DevOps Daily with Fexingo: CI/CD, Kubernetes, and Modern Software Operations

Lucas and Luna dive into the hidden performance cost of running a service mesh in Kubernetes. Using Istio and Linkerd as examples, they unpack how sidecar proxies add 2-5 milliseconds of latency per hop, and how mesh overhead compounds in deep call chains. They walk through a real-world scenario where a 12-microservice transaction saw tail latency spike from 50ms to 210ms after enabling mutual TLS and traffic policies. The hosts discuss tuning strategies: reducing proxy CPU limits, disabling unused features, and choosing between Istio's Envoy and Linkerd's Rust-based proxy. They also cover when a service mesh is premature — and when it's worth the cost. This episode assumes you already know what a service mesh does; it's about the tradeoffs you rarely see in vendor docs.

#Kubernetes #ServiceMesh #Istio #Linkerd #EnvoyProxy #Latency #Microservices #SidecarProxy #MutualTLS #Performance #DevOps #CloudNative #CNCF #K8sNetworking #Technology #FexingoBusiness #DevOpsDaily #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

Episode 28 of DevOps Daily with Fexingo dives deep into a common but often overlooked performance pitfall: Kubernetes Persistent Volume Claims (PVCs). Lucas and Luna unpack a real-world case where a fintech company's database workloads suffered 40% latency spikes due to misconfigured PVC storage classes and access modes. They explain how volume binding modes, reclaim policies, and CSI driver settings can silently tank application performance, and share concrete tuning strategies like using WaitForFirstConsumer binding, ReadWriteOncePod access mode, and monitoring volume attach times. A must-listen for any DevOps engineer running stateful workloads on Kubernetes.

#Kubernetes #PersistentVolumeClaims #StoragePerformance #DevOps #StatefulWorkloads #K8sStorage #VolumeBindingMode #ReadWriteOncePod #CSI #Latency #Fintech #DatabasePerformance #KubernetesPerformance #CloudNative #CI/CD #Technology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

Episode 27 of DevOps Daily with Fexingo: Lucas and Luna dive into a silent performance killer in Kubernetes clusters—how Persistent Volume Claims with default storage classes can cause latency spikes and throughput bottlenecks. Using a real-world case from a fintech startup running Cassandra on AWS EBS gp2 volumes, Lucas explains why their 99th percentile read latency jumped from 5ms to 120ms under moderate load. The episode covers storage class tuning, volume binding modes, and why your CSI driver configuration matters more than you think. No fluff—just actionable insights for DevOps engineers managing stateful workloads in production.

#Kubernetes #PersistentVolumeClaims #StorageClass #Performance #DevOps #StatefulWorkloads #Cassandra #AWSEBS #CSI #Latency #Throughput #Fintech #Infrastructure #Technology #CloudNative #FexingoBusiness #BusinessPodcast #DevOpsDaily

Keep every episode free: buymeacoffee.com/fexingo

Episode 26 of DevOps Daily with Fexingo. Lucas and Luna dive into why Kubernetes rollbacks are riskier than most teams realize. They examine a case where a simple image tag rollback silently reverted a critical security fix, breaking compliance for days. They discuss why declarative state doesn't protect you from bad rollouts, how Helm and Argo CD handle (or fail to handle) rollback semantics, and what a safer rollback strategy looks like — including version pinning and canary analysis gates. Specific examples from a real-world incident involving a fintech startup's production outage in April 2026.

#Kubernetes #DevOps #Rollbacks #Helm #ArgoCD #CI/CD #IncidentResponse #ReleaseManagement #DeclarativeState #VersionPinning #CanaryDeployments #SiteReliabilityEngineering #Compliance #Technology #CloudNative #FexingoBusiness #DevOpsDaily #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

Episode 25 of DevOps Daily with Fexingo dives into a specific Kubernetes failure mode: Pod Disruption Budgets (PDBs) that prevent node drains during planned maintenance. Lucas and Luna break down a real incident where a cluster's PDBs blocked critical node updates, causing a multi-hour outage. They explain how PDBs work, why common configurations like minAvailable: 100% can lock you into a deadlock, and the practical fix: using percentage-based budgets with a small tolerance. The episode includes a concrete simulation using a three-replica deployment and two-node cluster. No fluff, just the one angle that will save your next maintenance window.

#Kubernetes #PodDisruptionBudget #NodeDrain #DevOps #SiteReliabilityEngineering #CloudNative #ContainerOrchestration #HighAvailability #MaintenanceWindow #ClusterOperations #K8sFailureMode #ProductionIncident #Technology #FexingoBusiness #BusinessPodcast #DevOpsDaily #ResilienceEngineering #RollingUpdate

Keep every episode free: buymeacoffee.com/fexingo

Episode 24 of DevOps Daily with Fexingo digs into a subtle but painful Kubernetes performance trap: the relationship between pod resource limits (CPU throttling) and application latency. Lucas and Luna examine a real-world case from a fintech company where setting CPU limits too low caused tail-latency spikes during normal traffic. They explain the difference between compressible and incompressible resources, how the Completely Fair Scheduler enforces CPU limits, and why setting requests without limits — or using the 'burst' capability wisely — often delivers better performance. The hosts also cover monitoring for throttling, the role of cgroup v2, and the trade-offs of over-provisioning. This episode is essential for any team running latency-sensitive workloads on Kubernetes.

