I really like it, and for me it's no brainer - we've been casually discussing that we need to do that one sooner rather than later and I like the attention to detail and the analysis done by Niko.
My only ask will be to split it into smaller, independent PRs when it gets implemented, so that we can asses better the consequences of some changes (and whether we have to potentially account for backwards compatibility for users who implemented their own executors (not a very popular one also because of those problems this AIP is aiming to solve, but still we have to account for that). I personally think if there are no objections, this one is ready to start voting on. J. On Tue, Nov 8, 2022 at 1:42 AM Oliveira, Niko <[email protected]> wrote: > > Hey folks! > > I went ahead and wrote an AIP for this proposal. It can be found here: > https://cwiki.apache.org/confluence/display/AIRFLOW/AIP-51+Removing+Executor+Coupling+from+Core+Airlfow > > > Please leave any feedback here or in the Confluence comments. > > Thanks for your time! > > ________________________________ > From: Oliveira, Niko <[email protected]> > Sent: Monday, October 24, 2022 2:28 PM > To: [email protected] > Subject: [EXTERNAL] Proposal to Remove Executor Coupling in Core Airlfow Code > Base > > > Hey all! > > Recently I have spent some time investigating the occurrences of hardcoded > Executor logic within core Airflow code and put together a mini-AIP of sorts > on Github Discussions (it was nice to use GH markdown and automatic code > snippets). > > I'm particularly interested to hear if folks think an AIP would be reasonable > for this set of changes or if the community is fine with using Discussions > alone and beginning development without an AIP. > > https://github.com/apache/airflow/discussions/27241 > > Thanks for you time! >
