Autobanning Bad Requests (Part 2)

2026-10-11 Autobanning Bad Requests (Part 2) I've been reviewing the auto-banning solution that I described last week and discovered that it isn't performing as well as I had hoped.1 The issue was something that I predicted: Eventually, I might speed up the detect and block loop. There are a few minute of latency between bad requests being made, and the client being blacklisted. That's enough time for the bot complete scraping the site which isn't ideal, but whatever. I didn't realise at the time how fast that loop would need to be. It's now apparent that the bots start and finish their requests within seconds. I need a zero-latency detect and block cycle which is impossible to achieve using my existing solution. I dug into the HAProxy documentation and discovered that there is a feature called a stick-table.2 It's basically an array of state that can be called upon when deciding whether to accept or deny a request. I'm using it to track the IP addresses making requests that exhibit…