Available inEnglishFrenchGermanHindiPortugueseRussianSpanish
CKA Exam Prep: Certified Kubernetes Administrator
Name the object, the field and the fix behind a broken cluster: ENDPOINTS none, READY 0/1, a Pod stuck Pending. For engineers who run kubectl daily with a CKA date booked; kubectl fluency is assumed, cluster administration is not. Eighty cards weighted to the CNCF curriculum's five domains and current to the Kubernetes version the exam runs; it drills recall, not the hands, so keep a lab alongside.
One box behind the API server is hidden: the only component that durably stores cluster state. Name it.
etcd
— Not a Kubernetes binary. It is the key-value store kubeadm installs first.
Source
etcd is a consistent, watchable key-value store, and it is the cluster's entire memory: lose it and no Deployment, Secret or Node exists any more. It serves clients on 2379 and peers on 2380. Backing it up is core CKA work: etcdctl snapshot save with --endpoints, --cacert, --cert and --key (the certs live under /etc/kubernetes/pki/etcd), then etcdutl snapshot restore into a fresh data directory (etcd 3.6, the version v1.35 ships, removed restore from etcdctl), after which you point etcd's static Pod manifest at it. Everything else in the control plane is stateless and can be restarted at will.