I'm trying determining whether karpenter makes sense when you aren't automatically consolidating with: consolidationPolicy: WhenUnderutilized I understand that if consolidation is happening, that I could see a great cost benefit. However if I'm not using consolidation (as some of my workloads just can't be stopped and restarted), then is there still a value in Karpenter? Without auto-consolidation, I could see a large node spinning up and the workload completing, but if there is still a single long lived app there, that node would continue to run and eat a lot of cost. now if that node is ev ..read more

I need to synchronize my local files with a pod every time I save a file in VSCode. Here are the constraints and approaches I have considered: Constraints: I want to use VSCode to push changes on save. I cannot use GitHub, GitLab, or perform any git push operations on the pod. I can get the pod's IP address using kubectl get pods -o wide | grep username. Approaches Considered: Using VSCode's RSYNC Extension: I am considering using the VSCode RSYNC extension to push changes automatically on save. Example settings.json configuration for Sync-Rsync: { "rsync.target": "user@pod-ip:/path/to ..read more

We are new to K8s. From a K8s management perspective, is it a better choice to run 2 separate EKS clusters (with one cluster managing only Linux-based nodes and the other cluster managing only Windows-based nodes) or run a cluster with both Linux-based nodes and Windows-based nodes ? We understand that Windows containers don't have good K8s ecosystem supporting tools. We are not clear in a heterogenous cluster of Linux and Windows nodes would complicate EKS cluster management. Thank you ..read more

I am trying to copy all the files from an image to a volume. However, kubernetes seems to not interpret the wildcard correctly. So, when I use the following, it fails with "not found": initContainers: - name: init-container-name image: <image path> command: - cp - /foo/bar/* - /output How can I get Kubernetes to interpret the * correctly? Is there a way to copy the contents without the surrounding bar folder? I can copy the whole folder: initContainers: - name: init-container-name image: <image path> command: - cp - -r - /foo/bar - /output But this results ..read more

I have an Azure Kubernetes Service (AKS) cluster that I built using Terraform. As part of the provisioning of the cluster, I also created the App Gateway as part of and integrated with the cluster. resource "azurerm_kubernetes_cluster" "aks" { name = var.aks_cluster_name resource_group_name = var.aks_rg location = var.aks_location dns_prefix = var.dns_prefix automatic_channel_upgrade = var.upgrade_channel private_dns_zone_id = data.azure ..read more

I have a ongoing problem on Kubernetes. I'm currently using: Kubernetes (1.30.2) on Master and Worker nodes. Calico (3.27.0) Metallb (0.14.5) I came across a problem where, if an unexpected course of events occur where one of my worker nodes reboots or shuts down (either maintenance or someone rebooting the VM), the worker node is unable to join back to the master node. When I ping the master node from the rebooted or recently shut down worker node, it becomes Destination Port Unreachable. It becomes a huge hassle since the only workaround that I do everytime for something like this, i ..read more

I'm currently running a Selenium Grid setup on a Kubernetes cluster, with edge nodes configured to scale between a minimum of 5 replicas and a maximum of 100 replicas. My use case involves executing 1000+ concurrent sessions, where each session takes at least 10 minutes to complete. This means that each node will run approximately 10 sessions simultaneously, and the total time to complete all sessions will be around 100 minutes. Consequently, the last session will need to remain in the queue for up to 90 minutes. Given this scenario, I have a few questions about the following Selenium Grid env ..read more

I am trying to display the rate of failing Kubernetes jobs using the kube_job_failed metric. It will display 1 job failed, 0 if not. I had the plan to take the rate()and then aggregate using sum() which does not work: Note that the first failing job happened a little while after 11:00, so something is happening. I am assuming the problem is that the specific metric for the failing/successful job does not exist before it fails/completes. So the rate of change for all jobs is zero. The closest I came to a solution is using the sum() operator but I can not figure out how to get the rate of chang ..read more

I'm running docker-desktop with WSL2.0 ubuntu on windows. I created a k8s deployment in the wsl ubuntu host. The deployment has one pod where I'm running a simple server program(runs forever, accepts udp connections, prints received buffer from client("Hi Server"), returns "Hi Client"). Server is running on port 5000. netstat inside the container gives me this - # netstat -au Active Internet connections (servers and established) Proto Recv-Q Send-Q Local Address Foreign Address State udp 0 0 0.0.0.0:5000 0.0.0.0:* I've ..read more

We're currently trying to provision a High-Availability ArgoCD instance onto our kubernetes cluster using the argocd helm chart (7.2.0) to provision our applications. Our infrastructure is on AWS and put together via AWS CDK written in TypeScript. So far, we've managed to deploy the argocd helm chart using the following cdk snippet: const argoCDHelmChart = this.cluster.addHelmChart('argocd-helmchart', { repository: 'https://argoproj.github.io/argo-helm', chart: 'argo-cd', version: '7.2.0', namespace: 'argocd', release: 'argo-cd', values ..read more