KCNA flashcards
KCNA · Build recall with a focused review session.
32 cards. Read the term and the definition here. The deck at the top of this page is optional.
- etcd
- Distributed key-value store that holds all Kubernetes cluster state; every object lives here, and loss of quorum makes the cluster unable to accept writes.
- kube-apiserver
- Front end of the control plane; it validates and serves the Kubernetes REST API and is the only component that reads and writes etcd directly.
- kube-scheduler
- Control-plane component that picks a node for each new Pod by filtering feasible nodes and scoring them; it never starts containers itself.
- kubelet
- Node agent that watches the API server for PodSpecs bound to its node, starts containers through the CRI, and reports pod and node status.
- Kubernetes Pod
- Smallest deployable unit; one or more containers that share a network namespace, IP address, and optionally mounted volumes on the same node.
- Deployment versus StatefulSet
- Deployments run interchangeable pods with rolling updates; StatefulSets give stable network names, ordered start and stop, and per-pod persistent storage.
- Kubernetes Service types
- ClusterIP exposes a virtual IP inside the cluster, NodePort opens a port on every node, LoadBalancer provisions an external balancer, ExternalName maps to a DNS CNAME.
- Requests versus limits
- Requests reserve resources for scheduling and guarantee a minimum; limits cap usage through cgroups, causing CPU throttling or, for memory, OOM termination.
- Container versus virtual machine
- Containers share the host kernel and isolate processes, so they start fast and stay small; virtual machines run a complete guest OS on a hypervisor.
- OCI image layers
- Layers are read-only and content-addressed, so identical layers are cached and shared across images; a writable layer is added only when a container runs.
- Container Runtime Interface
- CRI is the interface between the kubelet and container runtimes such as containerd or CRI-O, allowing runtimes to be swapped without changing Kubernetes.
- Horizontal Pod Autoscaler
- Controller that changes a workload's replica count based on observed metrics, typically CPU utilization measured as a percentage of the requested CPU.
- Rolling update parameters
- maxSurge is how many extra pods may exist during a rollout; maxUnavailable is how many may be down; both accept integers or percentages.
- Liveness readiness startup probes
- Liveness restarts an unhealthy container, readiness removes a pod from Service endpoints, and startup shields slow-starting containers from liveness checks until the first success.
- Self-healing reconciliation
- Controllers continuously compare desired state held in the API server with observed state and act on the difference, so deleted or failed pods are replaced automatically.
- Namespace scoping
- Virtual cluster partition used for name uniqueness, quotas, and RBAC scoping; objects in different namespaces may share a name, but names must be unique within one namespace.
- Helm chart
- Package of Kubernetes manifest templates plus default values and metadata; Helm renders the templates and tracks each installed instance as a release.
- GitOps
- Operational model where a Git repository holds the desired state and an agent inside the cluster continuously reconciles live state against it, detecting or reverting drift.
- Canary deployment
- Release pattern that sends a small share of traffic to the new version, checks error and latency metrics, then gradually increases traffic or rolls back.
- Blue-green deployment
- Release pattern with two identical environments; traffic switches from blue to green in one step, and rollback means switching the pointer back.
- Kustomize
- Kubernetes-native configuration tool that composes a base set of manifests with per-environment overlay patches, avoiding a templating language entirely.
- Immutable image tags
- Tags such as a git commit SHA or an image digest identify one exact build; mutable tags like latest make deployments unreproducible and break deterministic rollback.
- Progressive delivery
- Extension of canary and blue-green releases that gates the rollout on automated metric analysis and triggers rollback when defined thresholds are breached.
- Declarative deployment
- Approach where the desired end state is submitted to the API server and controllers converge toward it, rather than issuing step-by-step imperative commands.
- Twelve-factor app
- Methodology for cloud native applications: configuration in environment variables, stateless processes, disposability, dev and prod parity, explicit dependencies, and logs as event streams.
- Three pillars of observability
- Metrics are aggregated numeric time series, logs are discrete timestamped events, and traces follow one request across services; together they explain system behavior.
- Service mesh
- Infrastructure layer of sidecar or node proxies that adds mutual TLS, retries, traffic splitting, and telemetry to services without changing application code.
- CAP theorem
- A distributed data store facing a network partition must sacrifice either consistency or availability; partition tolerance is assumed because real networks drop messages.
- Loose coupling
- Components interact through stable interfaces or asynchronous messages, so one service can change, scale, or fail without breaking its consumers.
- Serverless cold start
- Scale-to-zero platforms must initialize a runtime on the first request after an idle period, adding latency that already-warm instances do not incur.
- Idempotency in distributed systems
- An operation is idempotent when repeating it has the same effect as performing it once; required because retries are unavoidable under at-least-once delivery.
- CNCF
- Cloud Native Computing Foundation, which hosts projects such as Kubernetes, Prometheus, Envoy, and containerd and grades them as sandbox, incubating, or graduated.