| Image / container | A program packed with everything it needs / one running copy of an image; `docker run IMAGE` starts one | Images and containers |
| Container vs VM | Shares the host kernel / brings its own | Containers and VMs |
| Container | An ordinary process shown a smaller world | What a container is |
| pid, net, mnt, uts, ipc, user, cgroup | Namespaces: one wall per kind of thing | What a container is |
| cgroups | Limits on CPU, memory and process count | What a container is |
docker run --memory 256m --cpus 0.5 | Set the cgroup limits at run time | What a container is |
| Exit code 137 / 143 | 128 + 9 (SIGKILL, or out of memory) / 128 + 15 (SIGTERM) | Stopping and exit codes |
| OCI | Image, runtime and distribution standards: any tool builds, any runtime runs | The standard |
| containerd, CRI-O | Runtimes Kubernetes uses since dockershim was removed (1.24) | The standard |
crictl ps , crictl pods , crictl logs ID | Containers as the kubelet sees them, on the node itself | crictl |
docker run -d -p 8080:80 nginx | Publish container port 80 on host port 8080 | Ports |
docker ps logs exec -it … sh stop rm | List, read output, open a shell, stop, remove | Ports |
-p 127.0.0.1:8080:80 | Publish on the loopback address only; `-p 8080:80` listens on every interface | Publishing ports |
-v data:/var/lib/postgresql/data | A named volume: data outlives the container | Volumes |
| Bind mount | `-v $(pwd):/app`: a host folder inside the container | Volumes |
| Named volume / bind mount / tmpfs | Managed by Docker / a host path you name / memory only | Kinds of mount |
-e KEY=value / --env-file | Settings from outside the image | Environment |
| User-defined network | Containers find each other by name | Networks |
compose.yaml | services, ports, volumes, environment in one file | Compose |
docker compose up -d / down -v | Start everything / stop and delete volumes too | Compose |
depends_on: condition: service_healthy | Wait for the health check of the other service, not just its start | Start order |
| Layer | One per filesystem-changing instruction (RUN, COPY, ADD) | Layers |
docker build -t app:1.2 . | Build and tag an image | Layers |
| ENTRYPOINT / CMD | The program / its default arguments; in a pod, `command` / `args` | ENTRYPOINT and CMD |
| Cache order | Copy the dependency list and install before copying the code | The build cache |
| Tag vs digest | `app:1.2` can move / `app@sha256:…` never does | Tags and digests |
| imagePullPolicy | Default Always for `:latest` or no tag, IfNotPresent otherwise | Names and pull policy |
| Small base | alpine, slim or distroless instead of a full OS | Small images |
docker image history app | Size of each layer | Small images |
| Multi-stage build | `FROM … AS build` then `COPY --from=build` into a small image | Small images |
| Registry | Docker Hub, ECR, ACR, Artifact Registry; pulls only missing layers | Registries |
| Image index | One tag, a manifest per platform (linux/amd64, linux/arm64) | Inside a registry |
| Desired state | You state what should run; controllers keep it true | Why Kubernetes |
kind create cluster | A practice cluster in containers | Why Kubernetes |
| Control loop | Watch, compare, act | Desired state |
| API server, etcd, scheduler, controller manager | The control plane | Inside a cluster |
| kubelet, kube-proxy, container runtime | On every worker node | Inside a cluster |
/etc/kubernetes/manifests | Static pod files: API server (6443), etcd (2379), scheduler, controller manager | Where the parts run |
kubectl apply -f app.yaml | Send desired state to the API server | One request |
--dry-run=client -o yaml | Print the manifest a create command would make, and create nothing | Working with kubectl |
| apiVersion, kind, metadata, spec | Every manifest's shape; the cluster writes status | Manifests |
kubectl get pods -o wide / describe / explain | List, details and events, field docs | Manifests |
