Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
GODRIVER-3860Backlog
Improve performance monitoring and alerting
{panel:bgColor=#fafbfc}
h3. *Summary*
Make the Go Driver performance monitoring reliable, and make the performance alerts actionable.
Currently, we don't trust that the performance monitoring reliably points out meaningful changes, and we typically ignore performance monitoring alerts. We should improve the performance monitoring so we trust it to tell us something useful, and make the alerts actionable so we can do something when we get alerted. Consider borrowing from Node team's approa
Components
PerformancePerformance Benchmarking
SERVER-109955ClosedFixed
Add Fuzztest to build pipeline
Add [Fuzztest|https://github.com/google/fuzztest] to 10gen/mongo build pipeline replacing deprecated Libfuzzer tests.
Fixed Versions
8.3.0-rc0
SERVER-137428ClosedFixed
Implement query shape and query stats key for findAndModify
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137426ClosedFixed
Extend ParsedUpdate and ParsedDelete serialization for findAndModify
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-126317ClosedDone
Evergreen should mark FuzzTests as failed when they crash
FuzzTests are intended to run for a long period of time, and potentially surface multiple issues. As a result, the harness will observe crashes, record them to a special crash directory, and keep fuzzing. It seems to have a trick to keep itself from repeatedly hitting the same condition. After the target fuzz time elapses, the harness does not return a bad exit code to the shell. This causes resmoke to believe that the test "passed", and evergreen will mark the test as successful.
An an examp
Fixed Versions
9.1.0-rc09.1.0-rc1024
SERVER-108378ClosedFixed
Better cache lookup for `ReadConcern`
h2. Problem
On a 100% findById workload, {{_extractReadConcern}} is called on every query via the primary dispatch path: {{ServiceEntryPointShardRole::handleRequest}} → {{ExecCommandDatabase::_extractReadConcern}} → {{ReadWriteConcernDefaults::getDefault}} → {{ReadThroughCache::acquire}} → {{pthread_mutex_lock}}. In the common case — no cluster-wide read concern (CWRC) configured, no explicit read concern on the request, no transaction, external client — the function performs substantial work
Fixed Versions
9.0.0-rc0
NODE-6076Backlog
Add reasons to SDAM events
h3. Use Case
As a user of the driver,
I want driver events that can be emitted in multiple places to contain information about the cause of the event,
So that I can determine the cause of a given event without cross referencing other driver events and state when debugging.
Some driver non-error events can be emitted from multiple locations. `TopologyDescriptionChangedEvent`s, for example can be emitted when the client connects, as a response to a server heartbeat event, or as a res
Components
SDAM
NODE-7837Backlog
js-bson: Use Docker to run endian tests instead of zseries
h3. Use Case
As a... Node Driver developer
I want... run BSON CI tests in Docker, not on zseries hosts
So that... we don't have to rely on a small number of zseries hosts
h3. User Experience
* BSON endianess tests run in a few minutes in Docker, instead of taking days waiting for a zseries host
* old zseries BSON endianess tests are disabled, so they are never scheduled, but can be kicked off manually
h3. Dependencies
* none
h3. Risks/Unknowns
* Docker endian tests may fail i
NODE-4015Backlog
Standardize Connection String Validation Errors and MongoParseError
We use MongoAPIError interchangeably with MongoParseError when parsing connection string, and, moreover, the connection string module has its own version of the MongoParseError. It would be good to standardize the returned error from validating MongoClient options to at least extend from the same class in all of these different cases. We should probably have a separate type for connection string validation errors because we also use MongoParseError when we are literally unable to parse a message
Components
Connection StringError
SERVER-137854ClosedFixed
Default to 16KB leaf_page_max for new collections on disaggregated storage
Change the default {{leaf_page_max}} for newly created collections on disaggregated storage from WiredTiger's default of 32KB to 16KB. Attached storage should remain unchanged.
