Does Prefect store the data between tasks somewhere? If not, when task retries in a cluster of nodes, how the data is resent to the task? Does Prefect guaranteed the retried task goes to the same node? Sorry for so many questions

Hi @John Song - no worries for the questions! That's what this community is for. For retries with distributed computing, you'll need to configure results handlers for your tasks. I'd recommend taking a look at the results docs for more info on that.

Thanks, I will take a look