Indexing Strategies — Azure DocumentDB

Companion skill to documentdb-query-optimizer. That skill answers "why is this query slow?"; this one answers "which index should I create, and what shape should it take?".

Azure DocumentDB supports the standard MongoDB index types. Only _id is created automatically — every other index must be created explicitly. Default limit: 64 single-field indexes per collection (extendable to 300 on request).

Index engine types

_id vs regular indexes: The _id index is a B-tree, created automatically, and cannot be dropped. For sharded collections the _id key is composite — it includes a hash of the shard key. All other indexes created via createIndex are RUM indexes; the exception is geospatial indexes (2dsphere, 2d), which are GiST indexes.

Rules

• index-single-field — When a single-field index is enough; direction, options (unique, sparse, partial, collation).
• index-compound-esr — Compound index design via ESR (Equality → Sort → Range); prefer one compound over many singles.
• index-multikey-arrays — Indexing array fields; the one-array-per-compound (parallel-array) restriction; multikey can't cover queries.
• index-text-prefer-textsearch — On Azure DocumentDB, prefer the textSearch index + $search over community $text indexes.
• index-wildcard-dynamic-schemas — Wildcard indexes for truly dynamic schemas; cost vs benefit; scope the prefix.
• index-hashed-shard-keys — Hashed indexes for even distribution; when sharding becomes necessary; shard-key alignment; range-query caveats.
• index-2dsphere-geospatial — GeoJSON types, [longitude, latitude] order, $near / $geoWithin / $geoIntersects.
• index-ttl-expiry — TTL indexes: expireAfterSeconds semantics, date-field requirement, monitoring.
• index-count-budget — Keep 5–15 indexes per collection; review $indexStats; drop unused.
• index-lifecycle-drop-hide — Safe lifecycle: inventory → detect redundancy → hideIndex → dropIndex. The _id index cannot be dropped.
• index-pattern-cookbook — Query-pattern → index-shape cookbook (equality+sort, multi-equality, selectivity ordering, $eq + $in mix, range+sort, equality+range, hybrid, partial index for large arrays).