https://lckhd.eu/FXLuN5 Kubernetes Pods are built to be disposable, which is great until your app needs to actually store something. The moment a Pod gets replaced, its filesystem is gone.

https://lckhd.eu/j3ZKBt LLM-based tools are useful for many things but they don't always know the latest details. Using Skills is one way to help get better responses than general-purpose AI for cases like creating Flux manifests. With general LLM use you will likely get responses that include

https://lckhd.eu/BntaVc Storing a credential in GitHub Actions Secrets may give lots of confidence but as with almost anything around security there are nuances to be aware of.

https://lckhd.eu/CUHWdT People expect voice-enabled AI agents to be interactive and would like to be able to have a more interactive back and forth discussion than each getting a turn and processing. Managing WebSocket connections, concurrent audio, and conversational state from scratch can make

https://lckhd.eu/LMkL6c AWS Lambda MicroVMs is one of the most interesting releases in a long time and this session from Ajay should be a good one.

https://lckhd.eu/Xc5d2I As teams roll out more AI agents, connecting them starts to become more of a concern. Each new integration means another point-to-point link, another set of credentials, and more custom routing. One approach involves the Agent-to-Agent (A2A) protocol. The example below uses

https://lckhd.eu/NcahJx Understanding how Kubernetes works under the hood is worth the effort, even if you only touch it occasionally.

https://lckhd.eu/nbkLyr Most teams set up Elastic Container Registry (ECR), connect it to the Elastic Kubernetes Service (EKS), and move on. There are some things you need to watch out for to ensure things work smoothly with this combo. This includes things like expired tokens during deployments,

https://lckhd.eu/JVzSG1 For many cases using containers on AWS, the Elastic Container Service (ECS) is the best fit. Auto scaling with ECS on AWS recently picked up high-resolution metrics, moving evaluation from 60-second to 20-second intervals. It sounds like a small tweak, but the effect on how

https://lckhd.eu/SOi2SU Testing Agentic workflows is going to be one of the biggest challenges for many teams. Testing traditional software is fairly settled using unit tests and integration tests with deterministic outputs you can assert on.

https://lckhd.eu/xu9ZYc Most of us learn IAM as users, roles, and policies, and stop there. But once an organisation runs dozens or hundreds of AWS accounts, you need controls that sit above the individual identity. The article below talks about the two features that handle exactly that. Both

https://lckhd.eu/rZN3Q7 Keeping your Kubernetes clusters up to date has always been important but one challenge is that control plane upgrades have always been a one-way door.

https://lckhd.eu/2J6NlM Gateway API is the newer approach to managing traffic with Kubernetes. When using traditional Kubernetes Ingress, things like traffic splitting or header-based routing can be a challenge. Every controller has its own annotations, and none of them are portable across

https://lckhd.eu/hRmda3 Many teams have been using Transit Gateway on AWS for a long time but may be considering moving to Cloud WAN.

https://lckhd.eu/SwhEpU Many teams like to have a place to run code locally without spending cloud $$$s. It isn't totally clear how much value and reproducability these offer but it's great to have options.

https://lckhd.eu/kymWvP One of the key things that gets left out with many agentic AI solutions is security. As apps move from prototype into production, the security questions become more real.

https://lckhd.eu/QXu4QQ If you have run workloads on the Elastic Kubernetes Service (EKS), you know the challenges of setting up IAM permissions for pods. IAM Role for Service Accounts (IRSA) involves juggling OIDC providers, trust policies, and annotations across every cluster. EKS Pod Identity

https://lckhd.eu/Cm22sx AWS introduced Lambda Durable Functions about 6 months ago and it has some really interesting use cases. One case could be with multi-agent AI workflows. These are tricky to run reliably. Each step is slow, costly, and prone to failure, and a half-finished run wastes both

https://lckhd.eu/pteNpm Using the Elastic Container Registry (ECR) to store your container images works really well for your apps running in AWS.

https://lckhd.eu/or8Etr S3 has been around forever and in recent times, AWS has kept expanding what you can do with it. Recently S3 Files was announced and it has a lot of use cases where it makes sense. One typical pattern with Lambda functions is downloading objects to /tmp, processing them, and

https://lckhd.eu/4c95AK Using Argo Rollouts is one way to update applications in Kubernetes with minimal downtime. When you have a new version of your container, you typically update the image spec in a standard deployment and your pods get torn down and recreated, which means an interruption.

https://lckhd.eu/mVM4Vh When setting up Kubernetes clusters on your own, you have to come up with an approach for Load Balancers. The Kubernetes services from cloud providers handle setting these up but on bare-metal Kubernetes there's no cloud controller to hand you an external IP. The article

https://lckhd.eu/jXrNd2 I recently wrote an article which covers AWS Lambda Powertools for Python. It's the library I add to every Lambda function before writing any business logic, and I wanted to put everything I've learned from using it across real projects into one comprehensive guide. The

https://lckhd.eu/tanEPj Most people who have used Kubernetes have leaned on ConfigMaps and Secrets at some point. I like understanding the tools I use as deeply as I can,

https://lckhd.eu/570Pn6 Learn by example! The article below shows a serverless to-do app that goes beyond the usual browser-storage tutorial.

