-
Notifications
You must be signed in to change notification settings - Fork 4.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Task]: Update the minor version of cloudpickle library prior to Beam release. #23119
Comments
Looks like we are on the latest |
updated for 2.45: #25143 |
Moving to 2.47 since there is not a new release |
No updates to cloudpickle version. Moving to 2.51.0 |
It looks like there's been a major version upgrade (and no minor version upgrade) - https://pypi.org/project/cloudpickle/#history @tvalentyn I don't think this should be a release blocker and probably represents non-trivial work to investigate/upgrade. I'm going to move the release blocker and we should think about a good way to fund this going forward |
Similar status as before for cloudpickle, major increment to 3.0.0 was before the 2.52.0 branch cut. @tvalentyn any objections to rolling this up to 2.54? |
no objections. |
There's one week until the 2.54.0 cut and this issue is tagged for that release, if possible/necessary please complete the necessary work before then, or move this to the 2.55.0 Release Milestone. |
2.2.1 is still the latest of 2.x as of 2.55.0 cut |
No update here |
We should move to the next milestone each time (not just remove it) |
per #32617 cloudpickle 2.2.1 looks fine for Python 3.12, however we should revisit cloudpickle 3.0 later, move to the next milestone. |
#30528 (comment). Note the latest dask moves to cloudpickle >= 3.0.0 (https://github.com/dask/dask/blob/main/pyproject.toml#L33) |
What needs to happen?
If a Beam dependency has a flexible upper bound, users will download the most recent compatible version of a dependency at sdk installation time. Overtime, the version used at job submission may become newer than the version installed in a released Beam container. Given that forwards-compatiblity of a pickle library is not guaranteed, the pipeline may fail to unpickle at runtime.
To mitigate, pickle libraries should be specified in
install_requires
with tight version bounds limiting to a particular minor version. This can cause inconvenience to Beam user, if we depend on an old version of a library. Therefore, we should periodically update the version we use, at least once per release cycle.We can consider to close this issue when either condition is met:
Until then, don't close this issue, instead, move it to the next release milestone after updating the version in https://github.com/apache/beam/blob/master/sdks/python/setup.py
Issue Priority
Priority: 3
Issue Component
Component: sdk-py-core
The text was updated successfully, but these errors were encountered: