Navigation

mongos

MongoDB mongos instances route queries and write operations to shards in a sharded cluster. mongos provide the only interface to a sharded cluster from the perspective of applications. Applications never connect or communicate directly with the shards.

The mongos tracks what data is on which shard by caching the metadata from the config servers. The mongos uses the metadata to route operations from applications and clients to the mongod instances. A mongos has no persistent state and consumes minimal system resources.

The most common practice is to run mongos instances on the same systems as your application servers, but you can maintain mongos instances on the shards or on other dedicated resources. See also Number of mongos and Distribution.

Routing And Results Process

A mongos instance routes a query to a cluster by:

  1. Determining the list of shards that must receive the query.
  2. Establishing a cursor on all targeted shards.

The mongos then merges the data from each of the targeted shards and returns the result document. Certain query modifiers, such as sorting, are performed on each shard before mongos retrieves the results.

Changed in version 3.6: For aggregation operations that run on multiple shards, if the operations do not require running on the database’s primary shard, these operations may route the results back to the mongos where the results are then merged.

There are two cases in which a pipeline is ineligible to run on mongos.

The first case occurs when the merge part of the split pipeline contains a stage which must run on a primary shard. For instance, if $lookup requires access to an unsharded collection in the same database as the sharded collection on which the aggregation is running, the merge is obliged to run on the primary shard.

The second case occurs when the merge part of the split pipeline contains a stage which may write temporary data to disk, such as $group, and the client has specified allowDiskUse:true. In this case, assuming that there are no other stages in the merge pipeline which require the primary shard, the merge will run on a randomly-selected shard in the set of shards targeted by the aggregation.

For more information on how the work of aggregation is split among components of a sharded cluster query, use explain:true as a parameter to the aggregation() call. The return will include three json objects. mergeType shows where the stage of the merge happens (“primaryShard”, “anyShard”, or “mongos”). splitPipeline shows which operations in your pipeline have run on individual shards. shards shows the work each shard has done.

In some cases, when the shard key or a prefix of the shard key is a part of the query, the mongos performs a targeted operation, routing queries to a subset of shards in the cluster.

mongos performs a broadcast operation for queries that do not include the shard key, routing queries to all shards in the cluster. Some queries that do include the shard key may still result in a broadcast operation depending on the distribution of data in the cluster and the selectivity of the query.

See Targeted Operations vs. Broadcast Operations for more on targeted and broadcast operations.

Starting in MongoDB 4.4, mongos can support hedge reads to minimize latencies. See hedge reads for more information.

How mongos Handles Query Modifiers

Sorting

If the result of the query is not sorted, the mongos instance opens a result cursor that “round robins” results from all cursors on the shards.

Limits

If the query limits the size of the result set using the limit() cursor method, the mongos instance passes that limit to the shards and then re-applies the limit to the result before returning the result to the client.

Skips

If the query specifies a number of records to skip using the skip() cursor method, the mongos cannot pass the skip to the shards, but rather retrieves unskipped results from the shards and skips the appropriate number of documents when assembling the complete result.

When used in conjunction with a limit(), the mongos will pass the limit plus the value of the skip() to the shards to improve the efficiency of these operations.

Read Preference and Shards

For sharded clusters, mongos applies the read preference when reading from the shards. The member selected is governed by both the read preference and replication.localPingThresholdMs settings, and is re-evaluated for each operation.

For details on read preference and sharded clusters, see Read Preference and Shards.

Hedged Reads

Starting in version 4.4, mongos instances can hedge reads that use non-primary read preferences. With hedged reads, the mongos instances route read operations to two replica set members per each queried shard and return results from the first respondent per shard. The additional read sent to hedge the read operation uses the maxTimeMS value of maxTimeMSForHedgedReads.

Hedged reads are supported for the following operations:

Hedged Reads and Read Preference

Hedged reads are specified per operation as part of the read preference. Non-primary read preferences support hedged reads. See Hedged Read Preference Option.

For details on read preference and sharded clusters as well as member selection, see Read Preference and Shards.

Enable/Disable Support for Hedged Reads

By default, mongos instances support using hedged reads. To turn off a mongos instance’s support for hedged reads, see the readHedgingMode parameter. If the hedged read support is off, mongos does not use hedged reads regardless of the hedge option specified for the read preference.

Hedged Reads Diagnostics

The command serverStatus and its corresponding mongo shell method db.serverStatus() return hedgingMetrics.

Confirm Connection to mongos Instances

To detect if the MongoDB instance that your client is connected to is mongos, use the isMaster command. When a client connects to a mongos, isMaster returns a document with a msg field that holds the string isdbgrid. For example:

{
   "ismaster" : true,
   "msg" : "isdbgrid",
   "maxBsonObjectSize" : 16777216,
   "ok" : 1,
   ...
}

If the application is instead connected to a mongod, the returned document does not include the isdbgrid string.

