The Problem
Our lead generation automation broke under load. When a viral post triggered 2,000 webhooks in an hour, n8n crashed due to memory exhaustion.
The Solution
We decoupled the webhook ingestion from the heavy LLM processing. We set up an intermediary Redis queue to catch the webhooks, and then used a cron-triggered n8n workflow to process them in batches of 50.
The Results
- 100% webhook ingestion success rate
- CPU usage stabilized below 40%
- Zero dropped leads during traffic spikes