Nobody pushes a config change at 17:00 on a Friday twice
Nobody pushes a config change at 17:00 on a Friday twice.
The first time teaches you everything you need to know. The link comes up, the BGP session comes up, the graphs look clean, and then somewhere around Saturday morning a customer opens a ticket about something that has nothing to do with what you touched.
Most NOCs end up with an unwritten Friday rule. Some call it a change freeze. Some just call it common sense.
The reasoning is rarely about the change itself. It is about who is around when it goes wrong, and how long the gap is between the moment it breaks and the moment somebody who actually understands it is awake and looking at it.
That gap is the risk. The config is just the trigger.
The teams that run clean Friday changes are usually the ones with a tested rollback, a window that closes before the on-call handover, and someone watching who was in the room when the change was designed. When those three things are true, Friday is an ordinary day. When they are not, Monday finds out.
Our engineers are split on this. Half say freeze everything after Thursday afternoon. Half say a change you are afraid to make on Friday is a change you should not be making on Tuesday either.
Both camps have a point.
What is the rule on your team? Hard freeze, soft freeze, or it depends who is on call.
