Release Notes for MongoDB 8.2

terminateSecondaryReadsOnOrphanCleanup Parameter

MongoDB 8.2 adds the terminateSecondaryReadsOnOrphanCleanup parameter, which controls whether long-running read operations on secondary nodes automatically terminate before orphaned document deletion following a chunk migration.

By default, this parameter is set to true. If a read operation on a secondary node begins before the chunk migration commits, MongoDB automatically terminates the operation before deleting the orphaned documents. In MongoDB versions before 8.2, these operations continue executing after orphaned document deletion and may silently miss documents without returning an error.

To support this new behavior, the default value for orphanCleanupDelaySecs is now 3600 (previously 900).

Index Builds During Initial Sync

Starting in MongoDB 8.2, index builds during initial sync use 10% of available RAM by default. You can specify a different percentage of memory for MongoDB to use with the initialSyncIndexBuildMemoryPercentage parameter.

By default, the amount of available memory must be between 200 MB and 16 GB, inclusive. You use the following parameters to specify different minimum and maximum thresholds for memory usage:

• initialSyncIndexBuildMemoryMinMB

• initialSyncIndexBuildMemoryMaxMB

Query Statistics

Starting in MongoDB 8.2, the $queryStats output includes the following metrics on delinquent tickets:

• metrics.delinquentAcquisitions

• metrics.totalAcquisitionDelinquencyMillis

• metrics.maxAcquisitionDelinquencyMillis

MongoDB considers an execution ticket as delinquent when acquisition takes too long.

WiredTiger Cache Size Configuration as a Percentage

Starting in MongoDB 8.2, you can configure the WiredTiger internal cache size using a new percentage-based option. In addition to the existing --wiredTigerCacheSizeGB and storage.wiredTiger.engineConfig.cacheSizeGB gigabyte-based settings, you can now specify the cache size as a percentage of the available memory with the --wiredTigerCacheSizePct and storage.wiredTiger.engineConfig.cacheSizePct options.

New Inconsistency Type: Range Deletion Missing Shard Key Index

Starting in MonogDB 8.2, a new inconsistency type RangeDeletionMissingShardKeyIndex is implemented. The inconsistency type indicates that a sharded collection exists that doesn't have an index compatible with the collection shard key and there is at least one remaining range deletion task to complete.

serverStatus Output Change

serverStatus includes the following new fields in its output:

• metrics.operation.killedDueToRangeDeletion

• metrics.query.bucketAuto.spilledBytes

• metrics.query.bucketAuto.spilledDataStorageSize

• metrics.query.bucketAuto.spilledRecords

• metrics.query.bucketAuto.spills

• planCache.classic.replanned_plan_is_cached_plan

• planCache.sbe.replanned_plan_is_cached_plan

Explain Output for Disk Spilling Stages

Starting in MongoDB 8.2, the Explain Results for stages that spill to disk now include standardized metrics under consistent field names. This update provides a more unified way to track disk usage for memory-intensive operations.

The new fields include:

• spills

• spilledBytes

• spilledRecords

• spilledDataStorageSize

Targeted Mirrored Reads

Starting in MongoDB 8.2, you can selectively mirror read operations to specific servers that need their caches warmed up by tagging the nodes for read mirroring. Unlike general mirrored reads, targeted read mirroring allows you to target hidden nodes and mirror from both primary and secondary nodes.

You can configure targeted mirrored reads using the targetedMirroring field in the mirrorReads parameter.