You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@beam.apache.org by Jarek Potiuk <ja...@potiuk.com> on 2026/07/27 11:05:04 UTC

Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Hi Beam devs,

Writing from the Apache Airflow side with a short status note and one
question about 2.76.0 timing.

Can we count on ~ beginning of August release of 2.76.0 of python beam client?

Background: Airflow suspended its apache-beam provider earlier this year
[1]. We have a PR open to restore it [2] (tracking issue [3]), and at this
point the only thing still blocking us is dependency resolution.

Every released Beam up to and including 2.75.0 declares:

    envoy-data-plane<2,>=1.0.3; python_version >= "3.11"

The only release in that range is envoy-data-plane 1.0.3, which pins
betterproto==2.0.0b6. betterproto has never shipped a final 2.x - its
latest stable is 1.2.5, from 2020 - so that pin is only ever satisfiable
by a pre-release. Resolvers that do not opt into pre-releases for
transitive dependencies reject it outright.

The good news is that your team already fixed this in main.
The [4] removed envoy-data-plane and the
transitive betterproto from the Python SDK entirely, replacing them with a
small vendored Envoy RLS proto, and CHANGES.md lists it under 2.76.0. It
merged on 2026-07-07, one day after the 2.75.0 branch cut, so it just
missed that release.

So from our checks, a released 2.76.0 is the last thing standing between us
and un-suspending the provider. Once we can floor the provider at
apache-beam>=2.76.0, the problem disappears on every Python version we
support and we need no workaround on our side. Until then our options are
to wait, or to carry a local pre-release constraint that we would rather
not ship.

Hence the question: is the 2.76.0 branch cut still on for 5 August, per the
release calendar [5]? Nothing urgent, and no ask to change anything - we
just want to know whether it's reasonable to wait 2.76.0.

Our next release wave for providers is also slated for the beginning of August,
so even if we make some workaround and force installing a 2020 (!) pre-release
of betterproto release, we'd still prefer to wait for 2.76.0 as it
would be a very
temporary and pretty bad workaround.

Thanks for making the fix, and for Beam generally.

[1] https://github.com/apache/airflow/pull/61926
[2] https://github.com/apache/airflow/pull/66952
[3] https://github.com/apache/airflow/issues/66551
[4] https://github.com/apache/beam/pull/39213
[5] https://calendar.google.com/calendar/embed?src=0p73sl034k80oob7seouanigd0%40group.calendar.google.com
[6] https://cwiki.apache.org/confluence/spaces/AIRFLOW/pages/399278789/Release+Plan

Jarek Potiuk
Apache Airflow PMC

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Kenneth Knowles <ke...@apache.org>.
🎉🎉🚀🚀

On Wed, Sep 9, 2026 at 9:17 AM Jarek Potiuk <ja...@potiuk.com> wrote:

