If you board a German long-distance train often enough, you stop hoping for punctuality and start engineering around its absence. After three years of regular client travel I have come to think of Deutsche Bahn the way I think of any large distributed system I do not control: it has tail latency, hot shards, monitoring you can subscribe to, retry policies, SLA credits, and a small amount of folk knowledge about which paths through the topology are actually faster than the routing layer claims. The official planner gives you the shortest itinerary. Experience teaches you the most likely one. This post is the operator’s handbook I wish I had been handed on day one. It is not a complaint piece. Complaining about DB is a German national sport and the internet is well-stocked. This is the opposite: the things that actually help once you accept that the system will occasionally, with statistical certainty, fall over, and that the right response is not outrage but a better mental model.…
No comments yet. Log in to reply on the Fediverse. Comments will appear here.