The setup/configuration of agents, queues, etc. in Prefect 2 seems quite flexible. I imagine there’s optimal configurations for certain use cases. I wonder what that optimal configuration is for our use case - perhaps others have a similar use case.
We have a relatively small number of flows (~30; at least I imagine that’s relatively small), but many of the flows run for many hours (e.g., 7 hrs).
Clearly we don’t want the long running flows to fail due to some sort of exception related to the flow running for such a long period.
Is there some combination of number of agents, queues, etc. that will best support this use case?
The setup we’re currently using is 3 agents running on a single large VM, with each agent listening to 4 queues each. I can’t speak to how this performs under load because we haven’t completely moved from Prefect 1 yet.