dstack’s Post

🚀 Private cloud automation is broken. Too many tools. Too many scripts. Not enough clarity. Today at VergeIO, we’re changing that and we are excited to announce... End-to-end automation for VergeOS, with native support for Packer and Ansible. That means your infrastructure can finally live inside the same IaC workflows your teams already trust. Here’s the full chain, running on a single, integrated platform: 🔹 Packer to build standardized golden images 🔹 Terraform to provision VDCs, networks, and VMs 🔹 Ansible to enforce configuration and policies 🔹 Prometheus to monitor performance and health For VMware exit teams, this is big. You can build a clean VergeOS environment with code, migrate workloads with repeatable patterns, and stop inheriting years of legacy template debt. For CSPs and service providers, it means: Consistent onboarding across customers Lower operational overhead per tenant Faster time from “quote signed” to “platform live” Infrastructure that behaves the same every time. Defined in code. Delivered on VergeOS. #VergeOS #VergeIO #VMwareAlternative #Automation #Packer #Ansible #Terraform #Prometheus #InfrastructureAsCode #PrivateCloud #PlatformEngineering #DevOps #CSP

Great move by Verge.io. Adding HashiCorp Packer and Ansible support closes the loop on a true end-to-end automation chain for VergeOS. With image creation, provisioning, configuration, and monitoring all codified — teams can now treat their private cloud infrastructure like code, not just a one-off deployment. This significantly boosts repeatability, reduces drift, and makes large-scale environments far more manageable. Looking forward to seeing how this accelerates VMware exit strategies and streamlines operational workflows. https://lnkd.in/e8aMctv9

“Why are my Kubernetes costs going up even though I’m using cheap nodes?” This is one of those Kubernetes realities that surprises a lot of teams. In Kubernetes, cheap nodes can be more expensive than expensive nodes. Here’s why: Most people do capacity planning purely on CPU and memory. But Kubernetes scheduling has another silent limiter pod density. On cloud providers, smaller (cheaper) nodes usually come with: Fewer ENIs Fewer IPs per ENI A hard limit on how many pods can be scheduled on that node. So what happens in practice? Your node still has free CPU and memory. But it can’t schedule more pods because it ran out of IPs, Kubernetes adds more nodes. You pay for more infrastructure while leaving resources unused. From the outside, it looks like you’re scaling correctly. In reality, you’re bleeding efficiency. Bigger (more “expensive”) nodes often allow: Higher pod density Better IP availability Fewer nodes for the same workload Lower overall cluster cost Real capacity planning in Kubernetes isn’t just CPU and RAM. It’s CPU + memory + network limits + pod density. If you don’t account for density, autoscaling can quietly turn into over-provisioning. Kubernetes is one of the few places where unit price doesn’t tell the full cost story effective utilization does. #Kubernetes #CloudEngineering #Infrastructure #CostOptimization #DevOps

Learn Kubernetes Weekly 162 hot off the press, and those are my top picks: 🔐 Secure Your Kubernetes Cluster with Kong and Keycloak by Armel Tandeau de Marsac: This article shows how to use the Kong OIDC plugin together with Keycloak to secure cluster services and HTTP routes at the API gateway level. 🌐 Unpacking Cluster Networking for Amazon EKS Hybrid Nodes by Sheng Chen: This tutorial walks you through how to set up networking for a cloud-to-on-premises hybrid Amazon EKS Hybrid Nodes cluster, covering compatible CNIs, CIDR planning, routing, and load-balancer options for hybrid workloads. 📊 VictoriaLogs vs Loki: Benchmarking Results by Harshit Luthra: This benchmark shows that VictoriaLogs delivered ~94% lower query latency, ~40% less storage usage, and used less than half the CPU and RAM compared to Loki on a 500 GB/7-day log workload. 🛡️ Kexa: Cloud Compliance at Scale. This tool enables you to scan and enforce compliance across multi-cloud infrastructure with customizable YAML rules, alerts, and integrations. If you prefer to read online (or receive it next week in your inbox), you can do so here: https://lnkd.in/gDY67id3

To all my hyperscale Kubernetes buddies out there...this is some very cool tech! The obvious use cases illustrated in the video, but I've seen folks try to build similar architectures for DR applications in OT environments. Rugged, portable, and built for the edge—#BoozAllen’s EdgeXtend delivers hyperscale cloud services to remote, denied, and disconnected environments. See it in action: https://lnkd.in/evPTU3CZ

