MongoDB 8.0.5

What's new in this version:

Sharding:
- Remove collection name limit of 235 for unsplittable collections
- Do not force a Shard Registry lookup if topology time doesn't change
- Add test coverage for checkMetadataConsistency interruptability
- setFeatureCompatibilityVersion, createCollection and moveCollection could cause a 3-way deadlock in config shards
- Prevent sharding/moving a time-series buckets collection without the 'timeseries' options on FCV 8.0+

Replication:
- dbCheck oplog entries contain duplicate field names in batchStart/batchEnd

Query:
- $unionWith stage for explain does not use the router role API
- Indexes on array fields that contain subarrays does not include some results
- findAndModify: Mongos WriteConcernError Behavior Differs from Mongod
- Replace invariant with tassert when yielding
- Express path can return incorrect results when scanning a unique, multi-field index

Catalog:
- [Test only] collection_catalog_two_phase_drops.js assumes write with journaling will advance the latest timestamp
- Prevent config.system.sessions from being created as sharded with its chunk on the config server
- Invalid timeseries buckets collections prevent completion of FCV 8.0 upgrade

Storage:
- WiredTiger
- test_wt4803_history_store_abort hangs
- Investigate checkpoint cleanup for deleted pages
- Fix how compact walk handles EBUSY from checkpoint flush_lock
- Remove bounded cursor support when positioning truncate cursors
- Background compaction does not reset stats with run_once
- Migrate to AWS-based MacOS hosts
- Use internal session to create the history store table instead of default session
- Provide an option to control the number of RTS worker threads during startup/shutdown
- Create separate statistics for dirty leaf/internal pages
- Add WT stat to count create-with-import with repair
- Set up code ownership to guard backport PRs
- Build and Packaging:
- Configure folders created by installer
- Include collection type with create events.
- Introduce "matchCollectionUUIDForUpdateLookup" parameter in the change stream stage
- Remove swig installation from configure_wiredtiger