> Hello everyone.
>
> I am happy to announce that I just merged
> https://github.com/apache/airflow/pull/66952, which unsuspends the Apache
> Beam provider. Cross my fingers we don't find any blockers preventing the
> release with apache-beam 2.76.0; it should be included in the upcoming wave
> of Provider's releases (starting this week).
>
> Thanks to everyone for cooperating on this matter. This was a rather long
> one, but I think the end result is that both Airflow and Beam dependencies
> are fresher, which will also enable future "faster" dependency moves in the
> future. Given the current state of Supply Chain / Security issues (roughly
> a 10x increase in reported issues being fixed by nearly everyone),
> unblocking faster upgrade paths for our dependencies is crucial.
>
> For reference: Yesterday's Microsoft Patch Tuesday included 914 issues, 21
> of which were critical, and 2 are currently actively exploited.
> https://senserva.com/patch-tuesday.html#history
>
> This chart looks very similar across the board:
>
> [image: Screenshot 2026-09-09 at 15.16.54.png]
>
>
>
> J.
>
>
> On Mon, Jul 27, 2026 at 4:51 PM Jarek Potiuk <ja...@potiuk.com> wrote:
>
>> Yes. Our next release is planned for 4th of Aug, however, if we need to
>> support an unsuspended provider like Beam, it might be more involved. I
>> would say we will anyhow have to make an exception ad-hoc release whenever
>> Beam is released - that seems quite a bit safer than mixing it with other
>> provider releases.
>>
>> We have to account for the four-day cooldown required for installing
>> dependencies and rebasing and merging the "revert" PR and add >= 2.76.0
>> (there might be some surprises to fix - dependency resolution is too
>> complex to make 100% predictions.
>> This means that whenever Beam client 2.76.0 is released, we will either
>> need to wait 4 days or add a temporary exception to the cooldown (and
>> remove it after), and make sure that all our tests pass after we add >=
>> 2.76.0 in our deps.
>> So it's anyhow a more involved thing than a regular release.
>>
>> I would say releasing 2.76.Releasing 2.76.0 by the end of July could
>> allow us to add it to the "regular" wave to give us time to do all that -
>> so I'd say it's quite unlikely.
>>
>> J.
>>
>

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Kenneth Knowles <ke...@apache.org>.
🎉🎉🚀🚀

On Wed, Sep 9, 2026 at 9:17 AM Jarek Potiuk <ja...@potiuk.com> wrote:

> Hello everyone.
>
> I am happy to announce that I just merged
> https://github.com/apache/airflow/pull/66952, which unsuspends the Apache
> Beam provider. Cross my fingers we don't find any blockers preventing the
> release with apache-beam 2.76.0; it should be included in the upcoming wave
> of Provider's releases (starting this week).
>
> Thanks to everyone for cooperating on this matter. This was a rather long
> one, but I think the end result is that both Airflow and Beam dependencies
> are fresher, which will also enable future "faster" dependency moves in the
> future. Given the current state of Supply Chain / Security issues (roughly
> a 10x increase in reported issues being fixed by nearly everyone),
> unblocking faster upgrade paths for our dependencies is crucial.
>
> For reference: Yesterday's Microsoft Patch Tuesday included 914 issues, 21
> of which were critical, and 2 are currently actively exploited.
> https://senserva.com/patch-tuesday.html#history
>
> This chart looks very similar across the board:
>
> [image: Screenshot 2026-09-09 at 15.16.54.png]
>
>
>
> J.
>
>
> On Mon, Jul 27, 2026 at 4:51 PM Jarek Potiuk <ja...@potiuk.com> wrote:
>
>> Yes. Our next release is planned for 4th of Aug, however, if we need to
>> support an unsuspended provider like Beam, it might be more involved. I
>> would say we will anyhow have to make an exception ad-hoc release whenever
>> Beam is released - that seems quite a bit safer than mixing it with other
>> provider releases.
>>
>> We have to account for the four-day cooldown required for installing
>> dependencies and rebasing and merging the "revert" PR and add >= 2.76.0
>> (there might be some surprises to fix - dependency resolution is too
>> complex to make 100% predictions.
>> This means that whenever Beam client 2.76.0 is released, we will either
>> need to wait 4 days or add a temporary exception to the cooldown (and
>> remove it after), and make sure that all our tests pass after we add >=
>> 2.76.0 in our deps.
>> So it's anyhow a more involved thing than a regular release.
>>
>> I would say releasing 2.76.Releasing 2.76.0 by the end of July could
>> allow us to add it to the "regular" wave to give us time to do all that -
>> so I'd say it's quite unlikely.
>>
>> J.
>>
>

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Jarek Potiuk <ja...@potiuk.com>.
Hello everyone.

I am happy to announce that I just merged
https://github.com/apache/airflow/pull/66952, which unsuspends the Apache
Beam provider. Cross my fingers we don't find any blockers preventing the
release with apache-beam 2.76.0; it should be included in the upcoming wave
of Provider's releases (starting this week).

