Triggerer process does not reload trigger implementations after git-sync when deployed on kubes #33126
Unanswered
JossWhittle
asked this question in
Q&A
Replies: 1 comment 2 replies
-
This is as expected. Triggerer does not reload Triggers dynamically, it needs to be restarted. PRs to change the behaviour are welcome. |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Consider the scenario:
A DAG using a custom Sensor with accompanying custom Trigger is already running, and the Airflow
triggerer
container has spawned, git-cloned, and started running the custom Trigger.The currently running Trigger continues to execute as expected. It responds correctly to it's event and the deferred task is resumed correctly.
But when the Sensor defers and the
triggerer
container picks up the request to run the custom Trigger it reuses the original implementation.Getting the custom Trigger to refresh to the latest code requires killing the
triggerer
container and letting kubes recreate it.My current theories are:
triggerer
containerscheduler
container using the newly git-synced codetriggerer
container using the original first git-cloned code.This feels like a bug. I'd love to get other peoples inputs here.
Can it be made safe to implement deferable Sensors and Triggers in git-synced code?
Beta Was this translation helpful? Give feedback.
All reactions