Operate
- Start with query patterns, relevance requirements, data freshness, and retention.
- Treat mapping, analyzer, shard count, and indexing strategy as architectural choices.
- Optimize for predictable operations, not demo-search quality only.
- Distinguish logging/observability use-cases from user-facing search needs.
Default Standards
- Mappings should be explicit on important fields.
- Shards are not free; oversharding is an operational tax.
- Relevance tuning requires representative queries and evaluation, not intuition.
- Observe indexing throughput, search latency, heap pressure, GC, and shard health.
References
- Indexing and query design: references/indexing-and-query-design.md
- Operations and relevance: references/operations-and-relevance.md
- Mappings and analyzers: references/mappings-and-analyzers.md
- Query DSL and scoring: references/query-dsl-and-scoring.md
- Shards and cluster layout: references/shards-and-cluster-layout.md
- Indexing pipelines: references/indexing-pipelines.md
- Search relevance: references/search-relevance.md
- Aggregations and analytics: references/aggregations-and-analytics.md
- ILM and retention: references/ilm-and-retention.md
- Capacity planning: references/capacity-planning.md
- Security and multi-tenant: references/security-and-multi-tenant.md
- Backup, restore, and snapshots: references/backup-restore-and-snapshots.md
- Observability: references/observability.md
- Logging vs search clusters: references/logging-vs-search-clusters.md
- Incident runbooks: references/incident-runbooks.md