Compatibility Changes in MongoDB 8.0

Queries for null Don't Match undefined Fields

Starting in MongoDB 8.0, comparisons to null in equality match expressions don't match undefined values.

For example, consider these documents and query:

// create the people collection
db.people.insertMany( [
 { _id: 1, name: null },
 { _id: 2, name: undefined },
 { _id: 3, name: [ "Gabriel", undefined ] },
 { _id: 4, names: [ "Alice", "Charu" ] }
] )

db.people.find( { name: null } )

Prior to MongoDB 8.0, the preceding query would match documents where:

• The name field is null (_id: 1)

• The name field is undefined or contains an undefined array element (_id: 2 and _id: 3)

• The name field does not exist (_id: 4)

Starting in MongoDB 8.0, the preceding query does not match documents where the name field is undefined or contains undefined array elements. The query only matches documents where:

• The name field is null or contains a null array element (_id: 1)

• The name field does not exist (_id: 4)

This query behavior change also impacts these operations:

• $eq

• $in

• $lookup, because a null local field no longer matches an undefined foreign field.

To learn how to rewrite your queries or migrate your data to account for this behavior change, see Migrate Undefined Data and Queries.

Cannot Connect Directly to Shard and Run Commands

Starting in MongoDB 8.0, you can only run certain commands on shards. If you attempt to connect directly to a shard and run an unsupported command, MongoDB returns an error:

"You are connecting to a sharded cluster improperly by connecting directly
to a shard. Please connect to the cluster via a router (mongos)."

To run a non-supported database command directly against a shard, you must either connect to mongos or have the maintenance-only directShardOperations role.

MongoDB supports online transition from a replica set to a 1-shard cluster by allowing commands to be run directly against a shard. However, once the cluster has more than one shard, only the listed commands can be run directly against the shard without the maintenance-only directShardOperations role.

Write Concern Majority

Starting in MongoDB 8.0, write operations that use the "majority" write concern return an acknowledgment when the majority of replica set members have written the oplog entry for the change. This improves the performance of "majority" writes. In previous releases, these operations would wait and return an acknowledgment after the majority of replica set members applied the change.

Oplog Buffers

Starting in MongoDB 8.0, secondaries write and apply oplog entries for each batch in parallel. This introduces a breaking change for the metrics.repl.buffer status metric, as it now provides information on two buffers instead of one.

MongoDB 8.0 deprecates the following server status metrics:

• metrics.repl.buffer.count

• metrics.repl.buffer.maxSizeBytes

• metrics.repl.buffer.sizeBytes

It replaces them with these metrics:

• metrics.repl.buffer.apply.count

• metrics.repl.buffer.apply.maxSizeBytes

• metrics.repl.buffer.apply.sizeBytes

• metrics.repl.buffer.write.count

• metrics.repl.buffer.write.maxSizeBytes

• metrics.repl.buffer.write.sizeBytes

Shutdown Performance

Starting in MongoDB 8.0, Bulk.insert() and data ingestion workloads may perform better. However, shutting down MongoDB immediately after running these workloads might take longer because of the extra data being flushed to disk.

Concurrent Compact Commands Not Allowed

Starting in MongoDB 8.0, if you try to run multiple concurrent compact commands on the same collection, MongoDB returns an error.

Geospatial Input Validation

Starting in MongoDB 8.0, you cannot use geospatial queries with malformed input. In prior versions, certain geospatial queries accept malformed input without an error.

OIDC Identity Providers Can Share an Issuer

Starting in MongoDB 8.0, when multiple identity providers (IDP) are defined, the oidcIdentityProviders parameter accepts duplicate issuer values as long as the audience value is unique for each issuer. This is also available in versions 7.3 and 7.0.

Removed Parameters

MongoDB 8.0 removes the storeFindAndModifyImagesInSideCollection parameter.

serverStatus Output Change

Starting in MongoDB 8.0, wiredTiger.concurrentTransactions is renamed to queues.execution.

Time Series system.buckets Collections

When you upgrade to 8.2, if you have any system.buckets collections that are not time-series collections, you might need to drop or rename those collections before you upgrade, depending on your 8.0 patch release:

MongoDB 8.0.5 and later

You do not need to drop system.buckets collections that aren't time-series collections before you upgrade. However, you must drop or rename them after you complete your upgrade.

MongoDB 8.0.4 and earlier

You must drop or rename system.buckets collections that aren't time-series collections before you upgrade. All system.buckets collections must have valid time series options configured before you upgrade to versions 8.0.0 - 8.0.4.

To determine whether you have system.buckets collections that are not time-series collections, use the db.getCollectionInfos() method with a filter:

db.getCollectionInfos(
 {
 $and: [
 { name: { $regex: /^system\.buckets/ } },
 { 'options.timeseries': { $exists: false } }
 ]
 }
)

Sorting by Non-Existent Fields

When sorting by a nonexistent field, MongoDB does not guarantee any particular output ordering. The behavior in these cases may change from version to version.

$rank and $denseRank Behavior

Starting in MongoDB 8.0, null and missing field values in $denseRank and $rank sortBy operations are treated the same when calculating rankings. This change makes the behavior of denseRank and rank consistent with $sort.

$shardedDataDistribution Output

Starting in MongoDB 8.0, $shardedDataDistribution only returns output for a collection's primary shard if the primary shard has chunks or orphaned documents.

For details, see $shardedDataDistribution.

Upgraded TCMalloc

Starting in MongoDB 8.0, MongoDB uses an upgraded version of TCMalloc that uses per-CPU caches, instead of per-thread caches, to reduce memory fragmentation and make your database more resilient to high-stress workloads.

To use the new TCMalloc with better performance, see TCMalloc Performance Optimization for a Self-Managed Deployment.