Targeted Operations vs. Broadcast Operations

Generally, the fastest queries in a sharded environment are those that mongos route to a single shard, using the shard key and the cluster meta data from the config server. These targeted operations use the shard key value to locate the shard or subset of shards that satisfy the query document.

For queries that don’t include the shard key, mongos must query all shards, wait for their responses and then return the result to the application. These “scatter/gather” queries can be long running operations.

Broadcast Operations

mongos instances broadcast queries to all shards for the collection unless the mongos can determine which shard or subset of shards stores this data.

Read operations to a sharded cluster. Query criteria does not include the shard key. The query router ``mongos`` must broadcast query to all shards for the collection.

After the mongos receives responses from all shards, it merges the data and returns the result document. The performance of a broadcast operation depends on the overall load of the cluster, as well as variables like network latency, individual shard load, and number of documents returned per shard. Whenever possible, favor operations that result in targeted operation over those that result in a broadcast operation.

Multi-update operations are always broadcast operations.

The updateMany() and deleteMany() methods are broadcast operations, unless the query document specifies the shard key in full.

Targeted Operations

mongos can route queries that include the shard key or the prefix of a compound shard key a specific shard or set of shards. mongos uses the shard key value to locate the chunk whose range includes the shard key value and directs the query at the shard containing that chunk.

Read operations to a sharded cluster. Query criteria includes the shard key. The query router ``mongos`` can target the query to the appropriate shard or shards.

For example, if the shard key is:

{ a: 1, b: 1, c: 1 }

The mongos program can route queries that include the full shard key or either of the following shard key prefixes at a specific shard or set of shards:

{ a: 1 }
{ a: 1, b: 1 }

All insertOne() operations target to one shard. Each document in the insertMany() array targets to a single shard, but there is no guarantee all documents in the array insert into a single shard.

All updateOne(), replaceOne() and deleteOne() operations must include the shard key or _id in the query document. MongoDB returns an error if these methods are used without the shard key or _id.

Depending on the distribution of data in the cluster and the selectivity of the query, mongos may still perform a broadcast operation to fulfill these queries.

Index Use

If the query does not include the shard key, the mongos must send the query to all shards as a “scatter/gather” operation. Each shard will, in turn, use either the shard key index or another more efficient index to fulfill the query.

If the query includes multiple sub-expressions that reference the fields indexed by the shard key and the secondary index, the mongos can route the queries to a specific shard and the shard will use the index that will allow it to fulfill most efficiently.

Sharded Cluster Security

Use Internal/Membership Authentication to enforce intra-cluster security and prevent unauthorized cluster components from accessing the cluster. You must start each mongod or mongos in the cluster with the appropriate security settings in order to enforce internal authentication.

See Deploy Sharded Cluster with Keyfile Authentication for a tutorial on deploying a secured sharded cluster.

Cluster Users

Sharded clusters support Role-Based Access Control (RBAC) for restricting unauthorized access to cluster data and operations. You must start each mongod in the cluster, including the config servers, with the --auth option in order to enforce RBAC. Alternatively, enforcing Internal/Membership Authentication for inter-cluster security also enables user access controls via RBAC.

With RBAC enforced, clients must specify a --username, --password, and --authenticationDatabase when connecting to the mongos in order to access cluster resources.

Each cluster has its own cluster users. These users cannot be used to access individual shards.

See Enable Access Control for a tutorial on enabling adding users to an RBAC-enabled MongoDB deployment.

Metadata Operations

mongos uses "majority" write concern for the following operations that affect the sharded cluster metadata:

Command Method Note
addShard sh.addShard()  
create db.createCollection()  
drop db.collection.drop()  
dropDatabase db.dropDatabase() Changed in MongoDB 3.6
enableSharding sh.enableSharding()  
movePrimary    
renameCollection db.collection.renameCollection()  
shardCollection sh.shardCollection()  
removeShard    
setFeatureCompatibilityVersion    

Additional Information

FCV Compatibility

Starting in MongoDB 4.0, the mongos binary will crash when attempting to connect to mongod instances whose feature compatibility version (fCV) is greater than that of the mongos. For example, you cannot connect a MongoDB 4.0 version mongos to a 4.2 sharded cluster with fCV set to 4.2. You can, however, connect a MongoDB 4.0 version mongos to a 4.2 sharded cluster with fCV set to 4.0.

Connection Pools

Starting in MongoDB 4.2, MongoDB adds the parameter ShardingTaskExecutorPoolReplicaSetMatching that determines the minimum size (can vary during runtime) of the mongos instance’s connection pools to each member of the sharded cluster.

By default, for each replica set in the sharded cluster (i.e. shard replica set and config servers), mongos maintains connection pools to each replica set secondary that are at least equal to the size of its connection pool to the primary.

To modify, see ShardingTaskExecutorPoolReplicaSetMatching.

←   Config Servers Shard Keys  →