etcd

What is etcd?

• AA distributed, strongly consistent key-value store — used to hold critical configuration and coordination data
• BA distributed relational database — storing tables and rows queried with SQL across many nodes
• CA distributed message queue — buffering events between producers and consumers at high throughput
• DA distributed object cache — keeping hot data in memory to speed up reads from a slower backend

How many member failures can a 5-member etcd cluster tolerate?

• ATwo — a quorum of three remains, so the cluster keeps serving writes with two members down
• BFour — as long as one member survives it can continue committing writes on its own
• COne — any second failure immediately breaks consensus even though three members remain
• DThree — a minority of two members is still enough to form a working quorum for writes

What does defragmentation do in etcd?

• AReleases the free space left after compaction — returning it from the backend file to the filesystem
• BRemoves old key revisions from history — reclaiming the space that stale versions were using
• CRebalances keys across members — evening out how much data each node in the cluster stores
• DReorders keys on disk by access frequency — putting hot keys first to speed up future reads

Which Kubernetes component talks to etcd?

• AOnly the kube-apiserver — every other component reaches cluster state through the API server
• BEvery control-plane component — the scheduler and controllers each read and write etcd directly
• CThe kubelet on each node — it stores the node's Pod status straight into etcd as it changes
• DBoth the apiserver and kube-proxy — the two share direct etcd access to sync Service endpoints