kubectl api-resources | Every kind with its apiVersion, short name and whether it is namespaced | API groups |
| Pod | One or more containers sharing an IP and volumes | The pod |
| Init container / sidecar | Runs to completion before the main containers / an init container with `restartPolicy: Always` | Init containers and sidecars |
| Deployment → ReplicaSet → Pods | Who owns whom | Ownership |
| Rolling update | maxSurge 25%, maxUnavailable 25% by default | Rolling update |
kubectl rollout status / undo deploy/web | Watch a rollout / go back | Rolling update |
kubectl set image deploy/web web=web:1.3 | Start a new rollout | Rolling update |
| maxSurge / maxUnavailable | 25% each by default; surge rounds up, unavailable rounds down | maxSurge and maxUnavailable |
| Deployment, StatefulSet, DaemonSet, Job, CronJob | Interchangeable pods, stable identity, one per node, run to completion, on a schedule | Other workloads |
| CronJob defaults | concurrencyPolicy Allow, backoffLimit 6, keeps 3 successful and 1 failed Job | Jobs and CronJobs |
| Labels and selectors | `app=web`; `kubectl get pods -l app=web` | Labels |
| Service | A stable virtual IP and name in front of changing pods | Services |
kubectl expose deploy web --port 80 | Make a Service for a Deployment | Services |
kubectl port-forward svc/web 8080:80 | Reach a Service from your laptop | Services |
| ClusterIP, NodePort, LoadBalancer | Inside only; a port on every node (30000 to 32767); a cloud load balancer | Service types |
| port / targetPort / nodePort | On the Service / on the container / on every node, 30000 to 32767 | Service ports |
web.shop.svc.cluster.local | A Service's DNS name: service, namespace | DNS |
ndots:5 | Names with fewer than 5 dots are tried with each search ending first | The search list |
| Ingress | HTTP routing by host and path; needs an ingress controller | Ingress |
| Gateway API | The newer, role-oriented successor to Ingress | Ingress |
| Gateway / HTTPRoute | `gateway.networking.k8s.io/v1`: the entrance / the routing rules attached to it | Gateway API |
| NetworkPolicy | All traffic allowed until a policy selects a pod; needs a CNI that enforces it | NetworkPolicy |
| ConfigMap / Secret | Settings / sensitive settings, passed as env vars or files | ConfigMaps and Secrets |
kubectl create secret generic db --from-literal=password=… | Make a Secret | ConfigMaps and Secrets |
kubectl rollout restart deployment/api | Replace the pods so that they read changed variables | When settings update |
| Base64 | Encoding, not encryption; enable encryption at rest for etcd | Secrets |
imagePullSecrets | Names a `kubernetes.io/dockerconfigjson` Secret for a private registry | Secret types |
| Requests | What the scheduler reserves for the pod | Requests and limits |
| Limits | CPU is throttled at the limit; memory over the limit is OOMKilled | Requests and limits |
| LimitRange / ResourceQuota | Defaults and bounds per container / a total for the namespace | Namespace limits |
| Guaranteed, Burstable, BestEffort | QoS classes: evicted last to first under pressure | Quality of service |
| emptyDir / PersistentVolume | Lives with the pod / outlives it | Lifetimes |
| PVC → PV ← StorageClass | A claim asks; a class makes the disk on demand | Claims and volumes |
volumeBindingMode: WaitForFirstConsumer | Make the disk only when a pod is placed, in that zone; the default is Immediate | StorageClass |
volumeBindingMode: WaitForFirstConsumer | Make the disk in the zone where the pod lands | Moving |
| RWO, ROX, RWX, RWOP | One node writes, many read, many write, one pod | Access modes |
| Static binding | Class equal, access mode offered, capacity at least the request; one claim per volume | Static volumes |
| Reclaim policy | Delete (default for dynamic volumes) or Retain | Reclaim |
| volumeClaimTemplates | A StatefulSet gives each replica its own claim | Reclaim |
| PV phases | Available, Bound, Released, Failed | Volume phases |
