YingboWang commented on issue #5499: [WIP][AIRFLOW-3964] Build smart sensor
URL: https://github.com/apache/airflow/pull/5499#issuecomment-538674356
 
 
   @seelmann  We are deploying the smart sensor inside our company right now. 
Testing and monitoring it for further refinement. I will update the code for 
this PR and AIP document with the overall structure and some statistics based 
on our testing for the past several weeks. 
   
   @BasPH @feng-tao, Thanks for your feedbacks. It's my bad not update this PR 
on time. 
   The smart sensor operator itself is only used to create a control dag for 
taking over all sensors in the airflow cluster when the cluster is enable to 
"use_smart_sensor" by configuration. The smart sensor task is running as a 
service keep refreshing all active sensor tasks from DB and poking them. Users 
doesn't need to change their original sensor tasks when writing DAGs. 
   
   When use smart sensor, the sensor tasks state changing is different with 
regular sensor. The scheduler will persist sensor's poke context to DB and set 
the state to "Smart_pending" so that the task will not be queued and taking a 
separate process in any worker. Smart sensor task instance which is a long 
running process on the work will refresh DB to get all sensors that are in 
"Smart" state (smart_running or smart_pending) and poke them. If the poke 
return True it will set the corresponding task_instance state to "success" so 
that the scheduler will be able to kick off the downstream task_instance in the 
next parsing loop. There is only a cluster configuration change for using smart 
sensor. No need to write DAGs in a different way. I will update the PR and the 
description with latest change and also update the AIP to include all details 
for smart sensor design. 

----------------------------------------------------------------
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.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

Reply via email to