Nowhere Without a Plan

Nowhere Without a Plan

In my practice, from time to time I face the need to carry out high-stakes operations that affect production. This could be a migration to another platform, server, cluster, and so on.

And it's not always possible to do a smooth switchover to the new setup, which means there's no room for error, and the time window for the procedure is usually very limited 🤷‍♂️🤷‍♂️🤷‍♂️

➡️ In this case, I proceed as follows: on some test environment I run through all the steps of the procedure and write them down as a plan, precise down to the command or a clear and concise instruction.

The plan itself should be understandable even a month later, so it's important to put it together properly. Since the time frame for the procedure is tight, there's no need to write out two volumes of instructions.

➡️ Next, once you have a plan, you can go through it completely once more, executing it step by step, and measure how much time it takes. After that, all that's left is to schedule the operation, notify the relevant people, and at the appointed moment get the job done. I usually schedule such things for early morning, so I can finish everything while everyone is asleep. Well, I love the smell of burnt servers in the morning, what can I say)