Shard Keys

On this page本页内容

The shard key is either an indexed field or indexed compound fields that determines the distribution of the collection’s documents among the cluster’s shards.

Specifically, MongoDB divides the span of shard key values (or hashed shard key values) into non-overlapping ranges of shard key values (or hashed shard key values). Each range is associated with a chunk, and MongoDB attempts to distribute chunks evenly among the shards in the cluster.

Diagram of the shard key value space segmented into smaller ranges or chunks.

The shard key has a direct relationship to the effectiveness of chunk distribution. See Choosing a Shard Key.

Shard Key Specification

You must specify the shard key when you shard the collection. You can use the mongo shell method sh.shardCollection() to shard a collection:

sh.shardCollection(<namespace>, <key>) // Optional parameters omitted
namespace Specify the full namespace of the collection (i.e. “<database>.<collection>”) to shard.
key

Specify a document { <shard key field1>: <1|"hashed">, ... } where

Shard Key Fields and Values

Existence

Starting in version 4.4, documents in sharded collections can be missing the shard key fields. A missing shard key falls into the same range as a null-valued shard key. See Missing Shard Key.

In version 4.2 and earlier, shard key fields must exist in every document for a sharded collection.

Update Field’s Value

Starting in MongoDB 4.2, you can update a document’s shard key value unless the shard key field is the immutable _id field. Before MongoDB 4.2, a document’s shard key field value is immutable.

For details on updating the shard key, see Change a Document’s Shard Key Value.

For more information on the sharding method, see sh.shardCollection().

Refine a Shard Key

Starting in MongoDB 4.4, you can use refineCollectionShardKey to refine a collection’s shard key. The refineCollectionShardKey adds a suffix field or fields to the existing key to create the new shard key.

For example, you may have an existing orders collection with the shard key { customer_id: 1 }. You can change the shard key by adding a suffix order_id field to the shard key so that { customer_id: 1, order_id: 1 } becomes the new shard key. For more information, see the refineCollectionShardKey command.

Refining a collection’s shard key allows for a more fine-grained data distribution and can address situations where the existing key has led to jumbo (i.e. indivisible) chunks due to insufficient cardinality.

In MongoDB 4.2 and earlier, the choice of shard key cannot be changed after sharding.

Shard Key Indexes

All sharded collections must have an index that supports the shard key; i.e. the index can be an index on the shard key or a compound index where the shard key is a prefix of the index.

If you drop the last valid index for the shard key, recover by recreating an index on just the shard key.

Unique Indexes

MongoDB can enforce a uniqueness constraint on a ranged shard key index. Through the use of a unique index on the shard key, MongoDB enforces uniqueness on the entire key combination and not individual components of the shard key.

For a ranged sharded collection, only the following indexes can be unique:

  • the index on the shard key
  • a compound index where the shard key is a prefix
  • the default _id index; however, the _id index only enforces the uniqueness constraint per shard if the _id field is not the shard key or the prefix of the shard key.

    Uniqueness and the _id Index

    If the _id field is not the shard key or the prefix of the shard key, _id index only enforces the uniqueness constraint per shard and not across shards.

    For example, consider a sharded collection (with shard key {x: 1}) that spans two shards A and B. Because the _id key is not part of the shard key, the collection could have a document with _id value 1 in shard A and another document with _id value 1 in shard B.

    If the _id field is not the shard key nor the prefix of the shard key, MongoDB expects applications to enforce the uniqueness of the _id values across the shards.

The unique index constraints mean that:

  • For a to-be-sharded collection, you cannot shard the collection if the collection has other unique indexes.
  • For an already-sharded collection, you cannot create unique indexes on other fields.
  • A unique index stores a null value for a document missing the indexed field; that is a missing index field is treated as another instance of a null index key value. For more information, see Unique Index and Missing Field.

To enforce uniqueness on the shard key values, pass the unique parameter as true to the sh.shardCollection() method:

  • If the collection is empty, sh.shardCollection() creates the unique index on the shard key if such an index does not already exist.
  • If the collection is not empty, you must create the index first before using sh.shardCollection().

Although you can have a unique compound index where the shard key is a prefix, if using unique parameter, the collection must have a unique index that is on the shard key.