| Filter, then score | Rule nodes out, then pick the best | The scheduler |
| Topology spread | Spread replicas across zones and nodes | Placement |
requiredDuringSchedulingIgnoredDuringExecution | A hard rule at placement; not enforced on pods already running | Affinity, written down |
| nodeSelector, affinity | Pull pods towards nodes or other pods | Steering |
| Taint / toleration | `kubectl taint node n1 gpu=true:NoSchedule`; pods must tolerate it | Steering |
| Pending | No node passes the filters: check requests, taints, affinity | Steering |
| NoSchedule / PreferNoSchedule / NoExecute | No new pods / avoid if possible / also evict running pods | Taint effects |
| HPA | desired = ceil(current × metric ÷ target); 10% tolerance; 5 min scale-down window | Scaling pods |
kubectl autoscale deploy web --cpu-percent=70 --min=2 --max=10 | Create an HPA | Scaling pods |
kubectl autoscale deployment web --cpu-percent=50 --min=2 --max=10 | Makes a HorizontalPodAutoscaler (autoscaling/v2) | The HPA, written down |
| Cluster Autoscaler, Karpenter | Add nodes when pods do not fit | Scaling nodes |
| Authentication, authorisation, admission | Three gates for every API request | The API request |
| Role + RoleBinding / ClusterRole + ClusterRoleBinding | What, and who, in a namespace / cluster-wide | RBAC |
kubectl auth can-i delete pods --as ana -n shop | Test a permission | RBAC |
| ClusterRole + RoleBinding | The cluster-wide role's rules, granted in one namespace only | RBAC, written down |
| Service account | A pod's identity; short-lived bound tokens | Service accounts |
system:serviceaccount:shop:deployer | A service account's user name, for `kubectl auth can-i --as` | The token |
| securityContext | `runAsNonRoot`, `readOnlyRootFilesystem`, drop capabilities | Security context |
| Privileged, baseline, restricted | Pod Security Standards, enforced per namespace | Security context |
pod-security.kubernetes.io/enforce: restricted | Namespace label that makes the cluster refuse pods below that level | Pod Security Admission |
trivy image app:1.2 / cosign verify | Scan for known flaws / check the signature | Images |
| Helm chart | Templates plus values.yaml | Helm |
helm install upgrade history rollback | Manage a release | Helm |
helm upgrade --install NAME CHART -f values.yaml | Install or upgrade in one command; later values override earlier ones | Value precedence |
| Kustomize | Base plus overlays; `kubectl apply -k` | Kustomize |
kubectl kustomize overlays/prod | Print what an overlay produces without applying it | kustomization.yaml |
| GitOps | Declarative, versioned, pulled, continuously reconciled | GitOps |
| Argo CD, Flux | Sync the cluster from Git; self-heal undoes manual changes | GitOps |
| prune / selfHeal | Delete what left Git / undo manual changes; both off by default | The Application |
| Sandbox, incubating, graduated | CNCF project maturity levels | The landscape |
| Pending / ImagePullBackOff / CrashLoopBackOff | Cannot place / cannot pull / keeps exiting | Finding the fault |
kubectl describe pod logs --previous get events | Events, the last crash's output, what changed | Evidence |
| Container, pod, node, cluster | Work outwards | Evidence |
| Liveness / readiness / startup probe | Restart it / send it traffic / wait for slow starts | Probes |
| Probe defaults | initialDelay 0, period 10 s, timeout 1 s, success 1, failure 3 | Probe timings |
| Upgrade order | Control plane first, one minor version at a time, then nodes | Upgrades |
kubectl drain node --ignore-daemonsets / uncordon | Empty a node / put it back | Upgrades |
kubeadm upgrade plan / apply v1.34.1 / node | Check / upgrade the first control plane node / upgrade every other node | kubeadm upgrade |
etcdctl snapshot save / restore | Back up and restore the cluster's state | Backup |
etcdutl snapshot restore FILE --data-dir=DIR | Builds a new data directory; then point the etcd static pod at it | Snapshot and restore |
Nothing matches. Try a shorter word.