Upstream Alert: We all fall for the same trap

A routine OS update took down Heroku for 23 hours on June 10th, 2025. The incident echoes similar outages at OpenAI and Datadog, reminding us that lessons fade.

Summary: https://www.heroku.com/blog/summary-of-june-10-outage/

What happened?

On June 10th, 2025, Heroku went offline. Web apps died quietly, CLI commands failed, and logging vanished. And the status page? Broken too.

The outage lasted 23 hours and impacted users around the globe. The cause? A routine OS update, the kind you expect to be boring.

One of Heroku’s servers received an automated patch that unexpectedly restarted its networking service. Well, that wasn’t supposed to happen in the first place. But even more challenging: Heroku relies on a legacy script that set-up proper routing during boot only. So when the networking stack restarted mid-run, the routes didn’t get reconfigured.

Result? Servers could no longer reach the outside world.