Elasticsearch

The translog's role in write durability and shard recovery before Lucene changes are safely committed.

Lucene inverted indexes, terms, postings, doc values, segments, and the distinction between search and aggregation data structures.

How Elasticsearch routes document writes and reads through shards, replicas, refresh visibility, and durability mechanisms.

Cluster nodes, indices, primary/replica shards, coordination, routing, distributed search, and recovery.

A workload-first approach to shard sizing, mappings, refresh, indexing, queries, caches, merges, and JVM/storage capacity.

Character filters, tokenizers, token filters, index/search analyzers, and why analysis is part of the search schema.

Distributed search and analytics built on Lucene: documents, indices, shards, replicas, inverted indexes, and near-real-time behavior.

Primary-replica sequencing, acknowledgment, refresh visibility, optimistic concurrency, and limits of relational-style consistency assumptions.

Reliable change propagation from a relational source of truth into a search index using CDC/outbox/replayable pipelines.

Core index, mapping, document, search, filter, aggregation, bulk, pagination, and alias practices.

Version-aware Elasticsearch installation and development setup with security, storage, memory, and cluster safeguards.

Near-real-time search visibility, refresh intervals, and why refresh is not a durability guarantee.

What a flush does to Lucene commits and the transaction log, and how it differs from refresh.

Why immutable Lucene segments are merged, how deletes are reclaimed, and the I/O/CPU trade-offs.

Optimistic concurrency using sequence numbers and primary terms, plus idempotent update and retry design.

Representative search/index benchmarks, warm-up, dataset/query realism, latency percentiles, and bottleneck measurement.

Operational trade-offs of cloud-managed Elasticsearch-compatible services: versions, networking, scaling, snapshots, observability, and lock-in.