Hi all,

I am cancelling the vote on the Airflow Providers wave prepared on 2026-08-01.

The wave was cut with https://github.com/apache/airflow/pull/70791, which caps
pandas below 3 for DataFrame XComs. That has since been reverted in
https://github.com/apache/airflow/pull/71100, and pandas 3 is supported properly
instead in https://github.com/apache/airflow/pull/71103. Releasing providers
carrying the cap would ship a regression we have already decided against, and
enough of the 48 providers in the wave are affected that removing them one at a
time would leave very little worth voting on. A few other providers
also had issues.

No binding votes had been cast, so nothing is lost by re-cutting rather than
trimming.

All providers will be re-cut and put up for vote again shortly, on an expedited
schedule. Those currently at rc1 come back as rc2. The three that were
already at
rc2 in this wave - common-ai, google and openlineage - come back as
rc3. Clickhouse will be removed from the wave, and its rc1 will be yanked.

In the meantime, I also worked with Rahul on preparing constraints for
Airflow rc candidates. Once I get the providers up (in a couple of
hours), the RC prepared for Airflow 3.3.3.3.1 will automatically use
the RC candidates from providers. We basically built-in the solution
that votes for providers and Airflow can be run in parallel. Airflow's
RC will use the Provider's RCs in constraints, which should make the
final Airflow release much closer to what's been tested.

Thanks to everyone who tested. Where a provider's commits are unchanged, the
check marks in the testing status issue carry over to the new wave, so please do
not redo that work.

J.

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to