Glossary

Project-specific terms used throughout the Elasticsearch codebase. Standard search-engine terminology (term, document, postings list, etc.) is omitted; consult the Lucene docs for those.

| Term | Meaning | | -------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------- | | Action | A request/response pair routed through ActionModule. Every cluster operation has an ActionType and one or more handlers (REST and transport). | | Allocation | The process of deciding which node hosts which shard. Driven by AllocationService and the desired-balance allocator. | | Aggregation | A faceted analytics computation (server/src/main/java/org/elasticsearch/search/aggregations). Bucket aggregations group documents; metric aggregations compute values. | | BWC | Backwards compatibility. Tested via qa/full-cluster-restart, qa/rolling-upgrade, and the :bwc Gradle subprojects. | | CCR | Cross-Cluster Replication (X-Pack). Replicates indices from a leader cluster to a follower. | | CCS | Cross-Cluster Search. Federates a search across multiple remote clusters. | | Cluster state | The strongly-consistent metadata document maintained by the master and broadcast to followers. See ClusterState. | | Coordinator | Two meanings: (a) the node that receives and orchestrates a request (the coordinating node); (b) the master-election service Coordinator (cluster/coordination/). | | Data stream | A logical write target backed by a rolling sequence of hidden indices. Used for time-series data. | | Engine | The per-shard wrapper around Lucene (InternalEngine, ReadOnlyEngine, FrozenEngine). | | **ESQL / ES | QL** | Elasticsearch Query Language — a piped query DSL implemented in x-pack/plugin/esql/. | | Feature flag | A boolean gate registered via FeatureService or per-plugin feature classes (e.g. EsqlCapabilities). Used for runtime feature negotiation across mixed-version clusters. | | FieldData / Doc values | Columnar in-memory or on-disk per-document field representation used for sorting and aggregations. | | Frozen tier | Storage tier where indices are kept on cheap object storage and accessed via searchable snapshots with a small disk cache. | | ILM | Index Lifecycle Management. Policies that move indices through hot/warm/cold/frozen/delete phases. | | Index | Logical collection of documents, split into shards. Distinct from a Lucene index, which is a single shard. | | Ingest pipeline | A chain of processors applied before indexing (server/src/main/java/org/elasticsearch/ingest). | | Master-eligible | A node role that can be elected master. | | Module | Code under modules/ — built-in plugin shipped with the default distribution and always loaded. | | Node | A single Elasticsearch process; instance of org.elasticsearch.node.Node. | | Painless | The default scripting language (modules/lang-painless/). | | Persistent task | Long-running task tracked in cluster state and recovered after restart (e.g. ML jobs, transforms). | | Plugin | Optional code extending the node, packaged separately (plugins/) or shipped as X-Pack. | | Primary / Replica | Two roles a shard can play. Writes go to the primary first, then are replicated to replicas. | | Refresh | Make recently indexed documents visible to search by reopening the Lucene reader. | | Reroute | A pass through AllocationService that re-evaluates shard placement. | | Repository | An external blob store used for snapshots (FS, S3, GCS, Azure, HDFS, URL). | | REST handler | Class implementing RestHandler that registers HTTP routes and translates requests to actions. | | Searchable snapshot | An index whose shards are backed directly by snapshot blobs in a repository, with optional local caching. | | Segment | A self-contained Lucene sub-index. Shards consist of multiple segments; merging combines them. | | Shard | One Lucene index file set; either primary or replica. Shard count per index is fixed at creation. | | Slow log | Per-shard log that records queries / index operations that exceed configured thresholds. | | Snapshot | Point-in-time backup of one or more indices stored in a repository. | | SLM | Snapshot Lifecycle Management — schedules and retains snapshots automatically. | | Stateless | Topology where the durable copy of an index lives in object storage and nodes are caches. See x-pack/plugin/stateless. | | Threading | Most actions run on named thread pools (generic, search, write, refresh, management, flush, force_merge, ...). | | TransportAction | Server-side handler for an action (Transport*Action). | | Transport version | Wire-format version negotiated between nodes; gates new fields in serialized messages. | | Translog | Per-shard write-ahead log that backs indexing durability between flushes. | | Watcher | X-Pack alerting feature that runs scheduled queries and triggers notifications. | | X-Pack | Commercial features (subject to the Elastic License). All under x-pack/. |