Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-132668ClosedFixed
Make ErrorDuringSecondBatchApply test less flaky
Fixed Versions
9.1.0-rc09.0.9
SERVER-138611ClosedWon't Fix
Indexed Join query in SBE can tassert when index is dropped simultaneously
An aggregation with a $lookup that's pushed to SBE can tassert with 6447401.
The query planner records the index during planning, multiplanning can yield and the SBE plan is built in a different snapshot where the index may no longer be available.
The impact is that a query can tassert if its stage building happens concurrently with an index drop for one of the foreign collections. Note that if an index on the "primary" collection is dropped at runtime, the query uasserts, so the user impa
Affected Versions
8.0.37
SERVER-137498ClosedFixed
Update PGO profiles
Updated PGO profile numbers for performance.
Fixed Versions
9.0.9
SERVER-137341ClosedFixed
Fix audit proxy endpoint assertions that assume distinct TCP source ports
h2. Problem
The shared audit test helper assumes that a proxied connection always has different remote and directRemote source ports. This is not guaranteed: the client\-to\-proxy and proxy\-to\-server TCP connections have different destination ports, so they may legally use the same ephemeral source port.
This causes intermittent failures in audit\_remote\_local\_intermediate.js even when the audit fields are correct. See BF\-45214.
h2. Evidence
\* BFG\-3468973: July 27, 2026, mongodb\-mongo\
Components
Networking
Fixed Versions
9.1.0-rc09.1.0-rc10259.0.9
SERVER-136240ClosedFixed
Revisit new FailPointEnableBlock functions (and use of tassert)
SERVER-135703 introduced some new functions to FailPointEnableBlock to wrap waitForTimesEntered. We should revisit these names since it's been suggested that a singular {{wait(int n = 1)}} function would be a better replacement for them.
In addition, the {{tassert}} should probably be {{invariant}} since there's little to no reason for test-only code to be raising {{tassert}}.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137609ClosedFixed
Update PGO profiles
Updated PGO profile numbers for performance.
Fixed Versions
9.0.9
GODRIVER-2579Needs Triage
Incorporate connection pool checkout into server selection loop
This ticket was split from {{DRIVERS-1262}}, please see that ticket for a detailed description.
Components
CMAPSDAMServer Selection
GODRIVER-2901Needs Triage
Separate batching, retry, and single-operation execution logic
Currently, all batching, retry, server and connection selection, single-operation execution, and error handling logic are part of the giant 500-line [Operation.Execute|https://github.com/mongodb/mongo-go-driver/blob/f8b88fc241a91c32455f61c57557921ba7d17d2d/x/mongo/driver/operation.go#L412] function. That function is becoming unmaintainable as more and more logic is added to it (e.g. RTT short-circuit, memory buffer reuse, additional error handling, etc). Difficulty maintaining {{Operation.Execut
Components
Retryability
GODRIVER-3012Needs Triage
Create a CRUD query filters helper
There are a number of CRUD APIs that accept a filter parameter, including:
Currently, most users define filters using {{bson.M}} or {{bson.D}}, which resembles the API used by {{mongosh}}. However, the syntax of {{bson.D}} can be extremely awkward (see GODRIVER-2271) and defining filters with {{bson.M}} requires users to know the proper filter structure and is subject to common field name typos (e.g. "_id").
A much better way to define filters would be using query filter helpers that inf
Components
CRUD
GODRIVER-3094Needs Triage
Use atomic.Value instead of sync.RWMutex to provide atomic read/writes to values
h3. *Context*
Go's {{atomic.Value}} type can perform a lot better than using a {{sync.RWMutex}} to guard values, especially when the use case is read-heavy and has a lot of concurrent contention. See some benchmarks [here|https://gist.github.com/dim/152e6bf80e1384ea72e17ac717a5000a]. We should use {{atomic.Value}} instead of a {{snc.RWMutex}} in those cases.
h3. *Definition of done*
* Replace all uses of {{sync.RWMutex}} to guard values that are read frequently and written infrequently with
GODRIVER-3372Needs Triage
Make writing integration tests easier
{panel:borderStyle=solid|borderColor=#21313c|titleBGColor=#21313c|titleColor=#fafbfc|bgColor=#fafbfc}
h3. *Summary*
_What is the problem or use case, what are we trying to achieve?_
h3. *Motivation*
h4. Who is the affected end user?