Thanks to everyone for cooperating on this matter. This was a rather long
one, but I think the end result is that both Airflow and Beam dependencies
are fresher, which will also enable future "faster" dependency moves in the
future. Given the current state of Supply Chain / Security issues (roughly
a 10x increase in reported issues being fixed by nearly everyone),
unblocking faster upgrade paths for our dependencies is crucial.

For reference: Yesterday's Microsoft Patch Tuesday included 914 issues, 21
of which were critical, and 2 are currently actively exploited.
https://senserva.com/patch-tuesday.html#history

This chart looks very similar across the board:

[image: Screenshot 2026-09-09 at 15.16.54.png]



J.


On Mon, Jul 27, 2026 at 4:51 PM Jarek Potiuk <ja...@potiuk.com> wrote:

> Yes. Our next release is planned for 4th of Aug, however, if we need to
> support an unsuspended provider like Beam, it might be more involved. I
> would say we will anyhow have to make an exception ad-hoc release whenever
> Beam is released - that seems quite a bit safer than mixing it with other
> provider releases.
>
> We have to account for the four-day cooldown required for installing
> dependencies and rebasing and merging the "revert" PR and add >= 2.76.0
> (there might be some surprises to fix - dependency resolution is too
> complex to make 100% predictions.
> This means that whenever Beam client 2.76.0 is released, we will either
> need to wait 4 days or add a temporary exception to the cooldown (and
> remove it after), and make sure that all our tests pass after we add >=
> 2.76.0 in our deps.
> So it's anyhow a more involved thing than a regular release.
>
> I would say releasing 2.76.Releasing 2.76.0 by the end of July could allow
> us to add it to the "regular" wave to give us time to do all that - so I'd
> say it's quite unlikely.
>
> J.
>

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Jarek Potiuk <ja...@potiuk.com>.
Hello everyone.

I am happy to announce that I just merged
https://github.com/apache/airflow/pull/66952, which unsuspends the Apache
Beam provider. Cross my fingers we don't find any blockers preventing the
release with apache-beam 2.76.0; it should be included in the upcoming wave
of Provider's releases (starting this week).

Thanks to everyone for cooperating on this matter. This was a rather long
one, but I think the end result is that both Airflow and Beam dependencies
are fresher, which will also enable future "faster" dependency moves in the
future. Given the current state of Supply Chain / Security issues (roughly
a 10x increase in reported issues being fixed by nearly everyone),
unblocking faster upgrade paths for our dependencies is crucial.

For reference: Yesterday's Microsoft Patch Tuesday included 914 issues, 21
of which were critical, and 2 are currently actively exploited.
https://senserva.com/patch-tuesday.html#history

This chart looks very similar across the board:

[image: Screenshot 2026-09-09 at 15.16.54.png]



J.


On Mon, Jul 27, 2026 at 4:51 PM Jarek Potiuk <ja...@potiuk.com> wrote:

> Yes. Our next release is planned for 4th of Aug, however, if we need to
> support an unsuspended provider like Beam, it might be more involved. I
> would say we will anyhow have to make an exception ad-hoc release whenever
> Beam is released - that seems quite a bit safer than mixing it with other
> provider releases.
>
> We have to account for the four-day cooldown required for installing
> dependencies and rebasing and merging the "revert" PR and add >= 2.76.0
> (there might be some surprises to fix - dependency resolution is too
> complex to make 100% predictions.
> This means that whenever Beam client 2.76.0 is released, we will either
> need to wait 4 days or add a temporary exception to the cooldown (and
> remove it after), and make sure that all our tests pass after we add >=
> 2.76.0 in our deps.
> So it's anyhow a more involved thing than a regular release.
>
> I would say releasing 2.76.Releasing 2.76.0 by the end of July could allow
> us to add it to the "regular" wave to give us time to do all that - so I'd
> say it's quite unlikely.
>
> J.
>

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Jarek Potiuk <ja...@potiuk.com>.
Yes. Our next release is planned for 4th of Aug, however, if we need to
support an unsuspended provider like Beam, it might be more involved. I
would say we will anyhow have to make an exception ad-hoc release whenever
Beam is released - that seems quite a bit safer than mixing it with other
provider releases.