Internals:
- Add time spent in the multi-planner to the explain output
- Abstract out similar code between data consistency/extra index keys check and make error handling consistent
- Test concurrent build/drop of indexes during dbCheck
- Enable disabled test because of dbcheck hook in replica sets passthrough suites
- Add dropped collection/index and newly created view tests for extra index keys check
- Handle writeConcernErrors for FLE in bulkWrite
- Enqueued refreshes under the same inProgressLookup entry are always called with the same cached value in the ReadThroughCache
- GDB itself segfaults when index cache is enabled while loading core dump on SUSE 12 and SUSE 15
- Add magic restore log path parameter and generate magic restore-specific logs
- Exceptions thrown while generating index keys fail collection validation
- Log value of field for missing key check in dbCheck
- Change stream tests do not recover from resumable change stream errors
- Add dbcheck parameters to health log entry output
- Create metrics related to restore tool usage
- Allow magic restore to insert oplog entries larger than 16MB
- make burn_in_tests.py understand multiversion configuration
- Investigate Windows-specific failures for magic restore named pipe infrastructure
- Test and fix FLE2 writeConcernError reporting in sharded writes
- Ensure minKey/maxKey is always logged in primary oplog for dbCheck
- Apply QuerySettings for timeseries collection queries
- Make validation checks for min/max and embedded timestamp robust against extended range events and changing granularity
- Replace invariant with tassert in awaitReplication
- Create targeted jstest for non-PIT sharded cluster magic restore
- Create targeted jstest for PIT sharded cluster magic restore
- Create targeted jstest for magic restore and chunk migration
- Create targeted jstest for magic restore and resharding
- Add serverStatus() metric to count replanning events
- Make sure that we skip dbcheck oplog entries during restore
- Audit references to shouldRecoverFromOplogAsStandalone and add checks for magic restore as needed
- Add targeted magic restore test for retryable writes
- Cap time spent applying dbcheck operations on secondaries
- Support a command that returns after an index build has started
- Ensure dbCheck only runs one job at a time on primary
- cache pressure causes one bucket per single document with time series collections
- Improve test coverage for collection registration upon moveCollection
- Add targeted test to perform a selective restore with magic restore
- Add targeted jstest test for magic restore with BalancerSettings
- tag analyze_shard_key_jscore_passthrough to require large hosts for tsan coverage
- Invalid role configuration for fsyncunlock command
- Override drop function and asserts in magic restore passthroughs
- Properly detect when OpenLDAP includes MozNSS-compatible shims
- Add testing matrix to backup and restore
- improve $map, $reduce, $filter
- Add shard renames to targeted sharded cluster magic restore tests
- Add targeted test for magic restore and config shard transitions
- Refactor magic restore utility assertions
- Add targeted test to run magic restore on data from an incremental backup
- Support PIT restore without entries after the checkpoint timestamp
- Prevent TTL Monitor from running collMod during rollback and initial sync fuzzers
- Allow dbHash command via direct shard connections
- Make magic restore hook more resilient to large PIT restore workloads
- Validate clustered collection record IDs in dbCheck
- Add test for clustered collections in dbCheck
- Verify index order and uniqueness in extra keys check in dbCheck
- libunwind deadlocks when called from signal handler while dl_iterate_phdr is running
- Log a warning message for sharded collections that have mismatched UUIDs
- Move magic restore main function out of main thread
- Relax the assertion when the document is not found in doDeleteRecord()
- Remove recursion from SessionCatalogMigrationSource::_fetchNextNewWriteOplog
- Recommit SERVER-71520
- Use correct nodeType for clusters in magic restore passthroughs
- Use timed retry loop when waiting for Azure OIDC Ingress Rules to Update
- Add targeted test to restore data files from a shard node into a replica set
- Fix dbcheck_stepdown.js
- Do not printStackTrace that does not warrant Atlas Failure Monitor
- Kill spillable queries when disk space is low
- Onboard MongoDB Server to Papertrail Service for SSDLC Compliance
- Executing an aggregation with transaction sub-stages that makes cursors on another shard can return NotARetryableWriteCommand error
- Update burn-in to handle multiple sub-suites in multiversion tasks
- Add a check to validation to check for changed bucketing parameters for time-series
- Linter that checks require_fcv_* tags are <= current version
- Port LDAP mongos health checker tests to use mock LDAP server
- cursor_hints_to_query_settings test suites do not close cursors for auto generated queries
- Use same backup cursor data files for all replica set nodes in magic restore tests
- Preserve approved cluster parameters across magic restore
- Clear sync source candidate if sync source selection fails
- Extract only $readPreference from $queryOptions in mongo test shell
- Make sure timeseries inserts perform a ShardVersion check before inspecting the existence of the buckets collection
- Replication coordinator should tolerate large list of locks
- change streams don't survive stepdowns when a rollback also occurs.
- [SBE] Produce EOF plan for pushed down $lookup when outer collection does not exist
- Resharding oplog application shouldn't use a unique lsid for each delete operation
- SetQuerySettingsCommandReply does not contain simplified query settings
- planningTimeMicros can include work not related to planning
- Test that dbCheck correctly catches missing _id indexes
- Remove the package_new_compile task
- In timeseries_agg_out.js allow 7268700 error code
- Enhance OCSP test code with debug information for response status tracking
- Update PyGithub to 1.58
- Update symbol mapper to use the original executed task
- Split out replica_sets_kill_primary_jscore_multiversion
- Fix behavior when running next() over a closed exhaust cursor
- Test magic restore specific logs without omitting test output
- Wait for replication in merge_let_params_size_estimation.js everywhere
- Create ShardedMagicRestoreTest class in magic restore testing module
- GeoHash might downcast with undefined behavior for point near the edge of 2d index bounds
- Bump cache_stuck_timeout_ms from 10min to 15min in TSAN
- Ensure collation is followed when using hash index in SBE IndexedLoopJoin
- Fix dbcheck_multiple_operations.js stepdown test
- Fix ShardRegistry::Time definition
- Avoid invariant due to in progress index build during initial sync
- Get mongod version and git version information without spinning up a mongod process
- Make variables in SASLGlobalParams atomic
- Drop config.placementHistory metadata in magic restore
- Ensure queryShapeHash is computed only on mongos in sharded clusters
- WiredTigerSessionCache does not properly dispose of sessions released after rollback
- Remove check for distinct keys after consecutive identical keys at the end of batch
- [v8.0] Avoid running incompatible_tsan tests under tsan on 8.0
- Do not convert StaleConfig for updateMany in a transaction to QueryPlanKilled
- Move user cache invalidation from OpObserver to onCommit handlers
- GlobalLock can potentially release the snapshot involuntarily
- Make collection_uuid_sharded.js temporarily not throw when reshardCollection fails due to concurrent index build
- Elide fuzzer timeouts due to slow multiversion runs
- Merge magic restore project
- Update TaskExecutorCursor behavior to be resilient to destruction during outstanding network operation
- Make pinning connection between mongod and mongot the default
- Improve testing of user cache invalidation
- Fix concurrency_replication_bulk_write with dbcheck
- ShardRegistry::scheduleReplicaSetUpdateOnConfigServerIfNeeded can write an incorrect config version
- Set enterprise-rhel-81-ppc64le timeouts on variant level
- Set runningWithBalancer for the sharding_jscore_passthrough_with_config_transition suite
- Increase the time spent waiting for balancer round to complete in enforce_zone_policy.js
- Make validate_timeseries_bucketing_parameters_change.js have hard-coded timestamps
- Non-pit logic in should_run_backup_or_restore seems incorrect.
- Support upsert duplicate key retry if unique index has collation
- Increase default secondary timeout limit for dbcheck
- Pin 10gen/jepsen to the latest jepsen-mongodb-master commit
- Address TSAN issues in unit tests explicitly calling AM->setAuthEnabled()
- Reduce powercycle timeouts to meet Evergreen constraints
- Nesting shard role into router role breaks collection metadata recovery
- Change ExpressionContext::getResolvedNamespaces invariant to a tassert
- Crash when stepping up while block user writes is enabled
- [Test-only] Implement a well behaved shut down process in standalone_in_queryable_backup_mode.js test
- Reduce resource usage for the random_moveChunk_timeseries_deletes.js FSM test on sanitizer builds
- Increase evg timeouts for blockprocessing and change stream fuzzers in additional build variants
- Replace optional chaining syntax in create_indexes_return_on_start.js
- Increase logging of direct shard connection errors and warnings
- The restore role should allow dropping system.views in any database
- Retry pipx install db-contrib-tool
- "_configsvrSetClusterParameter" command with "previousTime" parameter set cannot be reissued whenever the previous invocation sets the parameter and then fails
- Emit change stream events for the system.views collection (when showSystemEvents is enabled)
- CheckMetadataConsistency can trigger false positives due to looking at sharding metadata while critical section is active
- Reduce memory footprint for archived buckets in BucketCatalog
- Modify ignore_dbcheck_in_magic_restore.js to use refactored magic restore test utilities
- Reduce the number of pipelines generated by the fuzzer
- Add zstandard to TSAN denylist
- change_streams_shards_start_in_sync.js should temporarily hang shard1 instead of rejecting cursor establishing request from mongos
- Move ShardedMagicRestoreTest into separate file
- Sharded backup/restore tests may run out of memory on Windows variants for non-essential tasks
- CheckMetadataConsistency hook is not compatible with asio_transport_layer_integration_test
- Set reshardingOplogBatchTaskCount to 1 for upsert_unique_index_collation.js
- Annotate magic restore test utility classes with JSDoc-style comments
- Expose a command on the mongos to untrack a collection
- Add diagnostics to rollback recovery
- Create an observer for server lifecycle events
- Make CMConcurrency a no-op.
