Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-137677ClosedFixed
Add zstd decompress pointer-overflow to UBSan denylist
h2. Overview
The vendored zstd (1.5.5) computes a prefetch address in ZSTD_prefetchMatch (lib/decompress/zstd_decompress_block.c) as dictEnd + prefetchPos - offset. On corrupt input with no dictionary, dictEnd is NULL, which UBSan reports as "applying non-zero offset to null pointer". The address is only passed to PREFETCH_L1 and never dereferenced. Upstream fixed this in zstd 1.5.6 with ZSTD_wrappedPtrAdd/ZSTD_wrappedPtrSub.
The libfuzzer variant hits this within seconds in protocol_fuzzer (se
Fixed Versions
9.1.0-rc09.1.0-rc1025
CSHARP-6261Ready for Work
Make numCandidates optional in $vectorSearch query builders
This ticket was split from {{DRIVERS-3586}}, please see that ticket for a detailed description.
CSHARP-6151Ready for Work
Ensure compatibility bw autoEmbedding indexes with nestedRoot fields
This ticket was split from {{DRIVERS-3585}}, please see that ticket for a detailed description.
SERVER-136545ClosedFixed
Validate per-collection size and count DDL in steady state
DDL operations present edge cases when performing checkpoint validation:
# If a collection is created and dropped within the same checkpoint, it is absent from the primary's checkpoint, and the standby should not expect to see an entry for that collection.
# If a collection is created and dropped _across_ checkpoints, the collection is present in the primary's checkpoint as a container delete, and the standby should expect to see that collection in the checkpoint.
This is the motivation f
Fixed Versions
9.1.0-rc09.1.0-rc1025
JAVA-6340Ready for Work
Make numCandidates optional in $vectorSearch query builders
This ticket was split from {{DRIVERS-3586}}, please see that ticket for a detailed description.
SERVER-134782ClosedFixed
Implement step-up drain-complete handshake for the pipelined oplog applier
This ticket handles step-up: on a step-up signal, the applier drains and signals drain-complete so the replication coordinator can finish transitioning to primary.
This ticket should include:
* On the step-up signal, stop pulling and perform a full drain, then signal drain-complete to the replication coordinator.
* A drained pipeline implying no in-flight batches and no outstanding deferred writes, so the new primary starts from a fully-applied, consistent state.
Fixed Versions
9.1.0-rc0
PYTHON-6200ClosedDone
Wait for monitor threads/tasks to finish in test_close_closes_sockets
h3. *Context*
test_close_closes_sockets can be [flaky|https://spruce.corp.mongodb.com/task/mongo_python_driver_compression_zstd_ubuntu_22_test_standard_rapid_python3.15t_sync_noauth_nossl_standalone_d6fc1bdf06919d5f3bd07b1a3a93f19a69fb22f3_26_10_27_20_02_06/tests?execution=0&sorts=STATUS%3AASC] due to test not waiting for monitor threads/tasks to close before asserting.
h3. *Definition of done*
Make the test wait until monitor threads/tasks complete.
h3. *Pitfalls*
{color:#505f79}What s
SERVER-138109ClosedFixed
Skip multikey path-coverage validation for documents generating no index keys
For a multikey index, collection validation currently [always checks that a documents multikey paths are covered by the collection's multikey paths|https://github.com/mongodb/mongo/blob/6edb5ef80106eb8a2d243b611208c38c635adc58/src/mongo/db/validate/key_string_index_consistency.cpp#L1056-L1057]. We should skip that check if the document does not generate any keys, which could be the case for instance for a sparse index.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138121ClosedFixed
Retry addShard on write conflicts during ShardingTest initialization
{{ShardingTest}} initialization can fail when {{addShard}} encounters a transient write conflict. This occurs before test-local setup retries can run.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138096ClosedFixed
Handle write conflicts in create_indexes_return_on_start.js sharding setup operations
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-136813ClosedFixed
createIndexes returnOnStart test should run using PDIB
Currently, [the jstest for the {{createIndexes}} command's {{returnOnStart}} option|https://github.com/mongodb/mongo/blob/dd09974a1f0116b5d89cd84768d3437c6be4ef5f/jstests/noPassthrough/index_builds/create_indexes_return_on_start.js] is tagged with {{requires_commit_quorum}}, which means that it does not run in suites that enable primary-driven index builds.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-135897ClosedWon't Do
Affected foreign keys table and foreign key journal for delete processing
no description yet
SERVER-134801ClosedFixed
Misleading "non-array path became multikey during yield" error returned if the index has been dropped
As per , path arrayness can apparently not distinguish between the case where the index became multikey from the case where the index was dropped.
As a result, for an index drop, without any change of multikeyneess, the following error is returned to the client:
This error is misleading. If it is not possible to distinguish the two cases, then the error should be changed to "index dropped or non-array path became multikey..." or something along those lines.
Note that this error mess
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-136931ClosedFixed
Add (default off) enablement flag and maxDoP for parallel query support
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138214ClosedFixed
Fix test race in stepdown_write_straddling_under_load.js
If stepdown is too slow, then the node rejects the {{replSetFreeze}}. Retry the command to satisfy slow variants.
Fixed Versions
9.1.0-rc09.1.0-rc1025
PYTHON-6218ClosedDone
Fix failing fuzz tests
h3. *Name of Failure:*
Fuzzing
h3. *Link to task:*
https://spruce.corp.mongodb.com/task/mongo_python_driver_fuzz_rhel8_fuzz_f2c5c250ca870ad2d832e95273bfc489fe8bcb26_26_10_30_13_12_19/logs?execution=0
h3. *Context of when and why the failure occurred:*
The BSON C extensions don't reject invalid UTF-8 regex flags but the Python BSON package does.
h3. *Stack trace:*
SERVER-132098ClosedFixed
Upgrade timelib and timezone data files to 2026e
2026c includes permanent timezone updates going into effect in September and November of 2026.
The [timeZoneInfo parameter|https://www.mongodb.com/docs/manual/reference/configuration-options/#mongodb-setting-processManagement.timeZoneInfo] can be used to reference up-to-date time zone info on unpatched versions.
Fixed Versions
9.1.0-rc08.3.169.0.78.0.377.0.48
SERVER-138285ClosedWon't Do
Disable index_build_killop_secondary_before_commit.js on Windows
BF-47023: {{jstests/noPassthrough/index_builds/index_build_killop_secondary_before_commit.js}} failed on windows-debug-suggested. Re-disable the test on Windows by adding the {{incompatible_with_windows_tls}} tag to get the waterfall green, until the underlying test race is fixed in SERVER-138284 (bug ticket).
h2. Background
The second phase of the test intentionally crashes the secondary via fassert 51101 by sending a killOp after the index build has voted for commit. The test asserts the
SERVER-137717ClosedFixed
Denylist express_id_lookup_missing_index_race.js from multiversion testing until SERVER-135788 flag enablement
h2. Problem
{{jstests/sharding/query/change\_streams/express\_id\_lookup\_missing\_index\_race.js}} fails flakily on mainline multiversion sharding tasks \(e.g. {{sharding\_last\_lts\_2\-linux\-enterprise}}\), in the subtest _"updateLookup succeeds via the aggregation fallback on a relocated post\-image"_:
The lookup itself _succeeds_ via the aggregation fallback \({{aggregation.found = 1}}\); only the SBE primary's decline counter never moves.
h2. Root cause
* The subtest relies on the de
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-136233ClosedFixed
Implement parallel planner for collection scans (non-hybrid)
Fixed Versions
9.1.0-rc09.1.0-rc1025