n8n queueing every afternoon since I added AI steps. More cores, faster cores, or queue mode?
My current setup: 4 vCPU, 8GB, Docker Compose with n8n, Postgres and Redis behind Traefik. I have about 40 workflows and recently I have added LLM calls and a vector lookup to a few of them. What I'm seeing at peak: load average 6.8, 6.1, 5.4. The n8n container sits at 340% CPU in docker stats.