Sxnan opened a new issue, #185: URL: https://github.com/apache/flink-agents/issues/185
### Search before asking - [x] I searched in the [issues](https://github.com/apache/flink-agents/issues) and found nothing similar. ### Description Currently, with Python async execution supported, a Python action can be divided into multiple invocations before it is finished. And each invocation can potentially update the state. It is problematic when a checkpoint is taken while an async execution is not finished, but has already updated the state. Say we have an async action divided into two steps, 1 and 2. After step 1, it updates the state from S0 to S1. Now, a checkpoint is made. Now the checkpoint contains S1. When we recover from the checkpoint, we have to run the async action from step 1, but now it sees the state S1 instead of S0. To fix the problem, we have to keep the update to the state in memory, and only persist it to the state when the async action is finished. Note that the per-action state consistency has the same problem, so we can only put the state update to the action state store when the async action is all finished. ### Are you willing to submit a PR? - [ ] I'm willing to submit a PR! -- 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: issues-unsubscr...@flink.apache.org.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