Collection tables are created without an explicit {{{}leaf_page_max{}}}, so they inherit WiredTiger's 32KB default. Index tables already set {{{}internal_page_max=16k,leaf_page_max=16k{}}}. This ticket is to make disaggregated storage use 16KB leaf pages for collections as per this [decision doc|https://docs.google.com
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-87118ClosedWon't Fix
Invariant after a forced reconfig with a hybrid index build pending quorum
I have discovered that on v5.0 and 6.0 a foced changing the votes/priorities for a down member while a hybrid index build is in progress crashes the server:
For comparison:
- v7.0.5 is able to handle a forced reconfig with the index build failed on "ReadConcernMajorityNotAvailableYet: Committed view disappeared while running"
- v4.4.28 forced reconfig does not seem to be affecting the index build in progress at all. If the primary is stepped down, the new primary will commit the index bu
Affected Versions
6.0.135.0.24
SERVER-138741ClosedFixed
Give SBE blueprint rewrite rules a shared slot summary and a way to queue nodes
Prepare the SBE blueprint rewrite engine for dead slot elimination.
* {{rewriteSbeBlueprint()}} takes an optional slot summary ({{SbBlueprintSlotUsage}}) that rules read and keep up to date through the rewrite context
* Rules can queue a node to be rewritten again with {{enqueue()}}
* The engine returns the new root, the number of firings, and whether it converged, and leaves the reaction to non-convergence to the caller
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-128558ClosedFixed
Make feature flag checks in ExpressionContext tolerate a null opCtx
Fixed Versions
9.0.0-rc08.3.8
SERVER-134753ClosedWon't Do
Add maximum byte limit for POA worker queues
This ticket adds a stopgap backpressure signal that complements the main in-flight bytes and batch count signal (SERVER-134780): the dispatcher blocks when any single worker's queued bytes exceeds a per-worker cap (e.g., 100 MB).
This ticket should include:
* Byte accounting at enqueue/consume — each {{WorkItem}} carries its serialized byte size, and the pool tracks per-worker queued bytes.
* CV wait — the dispatcher blocks on a condition variable after flushing a batch's slices if any work
SERVER-138547ClosedFixed
Use target architecture in cross-compiled archive directory names
h2. Problem
Cross-compiled release archives use the build host architecture in their top-level extraction directory. Dylan Richardson reported this when downloading the s390x enterprise RHEL 9 binaries for MongoDB 9.0.7: the archive extracts into an aarch64-named directory even though mongod correctly reports s390x.
Observed directory:
Expected directory:
The supplied mongod --version output reports:
h2. Root Cause
Source inspection on master at b7ea0b259e4 shows:
* {{bazel/wrapper_hook/wri
Fixed Versions
9.1.0-rc09.1.0-rc10258.3.178.0.389.0.8
SERVER-135262ClosedFixed
A sparse index can be considered multikey even if the would-be multikey document generates no keys
Consider a sparse index with key {{{{}"a.b": 1{}}}}. A document in that index's collection collection like
will be considered multikey even though it generates no keys for that index.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138719ClosedGone away
Standby counts a prepared transaction's size and count twice when it also records the commit entry
h2. Problem
The fix for the companion hash ticket makes the standby record a prepared transaction's size and count from its {{commitTransaction}} entry, as the primary does. With that alone, the standby still counts the transaction's operations as well, so it fasserts with 13626005 at roughly double the primary's size and count:
h2. Cause
Per-operation deltas are not recorded when the operation is applied. They are recorded by on-commit handlers that {{src/mongo/db/repl/oplog.cpp}} regi
SERVER-34824ClosedFixed
Make prepareTransaction command write a prepare oplog entry and use its optime as the prepare timestamp
Components
Replication
Fixed Versions
4.1.1
GODRIVER-1247ClosedFixed
connection leak with change stream
Using the latest, when creating change streams with Watch(), it looks like connections may be getting leaked.
Here's a minimal test case that illustrates the behavior:
When the code gets into the sleep loop at the end, there are lots of connections still open against the server. Running the same code using an older version of the driver results in the connection count going back down. Using git bisect, it looks like the change in behavior was introduced in df0f8c44059d04f60f45ebc8510065d43
Components
CRUD
Affected Versions
1.1.0
Fixed Versions
1.1.1