Glossary¶
Status: AI draft, 2026-11-16. Not yet verified by me. Short definitions only; each term links to the page that explains it. Pages still marked "not written yet" list what they will cover.
In one line¶
One line per term, in plain words, with a link to the one page where the term is explained properly. Use it to decode a word in an interview question, then follow the link for the real answer.
A to C¶
| Term | In plain words | Explained in |
|---|---|---|
| Admission control | Code the API server runs on every create or update, which may change or reject the request | CRDs, operators and admission webhooks |
| ALB | AWS's layer-7 (HTTP) load balancer | ALB and NLB |
| AMI | A machine image that EC2 instances boot from | Auto Scaling and AMIs |
| Argo CD | A GitOps controller that keeps a cluster in sync with Git | Argo CD architecture |
| ARP | How a host finds the hardware (MAC) address for an IP on the same network | Layers and encapsulation |
| Assume role | Exchanging your identity for an IAM role's temporary credentials through STS | IAM roles and STS |
| Availability zone | One or more data centres in a region, isolated from the others | Accounts, regions and the CLI |
| Blue-green | Run the new version beside the old, then switch all traffic at once | Deployment strategies |
| Canary | Send a small share of traffic to the new version first | Deployment strategies |
| Capabilities (Linux) | Root's powers split into separate privileges that can be dropped | Pod security |
| cgroup | Kernel feature that limits and counts a group of processes' CPU and memory | What a container is |
| CIDR | Notation for an address range, such as 10.0.0.0/16 |
IP addresses and subnetting |
| ClusterIP | A Service's stable virtual address inside the cluster | Services and DNS |
| CNI | The plugin interface that gives pods their network | Pod networking and NetworkPolicy |
| Conntrack | The kernel's table of tracked connections, used by NAT and stateful firewalls | Firewalls and connection tracking |
| Container runtime | The software that actually starts containers, such as containerd or runc | Runtimes |
| CoreDNS | The cluster's DNS server | Services and DNS |
| CRD | A custom object type added to the Kubernetes API | CRDs, operators and admission webhooks |
| CronJob | Runs a Job on a schedule | Jobs and CronJobs |
D to H¶
| Term | In plain words | Explained in |
|---|---|---|
| DaemonSet | Runs one pod on every (or every matching) node | StatefulSets and DaemonSets |
| Deployment | Keeps a number of identical pods running and rolls out changes | Deployments and rollouts |
| Drift | Real infrastructure differs from what the code or Git says | Import, moved and drift |
| EBS | Block storage volumes for EC2 | EBS |
| EFS | Shared network file system for many instances or pods | EFS |
| EKS | AWS's managed Kubernetes control plane | EKS architecture and node options |
| Encapsulation | Each network layer wrapping the one above in its own header | Layers and encapsulation |
| EndpointSlice | The list of ready pod addresses behind a Service | Services and DNS |
| etcd | The key-value store that holds all Kubernetes state | etcd, backups and cluster upgrades |
| Eviction | The kubelet or API removing a pod to free resources or drain a node | Disruptions and PDBs |
| Federation | Signing in to AWS with an outside identity provider | Federation and SSO |
| Finalizer | A marker that holds an object after delete until cleanup is done | Finalizers and deletion |
| Fluent Bit | A light agent that collects and forwards logs | Fluent Bit |
| Gateway API | The successor to Ingress for routing traffic into a cluster | Ingress and Gateway API |
| GitOps | Git holds the desired state and a controller applies it | GitOps principles |
| Graceful shutdown | Finishing in-flight work after SIGTERM before exiting | Probes and graceful shutdown |
| Helm chart | A package of templated Kubernetes manifests with values | Charts, templates and values |
| HPA | Changes the number of pods from a metric | Autoscaling |
| HTTP status codes | Three-digit result of an HTTP request: 2xx success, 4xx client, 5xx server | HTTP |
I to N¶
| Term | In plain words | Explained in |
|---|---|---|
| IAM policy | A JSON document that allows or denies actions on resources | IAM policies and how they are evaluated |
| Idempotent | Running it twice has the same effect as running it once | Terraform workflow |
| Image | The read-only filesystem layers and settings a container starts from | Images and Dockerfiles |
| Ingress | Kubernetes object with host and path rules for HTTP into the cluster | Ingress and Gateway API |
| Inode | The file's metadata record on disk; a name points to it | Disks and filesystems |
| IRSA, Pod Identity | Ways for a pod's service account to get AWS credentials | Cluster access, IRSA and Pod Identity |
| Istio | A service mesh: sidecar or node proxies that handle traffic between services | Why a mesh |
| Job | Runs pods until a task completes | Jobs and CronJobs |
| JWT | A signed token carrying claims, used by OIDC and service accounts | OAuth 2.0, OIDC and JWTs |
| Karpenter | Launches right-sized nodes for pending pods | Karpenter in depth |
| kind | Kubernetes in Docker, a local test cluster | Architecture and the reconcile loop |
| kubelet | The node agent that starts pods and reports their status | Architecture and the reconcile loop |
| Label, selector | Key-value tags on objects, and the queries that pick objects by them | Labels, selectors and namespaces |
