Hi does anyone know in Prefect 2 why there would be a 30 second delay between the agent picking up a flow run and starting to execute it? Many thanks.

I have my flows stored in GitHub storage and I always assume this is the time it takes to clone everything and setup the infrastructure... But this is only an assumption and I would be happy to learn more about this!

The 5s probably come from PREFECT_AGENT_QUERY_INTERVAL: https://docs.prefect.io/api-ref/prefect/settings/#prefect.settings.PREFECT_AGENT_QUERY_INTERVAL

Well I am seeing that it checks for runs every few seconds but the delay is after it picks up a run but before it starts executing. Seems it takes a long time to create a subprocess on the box the agent is on

In general I also noticed there is a delay in running flows locally since we migrated to prefect 2

Hi, I put in debug statements to the agent and it's the get_directory call on the storage which takes the time. Seems the subprocess spawned by the agent copies everything into a directory under the user's temp

permissions might not be the same for the directories locally, since where your code and execution is stored is presumably owned not by the prefect account

Actually I found a way round the copy. I created a Process block with the working directory the same as the one I use for the deployment. No copy is therefore required and it works really nicely.