Hello guys, We have deployed apache flink kubernetes operator in our production environment using

reactive scheduler

and

standalone mode

. We are using HPA to scale up and down the task managers based on metrics, while flink takes cares of vertical scaling. Our pipeline consists of Kinesis as source and Dynamo as Sink. There are few issues we are seeing right now. 1- We are getting failing checkpoints with unalligned checkpoints, which causes taskmanager restart 2- Task manager restarts on high load not sure what is causing though as we cannot see any exceptions We have a very high load environments 10-12k events/s. Do you think there is any room for improvement? Are we handling it correctly?

@Ahmed Hamdy ^

@Ahmed Hamdy Your suggestions on this?