.“Why are my Kubernetes costs going up even though I’m using cheap nodes?” This is one of those Kubernetes realities that surprises a lot of teams. In Kubernetes, cheap nodes can be more expensive than expensive nodes. Here’s why: Most people do capacity planning purely on CPU and memory. But Kubernetes scheduling has another silent limiter pod density. On cloud providers, smaller (cheaper) nodes usually come with: Fewer ENIs Fewer IPs per ENI A hard limit on how many pods can be scheduled on that node. So what happens in practice? Your node still has free CPU and memory. But it can’t schedule more pods because it ran out of IPs, Kubernetes adds more nodes. You pay for more infrastructure while leaving resources unused. From the outside, it looks like you’re scaling correctly. In reality, you’re bleeding efficiency. Bigger (more “expensive”) nodes often allow: Higher pod density Better IP availability Fewer nodes for the same workload Lower overall cluster cost Real capacity planning in Kubernetes isn’t just CPU and RAM. It’s CPU + memory + network limits + pod density. If you don’t account for density, autoscaling can quietly turn into over-provisioning. Kubernetes is one of the few places where unit price doesn’t tell the full cost story effective utilization does. #Kubernetes #CloudEngineering #Infrastructure #CostOptimization #DevOps

The first production-ready version of our cloud-native virtualization platform is here. Whether you are building a private cloud from scratch or modernizing legacy infrastructure, Kubermatic Virtualization gives you the tools to run, protect, and scale workloads on your terms. What’s inside the box? It is an all-in-one platform powering modern private clouds for both VMs and Containers. 🔹 KubeOne: Automated lifecycle management for clusters (bare metal to tenant). 🔹Kubermatic Virtualization: The open-source virtualization layer. 🔹KubeOVN: Enterprise-grade SDN for advanced networking. 🔹 UI Console: A single pane of glass for end-to-end management. Experience High Availability (HA), Live Migration, and enterprise automation, all powered by Kubernetes. 👉 Get started today: https://hubs.li/Q03W5lLc0 #Kubernetes #Virtualization #CloudNative #PrivateCloud #OpenSource #Infrastructure #K8s

Why every Pod in Kubernetes gets its own IP address Kubernetes networking follows a simple but powerful rule: “Every Pod gets a unique IP, and all Pods can talk to each other without NAT.” Here’s why this design exists: 1. Pods are treated like standalone network endpoints. 2. Each Pod gets its own network namespace and virtual Ethernet interface. 3. The CNI plugin (Calico, Cilium, Flannel, etc.) assigns the Pod a unique IP. 4. Pod-to-Pod communication works directly across nodes because: a. CNI sets routing rules b. No NAT is required c. Every Pod is reachable flat across the cluster 5. When a Pod dies, its IP dies with it — a new Pod gets a new IP. Why Kubernetes does this: 1. Simplifies distributed application communication. 2. Makes service discovery easier. 3. Avoids port conflicts between containers. 4. Works reliably across multi-node clusters. 5. Allows network policies and service meshes to work cleanly. In simple words: Each Pod gets its own IP because Kubernetes treats Pods like independent machines in a flat, virtual network. #DevOps #k8s #eks #SRE #Cloud

[New Blog Post] This is the first post in a new series focused on VCF 5.2.2 on VxRail 8.0.361. In this article, I walk through a practical, end-to-end example that covers: - Deploying a VxRail Management Workload Domain cluster - Deploying VMware Cloud Builder for VxRail - Bringing up VCF 5.2.2 using a multi-vDS design - Applying the SDDC Manager token - Reviewing and validating the final deployment configuration If you’re planning or validating a VCF on VxRail deployment, this series is intended to provide real-world guidance and observations. https://lnkd.in/eFhAHNXk

AWS Storage Gateway now supports Nutanix AHV hypervisor - The AWS Storage Gateway service now supports the Nutanix AHV hypervisor as a deployment option for S3 File, Tape and Volume gateways. If you use Nutanix AHV hypervisor-based on-premises infrastructure, you can now deploy Storage Gateway in your environment to… https://lnkd.in/eYs-JqfB