- Decrease timeout value to fix flakey dbcheck test
- Count insertion failures in SASL SCRAM cache
- Provide link to raw logs when parts of a task time out
- Prevent CollectionRoutingInfoTargeter from iterating all chunk ranges during update and delete
- Make CLUSTERED_IXSCANs obey query settings allowed indexes
- Rephrase error message about inconsistent bucket collection on upgrade to 8.0
- Modify Time Series Collection Parameters to Support Autoscaling
- MultiUpdateCoordinator can transition to 'done' state without releasing previously obtained resources
- Use a new database for the bucket_unpacking_with_sort_extended_range.js test
- SSLManagerWindows won't allow multiple CRLs from different issuers
- Update version check for reshardingDelayBeforeRemainingOperationTimeQueryMillis in servers.js
- Introduce configurable time limit to scan chunks during auto-merging
- [v8.0] Delete Antithesis tasks on 8.0
- Update MSI to reference libsasl2.dll instead of libsasl.dll
- Old garbage config.cache.* metadata might block catalog cache refreshes
- Transform CommandOnShardedViewNotSupportedOnMongod into CommandNotSupportedOnView on all code paths
- Relax the diff percentage window for the number of sampled queries in sample_rates_rs.js and sampled_rates_sharded.js
- Remove references to linkbench in system_perf.yml
- Modify speculative authentication auditing
- Improve variable names in $map and $filter serialization
- tassert tripped on 1-shard sharded $unionWith + $search
- Don't apply distinct scan optimization when $top(N)/$bottom(N)'s output is constant and sortBy is empty
- Validate against invalid PQS hints
- Add more log lines to FCBIS
- Improve Query Settings Fallback tests
- Preserve additional specified cluster parameters in magic restore
- Fix inequality check in releaseSession code
- [v8.0] Use absl::bit_count for bsoncolumn_test.cpp
- Fix an issue where change streams might incorrectly output a "drop" event during resharding or unsharding of a collection that is or was using zone sharding
- Ban changeStream tests from running in sharding_csrs_continuous_config_stepdown
- Plan summary on SBE does not show clustered ixscan
- $jsonSchema parser should not use empty strings to distinguish between top level schemas and sub schemas
- timeseries_create.js inserts a bucket with incorrect _id timestamp
- Retrying skipped records can try deleting record without wuow
- Replace invariant with assertion and add more details
- Closing archived buckets should not re-create stats for dropped collections
- [8.0] Extend timeout in initial_sync_unsupported_auth_schema.js assert.soon statements
- Add al2023-x86-compile v8.0 variant
- Prevent deleting ident directories concurrently with creation
- ImportCollection retry on bad metadata needs to roll back
- Create collection may be wrongly acklowledged on sharded clusters when write concern not respected
- Fix change_streams_split_event_v1_v2_tokens.js reshardCollection's initial chunk split (v8.0)
- update authenticode key alias
- upgrade "nongnu" libunwind to v1.8.1
- Consult log file rather than ramlog for speculative auth tests
- Use different collections for mapreduce out in read_pref_cmd.js
- Avoid mongos to sleep for 3 secs before shutting down
- Add test for records in the skipped record tracker that are no longer found in the collection
- Do not set reshardingDelayBeforeRemainingOperationTimeQueryMillis when launching patch versions where this parameter doesn't exist
- Repoint sys-perf to use genny-tasks from DSI
- Fix Jepsen and Antithesis to https cloning
- Enable dbCheck v2 feature flag on master
- Add a fallback error code for when we get SSL_ERROR_SYSCALL without an associated error.
- Upgrade MozJS to latest ESR 115 Minor Version
- resmoke's check for rogue processes incorrectly identifies unrelated processes on macOS
- Disable insert_duplicates_unique_index.js from slow suite on windows
- Add missing redact() calls to "Plan executor error" warning logs
- Fix shutdown race condition in repl_monitor_refresh.js
- Fix formula for calculating programMajorMinorVersion in servers.js
- Clear journal files between incremental backups in sharded_backup_restore.js library
- Handle interruptions while accessing ASIO sockets
- collection validation doesn't lock early enough to prevent racing against rollback
- Avoid tassert in test command sysprofile
- [v8.0] Fix wrong import in jstests/noPassthrough/index_stepup_missing_skipped_record.js
- Tag array_index_and_nonIndex_consistent.js as tsan_incompatible
- Rename incorrect overriden function in implicit_timeseries_collections
- [v8.0] Swap v8.0's bazel toolchain from mciuploads onto the persistent s3 bucket
- [v8.0] Merge dbcheck and server restore tool projects into 8.0.5
- Drop collection "test" at start of explain_all_plans_execution_stats.js
- Expect a single dropIndexes event in change_streams_split_event_v1_v2_tokens.js
- [v8.0] timeseries_reopened_bucket_insert.js assertion that a compressed bucket will be reopened fails on 7.0 in multiversion suites
- Upgrade mongo-task-generator to 0.7.20
- Replace invalid db in query_settings_index_hints_tests when run with the no_passthrough suite
- Move the remainder of the sys-perf config into DSI
- timeseries_reopened_bucket_insert.js has manually constructed bucket with mismatch in control ids and observed IDs
- Update the "search_meta_in_subpipeline_sharded.js" test
- Re-introduce queryHash for backwards compatibility
- Lower max fromjson depth from 250 to 200