I don't normally get into dance, but I've watched this five times since I came across it yesterday and it's stunning every time. Beyond the incredible choreography - the clothing, colors, lighting, camera work, editing, everything. (TIL: CDK is a dance company in the Netherlands.) https://www.youtube.com/watch?v=REPPgPcw4hk #dance#cdk#gotye
I'm really glad I started day one by using my own domain and handle for #mastodon I also have to give a shout out to @mastohost for making it look like I admin an perfect instance.
Shout out to #aws and of course #cdk for helping me make Lebowski.Social not only a reality, but easy to admin. Am I paying a bit more premium than pure metal? Sure. Is the scaling, backups, and migrations easier? So far.
You just create a user, then create a group, then set up MFA, then add the user to the group, then create permissions policies for all of the hundreds of services the user needs to use, then create a Role, then assign the policies to the Role, then the user can use STS to assume the Role, then they have an access token they can use. Make sure that you tag your services and use attribute based access control to make sure your policies have limited access.
Right now, someone is sitting down and learning how to deliver infrastructure as code for the very first time. Maybe it’s #terraform, or #cdk, or #ansible. They’re hearing the word #idempotent for possibly the first time. They’re going to Google things and check StackOverflow. How we all comport ourselves in answering questions as the folks who ostensibly know what we’re doing has a tremendous influence on their ability to stay in the game. We were all newbies once. Be kind, don’t be a jerk.
I'm going to be speaking at Indy.Code() in #Indianapolis in August! Come check out my #AWS#CDK talk! From now until July 1st you can register with the discount code EarlyBird for $50 off!
TIL that when you manage #AWS Step Functions using #CDK or #CloudFormation, you have no control over JobDefinition revisions. So whenever you change a job definition, CloudFormation creates a new revision for it and inactivates the previously active one.
So in practice, when you update a StepFunctions job definition any existing long-running executions that depended on the old version will simply fail. And waste a ton of resources and potentially money.
This sort of problem is exactly what turns some engineering leaders off of #serverless for other architectures.
In particular, I wish AWS would spend more energy into making CloudFormation support better across the board. Like requiring support for new services and ensuring a minimum resource coverage as part of service teams' goals. Because it doesn't look like this is the case today. Something many people like @Quinnypig have been screaming (in the cloud) about for ages now.
Sorry for the rant. I like AWS and serverless, but this boiled my blood today because it had non-trivial impact into my operations and cost and took the team quite a while to figure out. Now we'll have to figure out an ugly workaround to implement. 🙄