You cannot specify a unique constraint on a hashed index.

Choosing a Shard Key

The choice of shard key affects the creation and distribution of the chunks across the available shards. This affects the overall efficiency and performance of operations within the sharded cluster.

The ideal shard key allows MongoDB to distribute documents evenly throughout the cluster. See also sharding strategy.

At minimum, consider the consequences of the cardinality, frequency, and rate of change of a potential shard key.

Note

  • Starting in MongoDB 4.4, you can use refineCollectionShardKey to refine a collection’s shard key. The refineCollectionShardKey adds a suffix field or fields to the existing key to create the new shard key.
  • In MongoDB 4.2 and earlier, once you shard a collection, the selection of the shard key is immutable.

Restrictions

For restrictions on shard key, see Shard Key Limitations.

Collection Size

When sharding a collection that is not empty, the shard key can constrain the maximum supported collection size for the initial sharding operation only. See Sharding Existing Collection Data Size.

Important

A sharded collection can grow to any size after successful sharding.

Shard Key Cardinality

The cardinality of a shard key determines the maximum number of chunks the balancer can create. This can reduce or remove the effectiveness of horizontal scaling in the cluster.

A unique shard key value can exist on no more than a single chunk at any given time. If a shard key has a cardinality of 4, then there can be no more than 4 chunks within the sharded cluster, each storing one unique shard key value. This constrains the number of effective shards in the cluster to 4 as well - adding additional shards would not provide any benefit.

The following image illustrates a sharded cluster using the field X as the shard key. If X has low cardinality, the distribution of inserts may look similar to the following:

Diagram of poor shard key distribution due to low cardinality

The cluster in this example would not scale horizontally, as incoming writes would only route to a subset of shards.

A shard key with high cardinality does not guarantee even distribution of data across the sharded cluster, though it does better facilitate horizontal scaling. The frequency and rate of change of the shard key also contributes to data distribution. Consider each factor when choosing a shard key.

If your data model requires sharding on a key that has low cardinality, consider using a compound index using a field that has higher relative cardinality.

Shard Key Frequency

Consider a set representing the range of shard key values - the frequency of the shard key represents how often a given value occurs in the data. If the majority of documents contain only a subset of those values, then the chunks storing those documents become a bottleneck within the cluster. Furthermore, as those chunks grow, they may become indivisible chunks as they cannot be split any further. This reduces or removes the effectiveness of horizontal scaling within the cluster.

The following image illustrates a sharded cluster using the field X as the shard key. If a subset of values for X occur with high frequency, the distribution of inserts may look similar to the following:

Diagram of poor shard key distribution due to high frequency

A shard key with low frequency does not guarantee even distribution of data across the sharded cluster. The cardinality and rate of change of the shard key also contributes to data distribution. Consider each factor when choosing a shard key.

If your data model requires sharding on a key that has high frequency values, consider using a compound index using a unique or low frequency value.

Monotonically Changing Shard Keys

A shard key on a value that increases or decreases monotonically is more likely to distribute inserts to a single shard within the cluster.

This occurs because every cluster has a chunk that captures a range with an upper bound of maxKey. maxKey always compares as higher than all other values. Similarly, there is a chunk that captures a range with a lower bound of minKey. minKey always compares as lower than all other values.

If the shard key value is always increasing, all new inserts are routed to the chunk with maxKey as the upper bound. If the shard key value is always decreasing, all new inserts are routed to the chunk with minKey as the lower bound. The shard containing that chunk becomes the bottleneck for write operations.

The following image illustrates a sharded cluster using the field X as the shard key. If the values for X are monotonically increasing, the distribution of inserts may look similar to the following:

Diagram of poor shard key distribution due to monotonically increasing or decreasing shard key

If the shard key value was monotonically decreasing, then all inserts would route to Chunk A instead.

A shard key that does not change monotonically does not guarantee even distribution of data across the sharded cluster. The cardinality and frequency of the shard key also contributes to data distribution. Consider each factor when choosing a shard key.

If your data model requires sharding on a key that changes monotonically, consider using Hashed Sharding.