https://lckhd.eu/SZhlPO If you manage Kubernetes configs across dev, staging, and prod, you know the pain of copy-pasting near-identical YAML files and editing replica counts by hand.

https://lckhd.eu/5qM3re New Kubernetes versions land every few months, which is great, but it also means upgrading regularly. Amazon EKS v1.36 is out now which means you should start looking at moving up if you haven't already.

https://lckhd.eu/oJZwk7 AWS released Lambda durable functions a number of months ago and it fits well for many use cases. Contact centers handle millions of voice calls a month, but turning those recordings into useful insights can be tedious. The hard part can be the retry logic, state

https://lckhd.eu/23JVj7 Most teams plan for infra and services to break down in production but many don't really test it and only find out too late that it doesn't work.

https://lckhd.eu/M7XpFJ Terraform gives you a cluster you can stand behind, but it's difficult to keep your Terraform state in sync with the real deployed resources.

https://lckhd.eu/R4RcxB A good percentage of apps being built today support multiple tenants and search across these has always meant a tradeoff. Either you isolate each tenant for clean data boundaries or pool them to keep costs down. The recent announcement of OpenSearch Serverless next-gen is a

https://lckhd.eu/hisvLM AWS just announced customer-routed control plane egress for EKS, which lets you route Kubernetes control plane traffic through your own VPC.

https://lckhd.eu/HYALRD You have likely dealt with the spreadsheet ritual where someone pastes a file in Slack, you scroll through 40 columns, run a query somewhere, then copy the result back.

https://lckhd.eu/vphNEA Having good monitoring is key to a successful platform running in the cloud. There are many tools out there you can use for Kubernetes monitoring on AWS with the Elastic Kubernetes Service (EKS) but one good combo involves using Prometheus, Grafana, Alertmanager, Node

https://lckhd.eu/XjwG5G AWS just announced Lambda MicroVMs, a new serverless approach built for running untrusted user or AI generated code in isolated environments that maintain state.

https://lckhd.eu/cSHwcL After attending the AWS Community Day Montreal this past weekend I am convinced more than ever these types of events are by far the best

https://lckhd.eu/JTtsCV Many people heard that "Kubernetes removed Docker support" a few years ago and there was a lot of confusion.

https://lckhd.eu/UzpFna AWS recently announced the Workload Credentials Provider, giving Serverless apps a new way to reach services like Secrets Manager.

https://lckhd.eu/KoVcPG GitHub Actions is widely deployed, but most teams only use a fraction of what it offers. The example below goes from a Hello World workflow to a real deploy pipeline,

https://lckhd.eu/yf0tZR Using the Elastic Kubernetes Service (EKS) on AWS makes using Kubernetes easier than setting up from scratch. One complication that many teams don't handle well is regularly updating Kubernetes versions as they get released (and the old ones get deprecated and start costing

https://lckhd.eu/mgt6fM Another security related example where a security audit at a fintech went quiet over a single pull request. A workflow using pull_request_target had quietly exposed production AWS credentials to any contributor for three months but nobody noticed. The audit caught it before

https://lckhd.eu/iritwa Using the Elastic Container Service (ECS) is a great way to get started on with containers on AWS and using Fargate compute is a good starting point. ECS/Fargate handles your infrastructure, but that convenience has a tradeoff where you lose visibility into what's happening

https://lckhd.eu/y4rt5g The article here discusses how a developer documented a database password in a PR description, base64-decoded, on a repo with forty contributors. When flagged, he shrugged: "the value is base64 encoded." That moment captures the core problem with Kubernetes Secrets in one

https://lckhd.eu/VO01d0 Using agents with objects in S3 has some limitations. The agents can list the objects but have no real sense of what they contain or which ones are relevant.

https://lckhd.eu/BgUZKD We have all moved from simple chatbots to agents to fully agentic solutions and from simple Github Co-pilot completion to coding harnesses like Claude Code that are super powerful. Having a way to build your own harness with Bedrock AgentCore is a very interesting

https://lckhd.eu/QBTmIR Most teams set up Elastic Container Registry (ECR), connect it to the Elastic Kubernetes Service (EKS), and move on. There are some things you need to watch out for to ensure things work smoothly with this combo. This includes things like expired tokens during deployments,

https://lckhd.eu/dSHMgD Keeping the compute in your Kubernetes cluster at optimal scale is key to not wasting $$$s. Autoscaling has evolved beyond resizing node groups.

https://lckhd.eu/kDLWfC Setting CPU limits in Kubernetes may feel like basic hygiene but you really need to consider if it makes sense. Many docs and best-practices lists tell you to do it, but that one line might be the reason your tail latency keeps spiking while your dashboards look perfectly

https://lckhd.eu/mDyzvt There have been lots of CloudWatch announcements for native OpenTelemetry and PromQL support this year including previews for

https://lckhd.eu/04Hc6I Security in Kubernetes clusters is not really understood well. By default, anyone with kubectl access can ship a pod that runs as root, pulls :latest, or ignores resource limits. These deploy fine and (in many cases), nobody notices or understands, and three weeks later it