potiuk commented on issue #33744: URL: https://github.com/apache/airflow/issues/33744#issuecomment-1790216854
> Thanks a lot for you answer. I did check with a small py script calling symbol_by_name from kombu with same argument and it worked. But the whole stack (Airflow 2.7.2) went OK only when I downgraded to redis 4.52. Also import depends on python level. That's why I mean - something is messed up in your env (but I have no idea why and only you can investigate it). Our CI and development environment already does "whole stack". Whern you install airflow in a given version you are supposed to use constraints - see the docs: https://airflow.apache.org/docs/apache-airflow/stable/installation/installing-from-pypi.html to achieve reproducible installation with "known good" set of dependencies. All those are tested, verified and they are known to work. And only that set is "guaranteed" to wrork. You can downgrade or upgrade dependencies within the requirements limit that airflow has (and in most cases you will be fine) - but then it might be that some dependenciex will cause some problems (it's impossible to test matrix of combinations of 670 dependencies). In Airflow 2.7.2 constriants for example for Python, we know that Airflow works with those dependencies described in our constraints: https://github.com/apache/airflow/blob/constraints-2.7.2/constraints-3.8.txt For example there: * celery==5.3.4 * redis==4.6.0 * kombu==5.3.2 I just run airfow 2.7.2. with those set of constraints and there is no problem - everything imports fine in Celery, everything works. You can check it yourself (and our CI does it automatically) by running development tool called `breeze` in main version of airflow source code: ``` breeze start-airflow --use-airflow-version 2.7.2 --executor CeleryExecutor ``` There - you have running airflow with celery and redis and everything works fine, no import problem. That's why I think somehting is broken in your environment. What It is, hard to say, but I recommend you to start with the constraint list and compare your installed dependency versions with those on the constraint lits and work it out from there. Maybe some other dependency you have installed is breaking things. What we can do is provide you with "known good" constraints, and if you diverge from it, you have all the means to investigate what's causing it if you see problem like this (or you can stick to the constraints of our that are tested, verified and provide reproducible installation. Unless you have a reason to diverge from them - I recommend you to follow the installation process we have (as this is the only one that providese reproducible, working installation that we had a chance to test. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
