Known Issues
5166 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-135655ClosedFixed
Shared lazily-initialized state in S2 classes allow data races
The S2 data structures {{S2EdgeIndex}} and {{S2Loop}} have {{mutable}} state that can be changed by {{const}} methods. Previously this was thought to be innocuous because it only affected integer counters. However, {{map}} and {{multimap}} data structures are also affected, which can result in a server crash when accessed concurrently.
* [S2Loop::vertext_to_index_|https://github.com/10gen/mongo/blob/efd5d7c5aa7dc2dfbed81dd081165cb4b3322e45/src/third_party/s2/s2loop.h#L333]
* [S2EdgeIndex::ma
Fixed Versions
9.1.0-rc08.3.169.0.78.0.377.0.489.1.0-rc1026
SERVER-134783In Code Review
Support live worker resize for the pipelined oplog applier
This ticket supports live worker resize: a runtime change to {{replWriterThreadCount}} (shared with classic) drains, restarts the pool, and resumes dispatch.
This ticket should include:
* On a {{replWriterThreadCount}} change: drain the pipeline, restart the {{PipelinedApplierWorkerPool}} with the new worker count, and resume dispatch.
* The pipeline is drained first, so no in-flight state remains at resize time.
Fixed Versions
9.1.0-rc09.1.0-rc1026
GODRIVER-4243Blocked
Change stream leaks the server cursor when Next's context expires or is canceled
When the context passed to {{ChangeStream.Next}} expires, the driver tries to resume using that expired context. The {{killCursors}} for the old cursor fails without being sent, but {{BatchCursor.Close}} sets the cursor ID to 0 anyway. A later {{{}ChangeStream.Close{}}}, even with a live context, therefore sends nothing. The server cursor stays open until the idle cursor reaper removes it.
Applications that bound each {{Next}} call with a timeout leak one server cursor per timeout. On deploym
GODRIVER-3024Backlog
Runtime Error 'index out of range' in MongoDB Go Driver ExtJSON Parsing
h3. Detailed steps to reproduce the problem?
_Include full traceback, if possible_
Reproducible with the following script:
Output:
h3. Definition of done: what must be done to consider the task complete?
An io.Reader should return io.EOF when the stack is exhuasted.
h3. The ex
Components
BSON & Extended JSON
GODRIVER-4260ClosedDeclined
mongo-go-driver - PR #2719: Merge release/2.10 into master
[mongodb-drivers-pr-bot[bot]|https://github.com/apps/mongodb-drivers-pr-bot] has created [PR #2719: Merge release/2.10 into master|https://github.com/mongodb/mongo-go-driver/pull/2719] in [mongo-go-driver|https://github.com/mongodb/mongo-go-driver]
SERVER-139249Needs Scheduling
Preserve traffic recording request snapshots during exhaust processing
Preserve traffic recording request snapshots during exhaust processing
SERVER-139496Needs Scheduling
S2 heuristic counters have incorrect behavior on overflow
Our S2 geospatial third party library uses {{int}} counters for heuristics, like for when the cost of building a data structure indexing polygon vertices is likely to be amortized.
* [S2Loop::num_find_vertex_calls_|https://github.com/10gen/mongo/blob/af22eb75e478f32fde03e16dfa748e4defa22055/src/third_party/s2/s2loop.h#L356]
* [S2EdgeIndex::query_count_|https://github.com/10gen/mongo/blob/af22eb75e478f32fde03e16dfa748e4defa22055/src/third_party/s2/s2edgeindex.h#L243]
For large collections
SERVER-137511ClosedDuplicate
Complete TODO listed in SERVER-123384
There is a TODO in code owned by your team which references a resolved ticket.
The TODO appears in:
* src/mongo/db/replicated_fast_count/replicated_fast_count_test_helpers.cpp:483
You can also search for all references to the resolved ticket here:
https://github.com/10gen/mongo/search?q=SERVER-123384&type=Code
The next steps for this ticket are to either remove the outdated TODO or follow the steps in the TODO if it is correct. If the latter, please update the summary and description of this
SERVER-139455ClosedFixed
Remove existence check in backports required for multiversion linter
This removes the requirement for entries in etc/backports_required_for_multiversion_tests.yml to match files on that branch.
Requiring existence for test files turned out to be too strict. It is common for a test to be renamed/relocated across backports, but the entry still needs to match the path used on master.
