is simpler—users would only need to
> switch the snapshot version in their export HIVE_VERSION command to start
> using the newer build.
> >
> > If there are no strong objections to this approach, I suggest we proceed
> with it for now. If later we feel there is a clear need for a separate
> repository, we can always revisit and move in that direction.
> >
> > -Ayush
> >
> > On Wed, 18 Feb 2026 at 15:27, Shohei Okumiya <ok...@apache.org> wrote:
> >>
> >> Hi,
> >>
> >> I'm glad to see many positive reactions and information new to me.
> Thank you!
> >>
> >> We should probably discuss which repository to use for this new
> >> purpose. I believe we can reuse apache/hive as long as we use a proper
> >> tag name. It is also an option to use apache/hive-nightly or something
> >> like Apache Solr. I am currently biased toward having a separate
> >> repository, i.e., apache/hive-nightly, for two reasons. From Docker
> >> users' perspective, pushing all in a single repo will likely obscure
> >> stable tags such as 4.2.0. From the committers' perspective,
> >> separating a repo would make management tasks, such as expiring old
> >> images, easier and safer. It's not a strong opinion.
> >>
> >> Once it's decided, I can try to create a pull request (and an INFRA
> >> ticket if we need a new repo), and we can discuss other details, such
> >> as naming conventions or frequency, there.
> >>
> >> Best,
> >> Okumin
> >>
> >> On Wed, Feb 18, 2026 at 4:39 PM Ayush Saxena <ay...@gmail.com>
> wrote:
> >> >
> >> > +1.
> >> >
> >> > I believe there is a policy covering this (likely in the incubator
> documentation), though I can’t locate it at the moment. As far as I
> remember, publishing such images is allowed provided they are suffixed with
> -SNAPSHOT and do not use tags like latest.
> >> >
> >> > I’ll look for the exact reference and share it here once I find it.
> >> >
> >> > -Ayush
> >> >
> >> > On Tue, 17 Feb 2026 at 17:52, Attila Turoczy <at...@cloudera.com>
> wrote:
> >> >>
> >> >> +1
> >> >> do not want to repeat why it is important and valuable. Very cool
> idea!
> >> >> What is the next step to make it happen? :)
> >> >>
> >> >> -Attila
> >> >>
> >> >> On Sun, Feb 15, 2026 at 3:22 PM Denys Kuzmenko <dk...@apache.org>
> wrote:
> >> >>>
> >> >>> +1 for the initiative!
> >> >>>
> >> >>> I think having nightly/snapshot Docker images would be really
> useful for Hive developers. It allows quick testing of the latest master,
> easier development of container-related features, and keeps the Docker
> build pipeline actively tested.
>
Re: [DISCUSS] Publish nightly or snapshot Docker images
Posted by Stamatis Zampetakis <za...@gmail.com>.
I agree, we can use the existing repository and just add an appropriate suffix.
Best,
Stamatis
On Tue, Mar 10, 2026 at 7:03 AM Ayush Saxena <ay...@gmail.com> wrote:
>
> I think we should use a single repository for now and rely on a proper suffix such as 4.3.0-SNAPSHOT. That version can be updated as new snapshots are produced.
>
> Maintaining a separate repository would add some operational overhead, and given that we don’t have a large number of volunteers maintaining things, keeping everything in one place seems more practical. From a user perspective as well, this approach is simpler—users would only need to switch the snapshot version in their export HIVE_VERSION command to start using the newer build.
>
> If there are no strong objections to this approach, I suggest we proceed with it for now. If later we feel there is a clear need for a separate repository, we can always revisit and move in that direction.
>
> -Ayush
>
> On Wed, 18 Feb 2026 at 15:27, Shohei Okumiya <ok...@apache.org> wrote:
>>
>> Hi,
>>
>> I'm glad to see many positive reactions and information new to me. Thank you!
>>
>> We should probably discuss which repository to use for this new
>> purpose. I believe we can reuse apache/hive as long as we use a proper
>> tag name. It is also an option to use apache/hive-nightly or something
>> like Apache Solr. I am currently biased toward having a separate
>> repository, i.e., apache/hive-nightly, for two reasons. From Docker
>> users' perspective, pushing all in a single repo will likely obscure
>> stable tags such as 4.2.0. From the committers' perspective,
>> separating a repo would make management tasks, such as expiring old
>> images, easier and safer. It's not a strong opinion.
>>
>> Once it's decided, I can try to create a pull request (and an INFRA
>> ticket if we need a new repo), and we can discuss other details, such
>> as naming conventions or frequency, there.
>>
>> Best,
>> Okumin
>>
>> On Wed, Feb 18, 2026 at 4:39 PM Ayush Saxena <ay...@gmail.com> wrote:
>> >
>> > +1.
>> >
>> > I believe there is a policy covering this (likely in the incubator documentation), though I can’t locate it at the moment. As far as I remember, publishing such images is allowed provided they are suffixed with -SNAPSHOT and do not use tags like latest.
>> >
>> > I’ll look for the exact reference and share it here once I find it.
>> >
>> > -Ayush
>> >
>> > On Tue, 17 Feb 2026 at 17:52, Attila Turoczy <at...@cloudera.com> wrote:
>> >>
>> >> +1
>> >> do not want to repeat why it is important and valuable. Very cool idea!
>> >> What is the next step to make it happen? :)
>> >>
>> >> -Attila
>> >>
>> >> On Sun, Feb 15, 2026 at 3:22 PM Denys Kuzmenko <dk...@apache.org> wrote:
>> >>>
>> >>> +1 for the initiative!
>> >>>
>> >>> I think having nightly/snapshot Docker images would be really useful for Hive developers. It allows quick testing of the latest master, easier development of container-related features, and keeps the Docker build pipeline actively tested.