For AI agents: the complete documentation index is available at /llms.txt, the full documentation bundle is available at /llms-full.txt, and this page is available as Markdown at /docs/databases/etcd.md.
Ivory

    etcd

    An etcd cluster — the key-value store that coordinates Raft consensus among its members, and the store Patroni elects its leader through. Ivory manages it directly because so many HA setups depend on it.

    Stage: BETA

    At a glance

    • Cluster side — etcd's v3 API on the client port, used as both Keeper and Database port (2379)
    • Keeper credentials — optional; only if etcd authentication is enabled
    • Replication — Raft: one leader, the rest follow
    • Lag — Raft entries a member is behind the leader; warns past 1000

    How Ivory reads the cluster

    • Members — etcd's member list comes from whichever node is asked, so the whole cluster is visible from one. Each member keeps its etcd name; members other than the node itself are shown as reported until their own status arrives.
    • Cluster identity — each node reports its etcd cluster id, so a node from a different cluster is flagged.
    • Roles — the member etcd names as leader is the leader, the rest are replicas.
    • Lag — the gap between a member's applied raft index and the leader's.
    • Also shown — the etcd version, the raft term, and whether a member is a learner.
    • Warnings — anything the member itself reports (alarms such as NOSPACE, a lost leader), a backlog of committed but unapplied entries large enough for etcd to refuse requests, and a member or leader Ivory cannot reach, naming the address it tried.

    Operations

    • Switchover — moves leadership to the member you choose; a candidate is required.

    etcd elects its leader by itself, so there is no failover, restart or configuration to manage from Ivory.

    Query console

    An etcdctl-style console: get (with --prefix, --limit, --keys-only), put, del, member list, endpoint status, endpoint health and alarm list. Saved queries include the member list, endpoint status and health, alarms, and the keys Patroni stores for its clusters.

    Deploying

    Both templates run quay.io/coreos/etcd and leave authentication off — on purpose, so the Patroni templates can use the cluster as their coordination store.

    • Etcd (Multi Host) — three nodes etcd1–etcd3, one per machine, client port 2379 and peer port 2380. The member list in each command names etcd1–etcd3 at 10.0.0.1–10.0.0.3; replace the addresses, and keep the node names matching the list.
    • Etcd (Single Host) — all three on one machine on client ports 2479, 2481, 2483 (peers one above each). It avoids the default ports deliberately, and is what the Patroni single-host template points at.

    See Deployment.