| Least privilege | Grant only the permissions a task needs | Least privilege and AccessDenied |
| Liveness probe | A check whose failure restarts the container | Probes and graceful shutdown |
| Load average | Runnable plus waiting-on-disk tasks, averaged over time | Performance triage |
| mTLS | TLS where both sides present certificates | mTLS and authorization |
| Namespace (Kubernetes) | A named group of objects for access, quotas and names | Labels, selectors and namespaces |
| Namespace (Linux) | A kernel feature that gives processes their own view of PIDs, network or mounts | What a container is |
| NACL | A stateless subnet-level firewall in a VPC | Security groups and NACLs |
| NAT | Rewriting addresses so private hosts can share public ones | Routing and NAT |
| NetworkPolicy | Kubernetes firewall rules between pods | Pod networking and NetworkPolicy |
| NLB | AWS's layer-4 (TCP, UDP, TLS) load balancer | ALB and NLB |
O to R¶
| Term | In plain words | Explained in |
|---|---|---|
| Observability | Being able to explain a system's behaviour from its metrics, logs and traces | Signals and methods |
| OIDC | An identity layer on OAuth 2.0; also how pipelines get cloud credentials without keys | OIDC from pipelines to AWS |
| OOM killer | The kernel killing a process to free memory | Memory and the OOM killer |
| Operator | A controller that manages an application through custom resources | CRDs, operators and admission webhooks |
| PDB | PodDisruptionBudget: how many pods may be down during voluntary disruptions | Disruptions and PDBs |
| Pod | One or more containers sharing a network address and volumes | Pods |
| Pod Security Admission | Built-in check that enforces privileged, baseline or restricted pods per namespace | Pod security |
| PrivateLink, VPC endpoint | Reaching AWS services or other VPCs privately, without the internet | NAT, endpoints and PrivateLink |
| PromQL | Prometheus's query language | PromQL |
| PVC, PV | A claim for storage, and the volume that satisfies it | Storage |
| QoS class | Guaranteed, Burstable or BestEffort, from requests and limits | Requests, limits and QoS |
| Quorum | The majority of members that must agree before a write counts | etcd, backups and cluster upgrades |
| RBAC | Roles listing allowed verbs, bound to users or service accounts | RBAC and service accounts |
| Readiness probe | A check whose failure removes the pod from Service endpoints | Probes and graceful shutdown |
| Rebase | Replaying commits on top of another branch | Daily workflow |
| Reconcile loop | A controller repeatedly making reality match the desired state | Architecture and the reconcile loop |
| Reflog | Git's local record of where branches pointed, used to recover lost commits | Undoing and recovering |
| Requests, limits | What a container is guaranteed, and the most it may use | Requests, limits and QoS |
| Resolver | The DNS server that looks names up on a client's behalf | DNS |
| Reverse proxy | A server that receives client requests and forwards them to backends | Load balancers and proxies |
| Route table | Rules that decide where packets for each destination go | Routing and NAT |
S to Z¶
| Term | In plain words | Explained in |
|---|---|---|
| SBOM | A list of the components inside a build or image | Registries and image security |
| SCP | An Organizations policy that caps what accounts can do | Organizations, SCPs and multi-account |
| Security group | A stateful firewall attached to network interfaces in a VPC | Security groups and NACLs |
| Service (Kubernetes) | A stable name and address in front of a changing set of pods | Services and DNS |
| Service account | An in-cluster identity that pods run as | RBAC and service accounts |
| Sidecar | A helper container running beside the main one in a pod | Sidecars and traffic flow |
| SIGTERM, SIGKILL | The polite stop signal, and the one that cannot be caught | Processes and signals |
| SLI, SLO, error budget | What you measure, the target for it, and how much failure the target allows | SLOs |
| SNI | The hostname a TLS client sends so the server picks the right certificate | TLS and certificates |
| StatefulSet | Pods with stable names and their own volumes | StatefulSets and DaemonSets |
| STS | AWS service that issues temporary credentials | IAM roles and STS |
| Subnet | A slice of a network's address range; in AWS, tied to one zone | VPC basics |
| systemd unit | A service, timer or other object that systemd manages | systemd and services |
| Taint, toleration | A node repelling pods, and a pod accepting that | Scheduling |
| TCP handshake | SYN, SYN-ACK, ACK: how a TCP connection opens | TCP and UDP |
| Terraform state | Terraform's record of which real resources it manages | State |
| TLS | Encryption and server identity for connections | TLS and certificates |
| Trace, span | One request's path across services, and one step of it | Tracing |
| Transit Gateway | A hub that connects many VPCs and networks | Peering, Transit Gateway and VPN |
| TTL | How long a DNS answer may be cached | DNS |
| VPA | Changes pods' CPU and memory requests from observed usage | Autoscaling |
| VPC | A private network in AWS | VPC basics |
| WAF | Rules that filter HTTP requests before they reach an application | CloudFront and WAF |
| Webhook (admission) | An HTTPS service the API server calls to check or change objects | CRDs, operators and admission webhooks |
| Zombie process | A finished process whose parent has not collected its exit status | Processes and signals |