_Who are the stakeholders?_
h4. How does this affect the end user?
_Are they blocked? Are they annoyed? Are they confused?_
h4. How likely is it that this problem or use case will occur?
_Main path? Edge case?_
h4. If the problem does occur, what ar
Components
Testing
GODRIVER-3721Backlog
Develop testcontainer-based alternative to prose tests in internal/test
h3. *Context*
Integration and prose tests in the mongo-go-driver currently rely on shared orchestration infrastructure, which creates a bottleneck and increases test execution time.
The Go Driver team should organically develop a testcontainer-based testing pattern as an alternative to the existing prose tests in the internal/test directory. This will allow tests to run in parallel containerized environments, scaling with the number of available CPU cores rather than competing for shared o
GODRIVER-3005Needs Triage
Automatically add the full command error in error messages
Automatically include [mongo.CommandError.Raw|https://pkg.go.dev/go.mongodb.org/mongo-driver/mongo#CommandError] in the error message formatted by [replaceErrors|https://github.com/mongodb/mongo-go-driver/blob/v1/mongo/errors.go#L49]. For example, pymongo does the following:
Java also does this: https://github.com/mongodb/mongo-java-driver/blob/r4.0.2/driver-core/src/main/com/mongodb/MongoCommandException.java#L50-L51
GODRIVER-2862Needs Triage
Merge driver.BatchCursor into mongo.Cursor
Once GODRIVER-2861 has been completed, there is no longer a valid reason to have the driver.BatchCursor abstraction. The team should consider merging the driver.BatchCursor into mongo.Cursor, replacing the driver.BatchCursor struct with the mongo.batchCursor interface.
SERVER-127904ClosedGone away
Thread ExpCtx directly through to explain or PlanExecutorSBE
Currently, in explain.cpp we access the ExpCtx through the CanonicalQuery that is stored on the executor (e.g. PlanExecutorSBE) ([example#1|https://github.com/10gen/mongo/blob/ce3dae58890ef7aa9b82fe6423f4f0bc39c993cd/src/mongo/db/query/explain.cpp#L116], [example#2|https://github.com/10gen/mongo/blob/ce3dae58890ef7aa9b82fe6423f4f0bc39c993cd/src/mongo/db/query/explain.cpp#L144]).
The code here makes the assumption that every executor will have a single CQ driving it and thus, it accesses the
SERVER-131913ClosedFixed
Disabled general mirrored reads during targeted tests
The mirrored reads counters get incremented for any mirrored read, general or targeted. Not explicitly setting samping to 0 on general mirrored reads leaves the on at a default value, so there's a small chance any request is sampled for a general mirrored read. Such a request may be delayed, and arrive after testing the baseline for a targeted mirrored reads test, thus making counters show an additional mirrored read. This seems to have caused [BF-44836](https://jira.mongodb.org/browse/BF-44836)
Fixed Versions
9.1.0-rc09.0.48.3.18
SERVER-132383ClosedFixed
Inconsistent defaultClientBaseBackoffMillis/defaultClientMaxBackoffMillis crashes server via BackoffWithJitter invariant
Discovered in failing fuzzer test
Fixed Versions
9.1.0-rc09.0.48.3.18
SERVER-132440ClosedFixed
Otel metrics and traces shutdown is not thread-safe
These can deadlock due to races with how the objects in the otel library obtain the relevant condition variable. This is only really a problem when the export time is too long, as when the next export triggers it unblocks the shutdown.
Fixed Versions
9.1.0-rc09.0.48.3.18
SERVER-134719ClosedFixed
Update performance microbenchmark thresholds
Update the instruction-count thresholds used to block PRs.
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.48.3.18
SERVER-122587ClosedFixed
Coverity: add GUARDED_BY annotations to replication coordinator external state
h2. What
Add `GUARDED_BY(_threadMutex)` to the member fields in `ReplicationCoordinatorExternalStateImpl` that are accessed without the mutex, and `REQUIRES(_threadMutex)` to the private helpers that read or write them. Fix all callers that must acquire the lock before invoking those helpers.
Files: `replication_coordinator_external_state_impl.cpp` (lines 351, 361, 512, 517, 522, 538, 595).
Resolves: SERVER-108741, SERVER-108742, SERVER-110036, SERVER-110582, SERVER-110600, SERVER-114009, SER
Fixed Versions
9.0.0-rc08.3.18