It looks like building from sources on our CI using base debian image (with enabled optimizations) "just works" and takes 1m36s. -> and taking into account that we already have pretty sophisticated CI-level remote caching that will allow to rebuild it only when either base image is released or python is released - building sounds like a very good idea.
J. On Thu, Jun 26, 2025 at 7:02 PM Damian Shaw <ds...@striketechnologies.com> wrote: > Going from source is certainly going to give the most control, I > personally choose python-build-standalone for my own images rather than > building from source to outsource all the build choices, such as what > optimization flags to enable, what compiler tool chains to use, etc. > > Damian > > > > -----Original Message----- > From: Jarek Potiuk <ja...@potiuk.com> > Sent: Thursday, June 26, 2025 10:15 AM > To: dev@airflow.apache.org > Subject: Re: [DISCUSS] Switching base Python container images to > python-standalone ? > > Thanks Damian - very insightful. And yes - I did not really want to > "diminish" the value of community images and work of the maintainers, it > was really more on the "we base our image security on the assumptions that > it's coming from "official" sources and surely there are some guardrails" - > which turned out to be not really true. > > But.... We might not have to even use python standalone. > > Aritra just tested installation of Python straight from sources -> > https://github.com/apache/airflow/pull/52265 and I was actually pretty > surprised how non-problematic and fast it was - and it seems it passes all > our CI tests. We just need to add a little pre-commit magic to get > notified when we should update patchlevel version of Python when a new one > is released, and we should be able to try it out in CI image - once it gets > battle-tested with CI/breeze etc. we can transfer this to PROD image as > well. We already have an idea how to do it - our PROD images are optimized > for size and do not contain "build essentials" - but I think we should be > able to build Python in the "build" segment and simply copy resulting > binaries to the "main" segment - since in both cases we use the same base > image, such 1-1 copy should **just work** - we already do the same with > installed python packages - we install them (including building when > needed) in build segment and we copy-over the installed .venv to the main > segment. > > So ... we might even not need a discussion - installing from Python > sources is THE BEST > > J. > > > On Wed, Jun 25, 2025 at 6:25 PM Damian Shaw <ds...@striketechnologies.com> > wrote: > > > First, I would like to thank the community members who have been > > maintaining the Python docker images, it's one of those thankless > > opensource infrastructure volunteer roles that they've been doing for > > a long time. Unfortunately Docker assigns the title "Official Image" > > for various community run images, which creates a misconception on the > > guarantees being provided, and if I were a suspicious person I would > > say Docker creates this misconception on purpose to both get free work > > from community members and make Docker seem more supported by third > > party organizations than it actually is. > > > > On the topic of python-build-standalone, I've been using it in > > production for several months now and I'm fairly happy with it. > > > > However, one minor reproducibility issue I have when installing > > python-build-standalone via uv, is that uv does not have an ability to > > pin to a specific build between uv versions, as uv hard codes a > > mapping of Python version and platform to a specific build and then > > updates that mapping between releases. So, updating the version of uv > > between runs, or having two users run different versions of uv to > > initialize the environment can change the results. > > > > While normally such build changes are trivial, if you look at the uv > > 0.7.8 and 0.7.9 changelogs you will see that sometimes they can have > > significant > > impact: https://github.com/astral-sh/uv/releases/tag/0.7.8. Also, this > > email finally prompted make an issue on this topic: > > https://github.com/astral-sh/uv/issues/14263. > > > > There are other ways to source python-build-standalone, such as > > pbs-installer or writing your own script, but I've not yet spent any > > time investigating them, so I can't comment on them. > > > > Damian > > > > -----Original Message----- > > From: Jarek Potiuk <ja...@potiuk.com> > > Sent: Wednesday, June 25, 2025 10:44 AM > > To: dev@airflow.apache.org > > Subject: [DISCUSS] Switching base Python container images to > > python-standalone ? > > > > Hello here, > > > > Together with Aritra we are looking into adding a few more things to > > our images (golang tool chain for CI image), also Shahar is > > experimenting with Rust tool chain and I also recently realized (by > > some of the issues we had) that 'Docker Official Python Image' that is > > part of 'Official Program' [1] is not as 'Official' as I thought so we > > discuss about changing the base of our images (first CI and then when > > we see it works fine - PROD) > > > > Currently we are using the 'Official' image - but after some issues > > and discussions with people at PyCon and FOSS Backstage (I had a > > chance to talk to Python maintainers and even had a few beers with > > them) - it turned out that the official Python Image' is maintained by > > 'a community's which really is a few pretty random people - and that > > explains for example why we have sometimes unpacked security > > vulnerabilities in setuptools etc. - because they made some > > compatibility choices and decisions that do not allow them to upgrade > > easily, also they had some delays in releasing updated Python > > versions. And Docker does not **really** do much vetting there. > > > > So I think it would be good to switch how we build the base for our > images. > > And following the experience of `uv python` [2] - it seems that maybe > > using "python-standalone" [3] project is a good alternative. It's > > managed by Astral now (so yes - another dependency on them), but what > > you have with it you have practically 100% complete Python interpreter > > installed in seconds. > > We could continue using debian-slim as a "base, base image" - and > > install python using "python-standalone". There are a few > > incompatibilities [4] of the distributions of Python, but there are > > very few and mostly related to some obscure systems (compatibilities > > with terminal in REPL, and gtk / UI integration that is anyhow not > > really working in "standard" Python distributions). > > > > I would love to hear what you think - happy to get any feedback/ > > insights, suggestions and answer additional questions, provide some > > links to past "troubles" we had with Python "Official" images etc. > > > > J. > > > > > > [1] Official Python Images - https://hub.docker.com/_/python [2] UV > > Python installation - https://docs.astral.sh/uv/guides/install-python/ > > [3] Python Standalone project - > > https://github.com/astral-sh/python-build-standalone > > [4] Python Standalone incompatibilities - > > https://gregoryszorc.com/docs/python-build-standalone/main/quirks.html > > ________________________________ > > Strike Technologies, LLC (“Strike”) is part of the GTS family of > > companies. Strike is a technology solutions provider, and is not a > > broker or dealer and does not transact any securities related business > > directly whatsoever. This communication is the property of Strike and > > its affiliates, and does not constitute an offer to sell or the > > solicitation of an offer to buy any security in any jurisdiction. It > > is intended only for the person to whom it is addressed and may > > contain information that is privileged, confidential, or otherwise > protected from disclosure. > > Distribution or copying of this communication, or the information > > contained herein, by anyone other than the intended recipient is > > prohibited. If you have received this communication in error, please > > immediately notify Strike at i...@striketechnologies.com, and delete > and destroy any copies hereof. > > ________________________________ > > > > CONFIDENTIALITY / PRIVILEGE NOTICE: This transmission and any > > attachments are intended solely for the addressee. This transmission > > is covered by the Electronic Communications Privacy Act, 18 U.S.C > > ''2510-2521. The information contained in this transmission is > > confidential in nature and protected from further use or disclosure > > under U.S. Pub. L. 106-102, 113 U.S. Stat. 1338 (1999), and may be > > subject to attorney-client or other legal privilege. Your use or > > disclosure of this information for any purpose other than that > > intended by its transmittal is strictly prohibited, and may subject > > you to fines and/or penalties under federal and state law. If you are > > not the intended recipient of this transmission, please DESTROY ALL > > COPIES RECEIVED and confirm destruction to the sender via return > transmittal. > > > ________________________________ > Strike Technologies, LLC (“Strike”) is part of the GTS family of > companies. Strike is a technology solutions provider, and is not a broker > or dealer and does not transact any securities related business directly > whatsoever. This communication is the property of Strike and its > affiliates, and does not constitute an offer to sell or the solicitation of > an offer to buy any security in any jurisdiction. It is intended only for > the person to whom it is addressed and may contain information that is > privileged, confidential, or otherwise protected from disclosure. > Distribution or copying of this communication, or the information contained > herein, by anyone other than the intended recipient is prohibited. If you > have received this communication in error, please immediately notify Strike > at i...@striketechnologies.com, and delete and destroy any copies hereof. > ________________________________ > > CONFIDENTIALITY / PRIVILEGE NOTICE: This transmission and any attachments > are intended solely for the addressee. This transmission is covered by the > Electronic Communications Privacy Act, 18 U.S.C ''2510-2521. The > information contained in this transmission is confidential in nature and > protected from further use or disclosure under U.S. Pub. L. 106-102, 113 > U.S. Stat. 1338 (1999), and may be subject to attorney-client or other > legal privilege. Your use or disclosure of this information for any purpose > other than that intended by its transmittal is strictly prohibited, and may > subject you to fines and/or penalties under federal and state law. If you > are not the intended recipient of this transmission, please DESTROY ALL > COPIES RECEIVED and confirm destruction to the sender via return > transmittal. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@airflow.apache.org > For additional commands, e-mail: dev-h...@airflow.apache.org >