We have to account for the four-day cooldown required for installing
dependencies and rebasing and merging the "revert" PR and add >= 2.76.0
(there might be some surprises to fix - dependency resolution is too
complex to make 100% predictions.
This means that whenever Beam client 2.76.0 is released, we will either
need to wait 4 days or add a temporary exception to the cooldown (and
remove it after), and make sure that all our tests pass after we add >=
2.76.0 in our deps.
So it's anyhow a more involved thing than a regular release.

I would say releasing 2.76.Releasing 2.76.0 by the end of July could allow
us to add it to the "regular" wave to give us time to do all that - so I'd
say it's quite unlikely.

J.

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Jarek Potiuk <ja...@potiuk.com>.
Yes. Our next release is planned for 4th of Aug, however, if we need to
support an unsuspended provider like Beam, it might be more involved. I
would say we will anyhow have to make an exception ad-hoc release whenever
Beam is released - that seems quite a bit safer than mixing it with other
provider releases.

We have to account for the four-day cooldown required for installing
dependencies and rebasing and merging the "revert" PR and add >= 2.76.0
(there might be some surprises to fix - dependency resolution is too
complex to make 100% predictions.
This means that whenever Beam client 2.76.0 is released, we will either
need to wait 4 days or add a temporary exception to the cooldown (and
remove it after), and make sure that all our tests pass after we add >=
2.76.0 in our deps.
So it's anyhow a more involved thing than a regular release.

I would say releasing 2.76.Releasing 2.76.0 by the end of July could allow
us to add it to the "regular" wave to give us time to do all that - so I'd
say it's quite unlikely.

J.

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Elad Kalif <el...@apache.org>.
> Is there a
particular timeline we need to hit in order to make your next provider
release?

Airflow providers are released every 2 weeks
These are the planned dates
https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=399278789#ReleasePlan-ProviderReleases

On Mon, Jul 27, 2026 at 4:10 PM Danny McCormick via dev <
dev@airflow.apache.org> wrote:

> Thanks for following up on this!
>
> > Can we count on ~ beginning of August release of 2.76.0 of python beam
> client?
>
> Yes, we should be releasing in early August as you noted. Is there a
> particular timeline we need to hit in order to make your next provider
> release?
>
> Thanks,
> Danny
>
> On Mon, Jul 27, 2026 at 7:05 AM Jarek Potiuk <ja...@potiuk.com> wrote:
>
> > Hi Beam devs,
> >
> > Writing from the Apache Airflow side with a short status note and one
> > question about 2.76.0 timing.
> >
> > Can we count on ~ beginning of August release of 2.76.0 of python beam
> > client?
> >
> > Background: Airflow suspended its apache-beam provider earlier this year
> > [1]. We have a PR open to restore it [2] (tracking issue [3]), and at
> this
> > point the only thing still blocking us is dependency resolution.
> >
> > Every released Beam up to and including 2.75.0 declares:
> >
> >     envoy-data-plane<2,>=1.0.3; python_version >= "3.11"
> >
> > The only release in that range is envoy-data-plane 1.0.3, which pins
> > betterproto==2.0.0b6. betterproto has never shipped a final 2.x - its
> > latest stable is 1.2.5, from 2020 - so that pin is only ever satisfiable
> > by a pre-release. Resolvers that do not opt into pre-releases for
> > transitive dependencies reject it outright.
> >
> > The good news is that your team already fixed this in main.
> > The [4] removed envoy-data-plane and the
> > transitive betterproto from the Python SDK entirely, replacing them with
> a
> > small vendored Envoy RLS proto, and CHANGES.md lists it under 2.76.0. It
> > merged on 2026-07-07, one day after the 2.75.0 branch cut, so it just
> > missed that release.
> >
> > So from our checks, a released 2.76.0 is the last thing standing between
> us
> > and un-suspending the provider. Once we can floor the provider at
> > apache-beam>=2.76.0, the problem disappears on every Python version we
> > support and we need no workaround on our side. Until then our options are
> > to wait, or to carry a local pre-release constraint that we would rather
> > not ship.
> >
> > Hence the question: is the 2.76.0 branch cut still on for 5 August, per
> the
> > release calendar [5]? Nothing urgent, and no ask to change anything - we
> > just want to know whether it's reasonable to wait 2.76.0.
> >
> > Our next release wave for providers is also slated for the beginning of
> > August,
> > so even if we make some workaround and force installing a 2020 (!)
> > pre-release
> > of betterproto release, we'd still prefer to wait for 2.76.0 as it
> > would be a very
> > temporary and pretty bad workaround.
> >
> > Thanks for making the fix, and for Beam generally.
> >
> > [1] https://github.com/apache/airflow/pull/61926
> > [2] https://github.com/apache/airflow/pull/66952
> > [3] https://github.com/apache/airflow/issues/66551
> > [4] https://github.com/apache/beam/pull/39213
> > [5]
> >
> https://calendar.google.com/calendar/embed?src=0p73sl034k80oob7seouanigd0%40group.calendar.google.com
> > [6]
> >
> https://cwiki.apache.org/confluence/spaces/AIRFLOW/pages/399278789/Release+Plan
> >
> > Jarek Potiuk
> > Apache Airflow PMC
> >
>

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Elad Kalif <el...@apache.org>.
> Is there a
particular timeline we need to hit in order to make your next provider
release?

Airflow providers are released every 2 weeks
These are the planned dates
https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=399278789#ReleasePlan-ProviderReleases

On Mon, Jul 27, 2026 at 4:10 PM Danny McCormick via dev <
dev@airflow.apache.org> wrote:

> Thanks for following up on this!
>
> > Can we count on ~ beginning of August release of 2.76.0 of python beam
> client?
>
> Yes, we should be releasing in early August as you noted. Is there a
> particular timeline we need to hit in order to make your next provider
> release?
>
> Thanks,
> Danny
>
> On Mon, Jul 27, 2026 at 7:05 AM Jarek Potiuk <ja...@potiuk.com> wrote:
>
> > Hi Beam devs,
> >
> > Writing from the Apache Airflow side with a short status note and one
> > question about 2.76.0 timing.
> >
> > Can we count on ~ beginning of August release of 2.76.0 of python beam
> > client?
> >
> > Background: Airflow suspended its apache-beam provider earlier this year
> > [1]. We have a PR open to restore it [2] (tracking issue [3]), and at
> this
> > point the only thing still blocking us is dependency resolution.
> >
> > Every released Beam up to and including 2.75.0 declares:
> >
> >     envoy-data-plane<2,>=1.0.3; python_version >= "3.11"
> >
> > The only release in that range is envoy-data-plane 1.0.3, which pins
> > betterproto==2.0.0b6. betterproto has never shipped a final 2.x - its
> > latest stable is 1.2.5, from 2020 - so that pin is only ever satisfiable
> > by a pre-release. Resolvers that do not opt into pre-releases for
> > transitive dependencies reject it outright.
> >
> > The good news is that your team already fixed this in main.
> > The [4] removed envoy-data-plane and the
> > transitive betterproto from the Python SDK entirely, replacing them with
> a
> > small vendored Envoy RLS proto, and CHANGES.md lists it under 2.76.0. It
> > merged on 2026-07-07, one day after the 2.75.0 branch cut, so it just
> > missed that release.
> >
> > So from our checks, a released 2.76.0 is the last thing standing between
> us
> > and un-suspending the provider. Once we can floor the provider at
> > apache-beam>=2.76.0, the problem disappears on every Python version we
> > support and we need no workaround on our side. Until then our options are
> > to wait, or to carry a local pre-release constraint that we would rather
> > not ship.
> >
> > Hence the question: is the 2.76.0 branch cut still on for 5 August, per
> the
> > release calendar [5]? Nothing urgent, and no ask to change anything - we
> > just want to know whether it's reasonable to wait 2.76.0.
> >
> > Our next release wave for providers is also slated for the beginning of
> > August,
> > so even if we make some workaround and force installing a 2020 (!)
> > pre-release
> > of betterproto release, we'd still prefer to wait for 2.76.0 as it
> > would be a very
> > temporary and pretty bad workaround.
> >
> > Thanks for making the fix, and for Beam generally.
> >
> > [1] https://github.com/apache/airflow/pull/61926
> > [2] https://github.com/apache/airflow/pull/66952
> > [3] https://github.com/apache/airflow/issues/66551
> > [4] https://github.com/apache/beam/pull/39213
> > [5]
> >
> https://calendar.google.com/calendar/embed?src=0p73sl034k80oob7seouanigd0%40group.calendar.google.com
> > [6]
> >
> https://cwiki.apache.org/confluence/spaces/AIRFLOW/pages/399278789/Release+Plan
> >
> > Jarek Potiuk
> > Apache Airflow PMC
> >
>

