Consul Ops Pro key terms

The agreement protocol Consul servers use to replicate the state of the cluster; a majority of voting servers must commit an entry before it is applied, which is why odd-numbered server counts are recommended.

A node that participates in the Raft cluster, stores the catalog, KV, and ACL state, and answers queries; agents forward requests to servers so clients never talk to the data store directly.

A lightweight agent on every node that runs local health checks, registers services, and forwards RPCs to servers; it does not hold Raft state and can be replaced without affecting cluster quorum.

The minimum number of voting servers, a strict majority, that must be available for the cluster to commit writes and elect a leader; losing quorum stops all catalog and KV mutations until servers return.

The Raft process by which servers choose one node to accept writes and replicate the log; if the leader fails, a new one is elected only when a majority of servers can communicate.

A periodic reconciliation in which a client agent compares its local service and check registrations with the catalog on the servers and re-syncs any differences, healing drift after network partitions.

The gossip layer embedded in Consul that handles membership, failure detection, and event broadcast over LAN and WAN pools; it is separate from Raft and uses its own encryption key.

The Serf membership group formed by all agents in one datacenter, used for fast failure detection and event propagation; it is distinct from the WAN pool that joins servers across datacenters.

The Serf group that only Consul servers join, spanning all datacenters in a federation; it carries cross-datacenter membership and is how servers learn about peers in remote datacenters.

Joining multiple Consul datacenters so a query in one can be answered from another, typically using prepared queries with a nearest or failover policy; each datacenter keeps its own servers and Raft state.

A server configuration that delays Raft bootstrap until the expected number of servers have joined, preventing an accidental single-node cluster from becoming the permanent source of truth.

The procedure for restoring a cluster that has lost quorum, usually by using the peers.json file to tell remaining servers which nodes should form the new configuration, then restarting them together.

The consul snapshot save command, which writes a point-in-time archive of the Raft state, including KV, catalog, ACLs, and intentions, used as the primary backup artifact for disaster recovery.

The consul snapshot restore command, which replaces the cluster state with a saved archive; it must be run against a server with a token that has operator write permission and is typically used before rejoining nodes.

An Enterprise feature that runs alongside Consul servers to take scheduled snapshots and optionally ship them to S3, Azure Blob, or GCS, providing automated backups without external cron jobs.

The act of declaring a service instance, its address, port, tags, and metadata to the local agent, either in configuration or via the HTTP API, so it appears in the catalog and can be discovered.

The JSON or HCL block describing a service, including name, ID, address, port, tags, meta, checks, and Connect sidecar settings; it is loaded by the agent and synced to the catalog.

A test the agent runs against a service or node, such as script, HTTP, TCP, TTL, or gRPC, whose status is reported to servers and used to filter discovery results and drive intention enforcement.

A health check whose status is set by the service itself calling the agent's check-update endpoint; if the service stops reporting within the timeout, the check turns critical automatically.

A service definition field that tells Consul to remove an instance from the catalog once its check has been critical for the given duration, cleaning up after crashed workloads that never deregister.

Consul's service mesh feature that provides mutual TLS identities to services and routes traffic through sidecar or native proxies, so services authenticate each other by certificate rather than network location.

A per-instance proxy, usually Envoy, registered alongside a service that terminates Connect mTLS, enforces intentions, and forwards traffic to the local application over plaintext on loopback.

A dependency a Connect-enabled service declares so its sidecar proxy can establish mTLS connections to that service's proxies, including the local listener address the application should call.

Authorization rules that specify which source services may connect to a destination service over the mesh, matched by service name, namespace, or partition and evaluated before mTLS traffic is allowed.

A secret credential presented with requests that grants the permissions of the policies and roles attached to it; tokens can be human, service, or node identities and may carry expiration or locality.

A named set of rules granting read, write, or deny access to resources such as services, nodes, keys, or operator endpoints; policies are attached to tokens or roles to grant permissions.

The one-time initialization that creates the initial management token when ACLs are enabled; it must be captured immediately and stored securely, because it cannot be retrieved later.

The default token used for requests that present no credential; its policy determines what unauthenticated callers can see, so it is usually restricted to node read or nothing at all.

A collection of policies and service identities that can be attached to many tokens, simplifying rotation and management when a group of workloads needs the same permissions.

An Enterprise ACL construct that grants a token permission to act as a specific service, including reading its Connect intentions and obtaining its mesh certificate, scoped to namespace and partition.

Symmetric encryption of Serf traffic using a shared key configured with encrypt, protecting membership and event data; the key must match across all agents and can be rotated with keyring commands.

Mutual TLS in which both sides present certificates; Consul uses it for server RPC, HTTPS API, and Connect mesh traffic, with separate certificate authorities and verify settings for each path.

The short-lived certificate issued by Consul's built-in CA to a service instance, encoding its identity and used by the sidecar proxy to prove who it is during Connect mTLS handshakes.

Consul's ability to use HashiCorp Vault as the Connect certificate authority and to store gossip and TLS keys there, centralizing secret management and enabling automatic rotation.

A query mode that lets any server answer from its local state without contacting the leader, trading freshness for lower latency and higher availability during leader elections.

The metrics Consul emits, such as Raft apply time, Serf member counts, and request rates, exposed in StatsD, Prometheus, or DogStatsD formats for dashboards and alerting.

An Enterprise feature that records every authenticated API request, including the token used and the resource touched, to a file or sink for compliance and forensic investigation.

A stored, named query template that can add failover across datacenters, nearest-node sorting, and automatic service failover, letting clients call a stable endpoint instead of raw catalog lookups.

The commercially licensed build adding namespaces, admin partitions, redundancy zones, network areas, audit logging, and the snapshot agent, features not available in the open-source binary.

An Enterprise scoping mechanism that isolates services, KV entries, and intentions within a datacenter, allowing multiple teams or environments to share servers without name collisions.

An Enterprise construct that divides one datacenter into separate administrative domains, each with its own catalog, KV, and ACL scope, often used to isolate teams or tenants on shared servers.