Good Morning!
I’ve moved completely off of Twitter on over to BlueSky and things over there are super rosy. I liveskeeted busting out some crappy Python and built an AWS NewsBot since AWS hasn’t gotten the memo that there’s a thriving AWS community over there yet. I’ve pondered why the AWS Developer YouTube suddenly spiked to millions of followers, yet the AI videos with "millions of views" seemingly get fewer comments/hits by humans than most baby seals. And I’ve posted some art I had commissioned.
But you don’t care about that–you care about the bevy of pre:Invent releases that AWS has hurled our way, and last week was a doozy. Let’s do this…
From the Community
Adrian Cockroft wonders whether AWS will have anything new to say about sustainability at re:Invent. I’m betting nothing of substance, because vanishingly few customers care about it beyond lip service.
As an example of the creeping platform enshittification, I learned that there’s now no way to block Amazon ads from Echo Show devices. Good luck on that "feature" pairing with the new Echo Show devices you dropped last week, Amazon…
Podcasts
Last Week In AWS: The Return of Old AWS
Screaming in the Cloud: Replay – Keep on Rockin’ in the Server-Free World with Michael Garski
Screaming in the Cloud: Standardizing Developer Freedom with Chris Weichel
Choice Cuts
Enhanced account linking experience across AWS Marketplace and AWS Partner Central – Two of the nineteen distinct kinds of Amazon accounts can now speak to one another.
Amazon API Gateway now supports Custom Domain Name for private REST APIs – FINALLY. We can now have internal references for things that aren’t named like somebody threw a cat onto a keyboard, then tried to give it a bath.
Amazon Aurora Serverless v2 supports scaling to zero capacity – After fighting it for years and failing at their attempt to retcon their own history, Amazon finally wakes up to the realization that Serverless does in fact mean "scales to zero."
Amazon CloudFront now supports Anycast Static IPs – For the low low price of $3K per month per list, you can still pretend it’s 1995 and avoid rewriting your application.
Amazon CloudFront now supports additional log formats and destinations for access logs – I’m amazed at the sheer number of times I’ve seen things kluged together to achieve this very thing. While the folks who implemented theirs a few days ago likely won’t be thrilled at the timing, this is a great release that’s long overdue.
Amazon CloudFront announces VPC origins – Whoa. Now you can completely remove your haunting fear of screwing up the security groups, and provide CloudFront access to instances / load balancers / other nonsense that lives entirely within a private subnet. This is huge.
Amazon CloudWatch launches full visibility into application transactions – Call me skeptical, since AWS can’t even give you full visibility into what resources are running in your AWS account–but the gem is hidden at the bottom of this item: "A new pricing option is also available , encompassing Application Signals, X-Ray traces, and complete visibility into transaction spans." What’s the new pricing? I don’t know! The CloudWatch Pricing page has (no exaggeration) 14 distinct tabs of pricing data, because if there’s one thing the CloudWatch team hates, it’s customer comprehension. This is gonna require a fair bit of digging I fear…
Amazon EC2 now provides lineage information for your AMIs – Ah yes, the Amazon Linux 2023.2 lineage – a most distinguished specimen indeed. Descended from the noble Red Hat Enterprise Linux bloodline, this pedigree demonstrates remarkable stability characteristics passed down through generations of careful upstream curation. One observes the refined Fedora DNA expressing itself in the regular security updates, while the AWS-specific traits – carefully bred in over years – manifest in enhanced cloud performance and integration with other AWS services. Most elegant. Would you care to examine its papers? The pedigree certification is, naturally, not available through AWS Systems Manager.
Amazon Q Developer in the AWS Management Console now uses the service you’re viewing as context for your chat – What the blue hell was it doing before?! It sure did pop up intrusively to interrupt you an awful lot for a thing that apparently had no bloody clue what it was you were doing.
Amazon WorkSpaces introduces support for Rocky Linux – Technically, all Linux is rocky the first time.
AWS App Studio is now generally available – This is an AI powered thing: you badly describe the thing you want, and App Studio creates a half-assed implementation like an AWS service on its release day.
AWS CloudTrail Lake launches enhanced analytics and cross-account data access – The cross-account feature explicitly mentions that you can query event stores between accounts only in the same region. I can’t stand it. Cross account, and cross region. Customers want both, AWS. I grow tired of saying it.