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Danny McCormick via dev <de...@beam.apache.org>.
Thanks for following up on this!

> Can we count on ~ beginning of August release of 2.76.0 of python beam
client?

Yes, we should be releasing in early August as you noted. Is there a
particular timeline we need to hit in order to make your next provider
release?

Thanks,
Danny

On Mon, Jul 27, 2026 at 7:05 AM Jarek Potiuk <ja...@potiuk.com> wrote:

> Hi Beam devs,
>
> Writing from the Apache Airflow side with a short status note and one
> question about 2.76.0 timing.
>
> Can we count on ~ beginning of August release of 2.76.0 of python beam
> client?
>
> Background: Airflow suspended its apache-beam provider earlier this year
> [1]. We have a PR open to restore it [2] (tracking issue [3]), and at this
> point the only thing still blocking us is dependency resolution.
>
> Every released Beam up to and including 2.75.0 declares:
>
>     envoy-data-plane<2,>=1.0.3; python_version >= "3.11"
>
> The only release in that range is envoy-data-plane 1.0.3, which pins
> betterproto==2.0.0b6. betterproto has never shipped a final 2.x - its
> latest stable is 1.2.5, from 2020 - so that pin is only ever satisfiable
> by a pre-release. Resolvers that do not opt into pre-releases for
> transitive dependencies reject it outright.
>
> The good news is that your team already fixed this in main.
> The [4] removed envoy-data-plane and the
> transitive betterproto from the Python SDK entirely, replacing them with a
> small vendored Envoy RLS proto, and CHANGES.md lists it under 2.76.0. It
> merged on 2026-07-07, one day after the 2.75.0 branch cut, so it just
> missed that release.
>
> So from our checks, a released 2.76.0 is the last thing standing between us
> and un-suspending the provider. Once we can floor the provider at
> apache-beam>=2.76.0, the problem disappears on every Python version we
> support and we need no workaround on our side. Until then our options are
> to wait, or to carry a local pre-release constraint that we would rather
> not ship.
>
> Hence the question: is the 2.76.0 branch cut still on for 5 August, per the
> release calendar [5]? Nothing urgent, and no ask to change anything - we
> just want to know whether it's reasonable to wait 2.76.0.
>
> Our next release wave for providers is also slated for the beginning of
> August,
> so even if we make some workaround and force installing a 2020 (!)
> pre-release
> of betterproto release, we'd still prefer to wait for 2.76.0 as it
> would be a very
> temporary and pretty bad workaround.
>
> Thanks for making the fix, and for Beam generally.
>
> [1] https://github.com/apache/airflow/pull/61926
> [2] https://github.com/apache/airflow/pull/66952
> [3] https://github.com/apache/airflow/issues/66551
> [4] https://github.com/apache/beam/pull/39213
> [5]
> https://calendar.google.com/calendar/embed?src=0p73sl034k80oob7seouanigd0%40group.calendar.google.com
> [6]
> https://cwiki.apache.org/confluence/spaces/AIRFLOW/pages/399278789/Release+Plan
>
> Jarek Potiuk
> Apache Airflow PMC
>

