Better Limits

Thoughts and notes · Work

Deployment – Pencils Down

My entire approach to deployment can be summarize as follows:

– Putting tested process/solution into production without any translation.

This is pencils down time — meaning we don’t change the solution. If anything needs to be changed it should be sent back to where it came from (e.g. change/development) and pushed out of deployment as ‘not ready’. getting into changing things only risk your outcomes and your reliability. Expect yourself/others to properly prepare for deployments, and don’t allow yourself, by feelings or request, to stray from the tested solution being deployed.

Step 0: You are responsible. The presumption upon deployment is that everything within reason has been tested for and will work as expected in the production environment. if you’re not capable of doing this (or at least assuming that you can and knowing when you need to stop) — then you shouldn’t be deploying.

The one thing that is true of all deployments is that blame will be put on someone when it goes wrong. Naturally and infallibly, the first person to look to will be (you) the person who deployed it. Don’t start deployments that can’t be completed.

Step 0.5: Is your bag packed. This is really more of just step one but before you actually initiate any part of a deployment you should do a basic check to make sure that anything you don’t control, that will use what you do control, is ready to go. If it isn’t and can wait until after deployment, you should make sure that notification are sent when you are done deploying to anyone in charge of those pieces/system.

Actual Deployment:

Step 1: Files First

– All resources are in place and correct. As with linux, most things with a computer are a file or generated from one, so while you might have to turn on a process or two and do validation, make sure any files that are needed are deployed first or automatically created. This may not explain what is being done but it doesn’t matter the resources (files) are required and no amount of understanding will change that. If they are missing, wrong or can’t be put where they should be, everything stops, and you can’t deploy. Do this first.

Step 2: Workflows, Dependencies, Containers

– This is the point where things can get a little confusing but where general knowledge of too down will come into play. Understand the overview of what you are doing and get things that connect your files and their outputs in place. Every resource/file should connect to something and be used in some way. If it isn’t you missed something and need to go back to check it again.

– Additionally many of things at this part of deployment need to exist in order to setup the rest. So do these, and then connect them. All your files/resources are already in place.

Step 3: Validation

this could be something as simple as downloading the file that you just uploaded, but whatever your checks are, deployment is not complete until you have everything in place and have looked at it a 2nd time (even if it’s an automated check) to make sure it exists as you think it does. (e.g. Download an upload a file (or run a hash on it once it’s uploaded. Either should allow you to confirm that the file didn’t get corrupted and is what you think it is.)

Step 4: External Dependencies and Connecting Solution.

– Just because something is deployed, doesn’t mean it is useable. You may not even be able to give someone the connection or access to your deployment (deployed solution) but what will likely matter is that you make them aware of it. Courtesy if nothing else.

Step 5: Communicate — do your best to notify anyone who might need to know when you are done. Don’t assume they will just know.