Hey folks,

This is a call to action for the multi-team project. We have had many 
interested parties reach out, but previously we lacked well-defined and 
well-scoped areas for contributors to tackle. But we now have some items that 
could use some help, specifically in the executors space.

Context:
Each executor needs changes before it can be used in a multi-team setup. The 
two major requirements are:
1) Ability to read team-based configuration - e.g., two CeleryExecutors may 
need different brokers, backends, or autoscaling settings, etc for different 
teams.
2) The executors must be safe to run concurrently within the same scheduler 
process  - e.g. no shared memory/queues, global DB tables, or shared 
filesystem, etc. This implies different things for different executors; some 
may need little to no changes some may need many.

The LocalExecutor [1], CeleryExecutor [2] and AwsEcsExecutor [3] have already 
been updated (or are in progress) to support multi-team. All other executors 
are outstanding.

We welcome folks to get involved with updating executors to support multi-team. 
I have created a tracking ticket here [4] which has sub-tasks per executor. 
Please add any that I have missed or leave a comment on the issue. Plenty of 
help and guidance will be provided!

Also, in particular, if anyone is familiar with k8s/KubernetesExecutor, I would 
love for that one to be available for 3.2 but I don't have the time/bandwidth 
to complete it personally.

Thanks for your time, and I look forward to working with you!

Cheers,
Niko

[1] https://github.com/apache/airflow/pull/59021
[2] https://github.com/apache/airflow/pull/60675
[3] https://github.com/apache/airflow/pull/55003
[4] https://github.com/apache/airflow/issues/60912

Reply via email to