This is partially to be expected although the delay for the workflow to actually start is normally around 1 - 3 seconds depending on hardware. We briefly mention this in the docs here: Execution modes and processes - n8n Documentation
Hello! Yeah, I saw that, but did not expect to be over 6s - that feels really bad. I’m running this in 8 vCore + 16 GB RAM VM on Proxmox, that has 2x E5-2450L. If I set the webhook node to respond immediately, I get ~1s, just like with EXECUTIONS_PROCESS=main.
I’m wondering if there is any optimization I could do to improve this…
You might be able to make some changes to docker itself, In my test environment I get ~ 3 seconds without changing anything. You would be better off using main and if you were worried about the instance falling over queue mode would be the best approach to take.
There could be some information out there on what docker settings could be tweaked to make the most of your hardware but we don’t really have anything in n8n itself.