1.2 KiB
1.2 KiB
The Flow itself
- No secrets, passwords or similar in place.
- No unnecessary hard-codings (if some piece of data acts as constant in several flows, it's better to make it part of the
env) - The right
reference_tasksare indicated, or the defaults are good enough - The flow has a proper clean-up/tear-down (if something fails, finishing activities like closing connections or deleting files will still happen)
- There is no unnecessary boilerplate imports, configs, variables, constants and functions in the script (if you can remove something and the flow will still always work properly, remove it)
- The flow has a way to be run and tested without consequences (i.e. write to a production database)
- You have run the flow succesfully at least once in its final state before registering
- You have run the flow succesfully at least once from the prefect server
Scheduling
- You have set up a slack warning so that failures are notified
- You have intentionally made the flow fail to verify that notifications work properly
- The right parameters are set with the scheduled run
- You have verified manually that the first scheduled run went well and produced the expected output