AWS Compute Optimizer now supports rightsizing recommendations for Amazon Aurora – This one I’m facing with a bit more skepticism: I’ve run it on an Aurora cluster and for some unknown reason, it doesn’t suggest turning the thing off and replacing it with an EC2 instance instead. I can’t understand why!
AWS Elastic Beanstalk adds support for Node.js 22 – It’s not dead after all!
AWS Lambda supports Amazon S3 as a failed-event destination for asynchronous and stream event sources – "WHAT THE HELL DID YOU JUST SAY?! I’M GOING TO TEAR YOUR HEAD OFF AND USE IT AS A BASKETBALL!" roars the S3 product owner who stopped reading the title at the word "failed."
Introducing an AWS Management Console Visual Update (Preview) – No, when you log in as a customer and the entire console looks different (as happened last week), that’s not a "preview." That’s "the default way the console looks" now. Have fun redoing your screenshots and video tutorials, everybody; some PM needed to meet an OKR in order to dodge a PIP.
The new AWS Systems Manager experience: Simplifying node management – Oh screw off with the "Amazon Q" example in this post. "For example, you can ask Amazon Q to “show me managed instances running Amazon Linux 1” yeah see, here’s the problem with that, assuming you’re an AWS product manager who’s entirely too close to the Amazon Q product. As a customer, it doesn’t occur to me that I can ask a question like that, so discoverability is terrible. Further, if I ask a question that it doesn’t know how to answer, I feel like an idiot–and will never ask such a thing again. This is why rolling out Amazon Q across the console before it could do anything useful was such a blunder. People don’t reevaluate their initial impressions of products like this that require thoughtful interaction. But hey, you checked a box on a requirements sheet somewhere by "using AI," so bully for you, I guess. Screw the customers, right?
AWS announces Block Public Access for Amazon Virtual Private Cloud – I like this feature very much, particularly how granularly it can be applied to individual VPCs, subnets, exemptions, etc. This cannot have been easy to build and implement.
Load Balancer Capacity Unit Reservation for Application and Network Load Balancers – Oooh. This is a step towards solving the autoscaling problem: "the capacity you need, twenty minutes after you needed it."
Announcing Idle Recommendations in AWS Compute Optimizer – This just obviated a pile of overpriced tooling in a bunch of third-party products with something that quite frankly is better at it. I spoke to the team about this one and went in highly skeptical–but from every angle I’m able to look at this from, they got it right. My mostly-idle dev instance that still has to be up all the time? It doesn’t suggest turning it off like the festering pile of crap that is Trusted Advisor does. The only use cases I can construct where it’d suggest turning off something that shouldn’t be turned off border on the absurd.
Announcing Savings Plans Purchase Analyzer – Hell. Yes. This solves the problem posed by every client I’ve ever suggested buy a savings plan: "The analyzer says that if I commit to $X, I’ll save $Y in spend. If I make a different commitment, how much will I save then?" The answer has either been "uhh… some money!" or else requires a bunch of analysis tooling be cobbled together to answer the question. We went with the latter, and I’m thrilled to be able to throw that crappy kluge onto the dust pile of history and have something that customers can model with themselves.
AWS Lambda turns ten – looking back and looking ahead – Jeff Barr opines thoughtfully on the decade of AWS Lambda. It’s been a long, strange road. I wonder where it’s going next.
Boost Engagement with AWS and Amazon Ads – If you read this article not from the perspective of a marketer, but rather as a human being who’s on the receiving end of all of this? It sounds abhorrent. It’s why I trashed my Echo Show this month after it wouldn’t stop showing ads on its home screen; I didn’t spend $250 to install a billboard for Amazon products in my kitchen.
Build fullstack AI apps in minutes with the new Amplify AI Kit – I have no idea what makes this different than App Studio, and honestly I’m not sure that I care to know. "More AI slop" is how I read it.
Important changes to CloudTrail events for AWS IAM Identity Center – If your CloudTrail consumption tooling breaks soon, this is likely why. I don’t know as I’d have released it the same week as the pre:Invent crush, but what do I know? I’m merely an overwhelmed customer.
Tools
aws-gate is a great way to use AWS Systems Manager Session Manager Manager Manager to connect to instances via their names, instead of their instance IDs. Almost like a human might use!
… and that’s what happened Last Week in AWS.