Change a Document’s Shard Key Value

When updating the shard key value

  • You must run on a mongos. Do not issue the operation directly on the shard.
  • You must run either in a transaction or as a retryable write.
  • You must include an equality condition on the full shard key in the query filter. For example, if a collection messages uses { activityid: 1, userid : 1 } as the shard key, to update the shard key for a document, you must include activityid: <value>, userid: <value> in the query filter. You can include additional fields in the query as appropriate.

See also the specific write command/methods for additional operation-specific requirements when run against a sharded collection.

Starting in MongoDB 4.2, you can update a document’s shard key value unless the shard key field is the immutable _id field. To update, use the following operations to update a document’s shard key value:

CommandMethod
update with multi: false
db.collection.replaceOne()
db.collection.updateOne()
db.collection.update() with multi: false
findAndModify
db.collection.findOneAndReplace()
db.collection.findOneAndUpdate()
db.collection.findAndModify()
 
db.collection.bulkWrite()
Bulk.find.updateOne()

If the shard key modification results in moving the document to another shard, you cannot specify more than one shard key modification in the bulk operation; i.e. batch size of 1.

If the shard key modification does not result in moving the document to another shard, you can specify multiple shard key modification in the bulk operation.

Warning

Starting in version 4.4, documents in sharded collections can be missing the shard key fields. Take precaution to avoid accidentally removing the shard key when changing a document’s shard key value.

Missing Shard Key

Starting in version 4.4, documents in sharded collections can be missing the shard key fields.

Chunk Range and Missing Shard Key Fields

Missing shard keys fall within the same chunk range as shard keys with null values. For example, if the shard key is on the fields { x: 1, y: 1 }, then:

Document Missing Shard KeyFalls into Same Range As
{ x: "hello" } { x: "hello", y: null }
{ y: "goodbye" } { x: null, y: "goodbye" }
{ z: "oops" } { x: null, y: null }

Read/Write Operations and Missing Shard Key Fields

To target documents with missing shard key fields, you can use the { $exists: false } filter condition on the shard key fields. For example, if the shard key is on the fields { x: 1, y: 1 }, to find the documents with missing shard key fields:

db.shardedcollection.find( { $or: [ { x: { $exists: false } }, { y: { $exists: false } } ] } )

If you specify an null equality match filter condition (e.g. { x: null }), the filter matches both those documents with missing shard key fields and those with shard key fields set to null.

Some write operations, such as a write with an upsert specification, require an equality match on the shard key. In those cases, to target a document that is missing the shard key, include another filter condition in addition to the null equality match. For example:例如：

{ _id: <value>, <shardkeyfield>: null } // _id of the document missing shard key

Set the Missing Shard Key Fields

To set missing shard key fields (which is different from changing the value of an existing shard key field), you can use the following operations on a mongos:

CommandMethodDescription描述
update with
multi: true
db.collection.update() with
multi: true
  • Can be used to set the missing key value to null only.
  • Can be performed inside or outside a transaction.
  • Can be performed as a retryable write or not.
  • For additional requirements, refer to the specific command/method.
update with
multi: false
db.collection.replaceOne()
db.collection.updateOne()
db.collection.update() with
multi: false
  • Can be used to set the missing key value to null or any other value.
  • To set to a non-null value, must be performed either inside a transaction or as a retryable write.
  • For additional requirements, refer to the specific command/method.
findAndModify
db.collection.findOneAndReplace()
db.collection.findOneAndUpdate()
db.collection.findAndModify()
  • Can be used to set the missing key value to null or any other value.
  • To set to a non-null value, must be performed either inside a transaction or as a retryable write.
  • Must include in the query filter an equality condition on the shard key. Since a missing key value is returned as part of a null equality match, to avoid updating a null-valued key, include additional query conditions as appropriate.
  • For additional requirements, refer to the specific command/method.
 
db.collection.bulkWrite()
Bulk.find.replaceOne()
Bulk.find.updateOne()
Bulk.find.update()
  • To set to a null value, you can specify multiple shard key modifications in the bulk operation.
  • To set to a non-null value, you can specify a single shard key modification in the bulk operation; i.e. batch size of 1.
  • To set to a non-null value, must be performed either inside a transaction or as a retryable write.
  • For additional requirements, refer to the underlying command/method.

Once you set the shard key field, to modify the field’s value, see Change a Document’s Shard Key Value.