Re: Beam 2.76.0 timing - last blocker for un-suspending the Beam provider in Airflow

Posted by Danny McCormick via dev <de...@airflow.apache.org>.
Thanks for following up on this!

> Can we count on ~ beginning of August release of 2.76.0 of python beam
client?

Yes, we should be releasing in early August as you noted. Is there a
particular timeline we need to hit in order to make your next provider
release?

Thanks,
Danny

On Mon, Jul 27, 2026 at 7:05 AM Jarek Potiuk <ja...@potiuk.com> wrote:

> Hi Beam devs,
>
> Writing from the Apache Airflow side with a short status note and one
> question about 2.76.0 timing.
>
> Can we count on ~ beginning of August release of 2.76.0 of python beam
> client?
>
> Background: Airflow suspended its apache-beam provider earlier this year
> [1]. We have a PR open to restore it [2] (tracking issue [3]), and at this
> point the only thing still blocking us is dependency resolution.
>
> Every released Beam up to and including 2.75.0 declares:
>
>     envoy-data-plane<2,>=1.0.3; python_version >= "3.11"
>
> The only release in that range is envoy-data-plane 1.0.3, which pins
> betterproto==2.0.0b6. betterproto has never shipped a final 2.x - its
> latest stable is 1.2.5, from 2020 - so that pin is only ever satisfiable
> by a pre-release. Resolvers that do not opt into pre-releases for
> transitive dependencies reject it outright.
>
> The good news is that your team already fixed this in main.
> The [4] removed envoy-data-plane and the
> transitive betterproto from the Python SDK entirely, replacing them with a
> small vendored Envoy RLS proto, and CHANGES.md lists it under 2.76.0. It
> merged on 2026-07-07, one day after the 2.75.0 branch cut, so it just
> missed that release.
>
> So from our checks, a released 2.76.0 is the last thing standing between us
> and un-suspending the provider. Once we can floor the provider at
> apache-beam>=2.76.0, the problem disappears on every Python version we
> support and we need no workaround on our side. Until then our options are
> to wait, or to carry a local pre-release constraint that we would rather
> not ship.
>
> Hence the question: is the 2.76.0 branch cut still on for 5 August, per the
> release calendar [5]? Nothing urgent, and no ask to change anything - we
> just want to know whether it's reasonable to wait 2.76.0.
>
> Our next release wave for providers is also slated for the beginning of
> August,
> so even if we make some workaround and force installing a 2020 (!)
> pre-release
> of betterproto release, we'd still prefer to wait for 2.76.0 as it
> would be a very
> temporary and pretty bad workaround.
>
> Thanks for making the fix, and for Beam generally.
>
> [1] https://github.com/apache/airflow/pull/61926
> [2] https://github.com/apache/airflow/pull/66952
> [3] https://github.com/apache/airflow/issues/66551
> [4] https://github.com/apache/beam/pull/39213
> [5]
> https://calendar.google.com/calendar/embed?src=0p73sl034k80oob7seouanigd0%40group.calendar.google.com
> [6]
> https://cwiki.apache.org/confluence/spaces/AIRFLOW/pages/399278789/Release+Plan
>
> Jarek Potiuk
> Apache Airflow PMC
>