#Kubernetes #DevOps #Performance #CPULimits #Latency #Throttling #CFS #CgroupV2 #PodResources #Fintech #SiteReliabilityEngineering #ContainerOrchestration #CloudNative #Infrastructure #Monitoring #FexingoBusiness #BusinessPodcast #Technology

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna dive into Kubernetes pod priority classes and preemption—a feature designed to ensure critical pods get resources, but one that can cause cascading failures and starvation for lower-priority workloads. They walk through a real scenario at a mid-size fintech that saw critical database pods preempting batch jobs, only to have those batch jobs reschedule and overwhelm the cluster. Lucas breaks down how priority classes work under the hood, the risks of misconfigured preemption policies, and the operational discipline needed—like setting PriorityClass namespaces and avoiding the default zero-priority trap. Luna challenges whether the feature is safe for multi-tenant clusters and shares a story about a team that accidentally set a pod priority higher than the cluster autoscaler's headroom. They close on the trade-off between reliability and fairness, and why priority classes should be a last resort, not a first instinct.

#Kubernetes #PodPriority #PodPreemption #ClusterOperations #Scheduling #K8sBestPractices #DevOps #CloudNative #FintechCaseStudy #PriorityClass #ResourceManagement #MultiTenancy #ClusterAutoscaler #KubernetesStarvation #BatchJobs #CriticalWorkloads #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

In this episode of DevOps Daily, Lucas and Luna break down why graceful shutdown is failing in Kubernetes clusters. They examine the lifecycle of a pod being terminated, explaining the sequence of SIGTERM, terminationGracePeriodSeconds, and preStop hooks. Using a real-world example of a Node.js web service that was dropping thousands of in-flight requests during rollouts, they reveal the hidden defaults that cause timeouts and busted connections. They walk through how to set proper terminationGracePeriodSeconds, configure readiness probes to align with shutdown, and test SIGTERM handling in your app. If you're managing microservices on Kubernetes, this episode will save you from silent data loss on your next deploy. Tagged with Kubernetes, graceful shutdown, SIGTERM, pod lifecycle, DevOps, microservices, terminationGracePeriodSeconds.

#Kubernetes #GracefulShutdown #SIGTERM #PodLifecycle #DevOps #Microservices #TerminationGracePeriodSeconds #PreStopHook #CloudNative #SiteReliabilityEngineering #ContainerOrchestration #RollingUpdate #ReadinessProbe #NodeJS #KubernetesPods #TechPodcast #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

A 45-second container shutdown sounds harmless until it costs you a production database connection pool. Lucas and Luna explore why most Kubernetes pods ignore SIGTERM, how a missing preStop hook turned a routine deployment into a 15-minute outage for a European payments platform, and the three-line fix that cuts pod termination from 30 seconds to under one second. They drill into the mechanics of the Linux process lifecycle inside containers, show why sleep 30 in a preStop hook is a common anti-pattern, and walk through a concrete health-check-aware shutdown sequence using Kubernetes 1.27's new pod status conditions. Listeners walk away with a specific checklist they can apply to any deployment tomorrow morning.

#Kubernetes #PodLifecycle #GracefulShutdown #SIGTERM #preStopHook #ContainerOrchestration #DevOps #SiteReliability #ProductionIncident #LinuxProcesses #PodTermination #CloudNative #Technology #FexingoBusiness #BusinessPodcast #SoftwareEngineering #Infrastructure #Observability

Keep every episode free: buymeacoffee.com/fexingo

In this episode, Lucas and Luna dig into the mechanics of Kubernetes Horizontal Pod Autoscaler — specifically why it often fails to keep up with sudden traffic spikes. They walk through a real-world scenario from a retail platform that saw request latency spike from under 100ms to over 2 seconds during a flash sale. The root cause wasn't resource limits or cluster size — it was the default HPA scaling metrics and cooldown windows. Lucas explains how target CPU utilization, stabilization windows, and custom metrics interact, and why relying solely on CPU-based HPA leaves you vulnerable. They discuss the alternative: using Kubernetes Event-driven Autoscaling (KEDA) with request-based metrics. If you're running Kubernetes in production and haven't stress-tested your HPA configuration, this episode will save you from a late-night incident.

#Kubernetes #PodAutoscaling #HPA #KEDA #SiteReliabilityEngineering #CloudNative #DevOps #IncidentResponse #Scalability #Microservices #Containers #ProductionEngineering #TrafficSpikes #Metrics #Technology #FexingoBusiness #BusinessPodcast #DevOpsDaily

Keep every episode free: buymeacoffee.com/fexingo