Components
Testing Infrastructure
Fixed Versions
9.1.0-rc09.1.0-rc1026
SERVER-126535Backlog
Comment field not propagated to shards in updateMany with pause-migrations flow
When a client attaches a {{comment}} to an {{update}} or {{delete}} command that triggers the pause-migrations coordinator path ({{_shardsvrCoordinateMultiUpdate}}), the {{comment}} field is silently dropped and never reaches the shard that executes the write. This means the comment never appears in the shard's slow query log, breaking user expectations and making cross-cluster observability difficult.
h2. Root Cause
In [coordinate_multi_update_util.cpp:135-139|https://github.com/mongodb/m
SERVER-139167Backlog
mongo shell segfaults (SIGSEGV, "Invalid access") running jstests/core/queryable_encryption/basic_crud.js on 8.0.35
h3. Summary
The {{mongo}} shell (8.0.35) crashed with SIGSEGV while running {{jstests/core/queryable_encryption/basic_crud.js}}. No stack trace file was produced, so the crash is unsymbolized.
Found in MongoDB CI for the mongosync (Cluster-to-Cluster Sync) project. The test runs against a replica set; replication and the cluster itself were healthy at the time.
h3. Log
resmoke recorded the test as exiting {{-11 (SIGSEGV)}}. Stack trace acquisition found nothing:
h3. What this is not
Th
SERVER-138002Investigating
Ensure long-running query spilling yields resources so global operations can make progress
h2. Problem
Long-running query spilling, spill-run consolidation, and recovery/merging of spilled state can execute inside a single stage/executor call while retaining locks or other execution resources. Where Global IS remains held, setUserWriteBlockMode may be unable to acquire Global X promptly. Queued Global X can also block new Global IS acquisitions and stall unrelated operations.
Interrupt checks alone do not release locks. Existing interrupt-delinquency fixes improve cancellation r
SERVER-139095In Code Review
Remove sbeFull parameter from getMinRequiredSbeCompatibility
{{getMinRequiredSbeCompatibility}} takes a redundant {{sbeFull}} parameter; every caller passes {{gFeatureFlagSbeFull.isEnabled()}} or a value that is always false.
* Read {{gFeatureFlagSbeFull}} inside the function instead.
* Drop the dead {{sbeFull}} parameter from {{shouldUseRegularSbeDeferredEngineSelection}}.
* No behavior change.
Fixed Versions
9.1.0-rc09.1.0-rc1026
SERVER-139353Backlog
Add additional safeguards on inadvertent modification of Queryable Encryption safeContent tags
Currently, doing a direct update or findAndModify to an encrypted collection's {{__safeContent__}} array when {{bypassDocumentValidation}} is {{false}} is only prevented if the operation is a replacement-type update, or a {{$set}} on the {{__safeContent__}} field itself. It does not error if you try to {{$set}} a dotted path like {{__safeContent__.0}} or if you use other modifier-style update operation like {{$push}} or {{$pop}} on it.
SERVER-139239Needs Scheduling
Avoid checksumming exhaust responses twice on the hot replication path
h1. Avoid checksumming exhaust responses twice on the hot replication path
SERVER-139492In Code Review
Split evergreen tasks for C++ integration tests
Currently, the C++ integration test tasks are organized according to the fixture they spawn or what feature flags and server paramters they use. For an example of simple fixture-based organization, integration_tests_replset is most C++ integration tests that run against a replica set. For an example of feature flagging, {{integration_tests_standalone_grpc}} runs a small set of C++ integration tests against a server with ingress gRPC enabled.
To support infinite burndown, this change splits thos
SERVER-139298ClosedFixed
Remove allowDeprecatedAfterOpTimeReadConcern from removed parameters lis
Now that SERVER-138119 is merged into master, remove this override.
Fixed Versions
9.1.0-rc09.1.0-rc1026
PYTHON-5533ClosedDone
Investigate changes in SERVER-109626: Adjust behavior for emitting additional fields in change streams
This ticket was split from {{DRIVERS-3278}}, please see that ticket for a detailed description.
Components
Change Streams
SERVER-111051Backlog
Shutdown cleanly when ESE is unable to acquire the KMIP Key
In AF-6822, we see a case where network causes us to fail acquiring the customer KMIP key. This triggers an invariant within wiredtiger, since wiredtiger_open on startup fails. Since this is not an "issue" in code, but rather a failure on startup, we should consider failing mongod more gracefully.
PYTHON-6166ClosedDone
Remove mention of waitQueueSize and waitQueueMultiple connection pool options from spec
This ticket was split from {{DRIVERS-3675}}, please see that ticket for a detailed description.
Components
CMAP