MQL reference navigation
Sharding commands
Sharding behavior and option support depend on the installed implementation.
shardCollection and reshardCollection
sh.shardCollection("bookings.flights", { route_no: "hashed" })
sh.reshardCollection("bookings.flights", { departure_airport: "hashed" })| Parameter | Description |
|---|---|
shardCollection | (Required) A string specifying the fully qualified namespace, database.collection. |
reshardCollection | (Required) A string specifying the fully qualified namespace, database.collection. |
key | (Required) A document specifying the shard-key specification; the backend validates supported hashed shard-key forms. |
numInitialChunks | (Optional, implementation-dependent) A positive integer requesting the initial number of chunks. Support and the maximum permitted count depend on the deployment's sharding implementation. |
On an implementation that supports numInitialChunks, either shell helper can pass the option:
sh.shardCollection("bookings.flights", { route_no: "hashed" }, { numInitialChunks: 4 })
sh.reshardCollection("bookings.flights", { departure_airport: "hashed" }, { numInitialChunks: 4 })Support for numInitialChunks depends on the installed sharding implementation, regardless of which client helper submits the request.
unshardCollection
db.adminCommand({ unshardCollection: "bookings.flights" })Use a raw request to send only the namespace. The shell's sh.unshardCollection() helper requires an additional toShard argument.
| Parameter | Description |
|---|---|
unshardCollection | (Required) A string specifying the fully qualified namespace. |
getShardMap
db.adminCommand({ getShardMap: 1 })Returns deployment-provided shard metadata. The default wire adapter passes no command-specific options to its backend query.
Other sharding commands
enableSharding is a no-op returning success; it does not change database state.
balancerStart, balancerStatus, balancerStop, listShards, and moveCollection can be implemented by some providers.
See wire protocol and common fields for the full dispatch support inventory.
Implementation
The backend command-document entry points parse numInitialChunks for both shardCollection and reshardCollection and apply it when creating the distributed collection. The count must be positive and no greater than the configured sharding limit.
The default open-source gateway adapter uses a legacy backend call that passes only the namespace, key, and whether the request is a reshard. It does not forward numInitialChunks; deployments must use an implementation that forwards the full command document to honor this option. That adapter also does not forward unique, presplitHashedZones, collation, forceRedistribution, or shardDistribution.
Source: gateway processor/data_description.rs, postgres/documentdb_data_client.rs, and backend pg_documentdb/src/sharding/sharding.c, including ParseShardCollectionRequest, ParseReshardCollectionRequest, and ParseUnshardCollectionRequest.