Known Issues
5166 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-135152ClosedFix Not Needed
[MFP][non-emergency] Non-aborting mode for the parse-all plugin in perf runs
{{parser_validator_test_plugin}} aborts the process on any parser rejection:
{{{}die() -> std::process::abort(){}}}. This behavior is correct for resmoke test suite but not suitable for perf tests as one false positive during the run kills the run and yield no data.
Add a mode where a parser rejection is logged and counted instead of aborting. Default stays abort.
h3. Tasks
* A parser rejection in perf mode does not abort the process.
* resmoke behaviour is unchanged.
* One artifact s
SERVER-135865ClosedFixed
Download native macOS LLVM toolchain from S3 instead of Homebrew
Homebrew no longer builds bottles for macOS 14, which broke `brew install llvm` on the macOS 14 dev image. This makes the native macOS toolchain download LLVM through Bazel, the same way our other toolchains work, so hosts no longer need LLVM from Homebrew.
- `bazel/toolchains/cc/mongo_apple/mongo_apple_toolchain.bzl`: replaces the `brew --prefix llvm@19/lld@19` lookup. The rule now downloads a sha256-pinned LLVM 19.1.7 archive for the host arch (arm64 or x86_64) from `mdb-build-public/toolchai
Components
Homebrew
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.98.3.188.0.39
SERVER-130139ClosedFixed
Account for o-vs-UUID consistency check
Tracks the server fix for the following vulnerability:
In {{applyOps}}, a command op ({{op:"c"}}) has its authorization and execution target diverge:
* Auth ({{OplogApplicationChecks::checkOperationAuthorization}}) remaps {{nss}} to the {{ui}} UUID's collection but keeps only its database, then authorizes the inner command parsed from the {{o}} body — i.e. it checks rights on {{<UUID-db>.<o-coll>}}.
* Execution (drop/collMod/dropIndexes applyFuncs) resolves the target purely by UUID via {{e
Fixed Versions
8.2.138.0.297.0.409.1.0-rc08.3.89.0.0-rc2
SERVER-120042In Code Review
Cannot validate collection with timestamp if not present in the database
When attempting to validate a collection using a snapshot, it will fail if the collection has been dropped from the active database.
SERVER-139402ClosedFixed
Remove unused BSONColumn::contains_forTest and its unit tests
{{BSONColumn::contains_forTest}} ({{src/mongo/bson/column/bsoncolumn.cpp}}) is dead code. SERVER-65804 added it in 2023 to speed up a time-series check, but no production code calls it. SERVER-74926, which would have finished it, was closed as Won't Do. Only seven unit tests call it ({{ContainsScalar*SimpleCompressed}} in {{bsoncolumn_test.cpp}}), so they test dead code.
It also has a bug, found by reading the code and not by running a test: after a Simple8b run it advances {{8 * numBlocks}} by
Fixed Versions
9.1.0-rc09.1.0-rc1026
SERVER-135628ClosedDuplicate
A brief KMIP outage during startup aborts the node instead of being waited out
h2. The problem
When a node starts and cannot decrypt its checkpoint because the KMIP server is briefly unavailable, the node aborts. It does so again on every restart, for as long as the outage lasts, instead of becoming a secondary once KMIP answers. The outage need not be long or permanent — the same stall was ridden out successfully in one incident and aborted the node in another, and what separated them was whether the last permitted retry happened to succeed.
The abort also names no
SERVER-137534Backlog
Remove Feature Flag for SPM-4895
Feature flags should be proactively removed to address technical debt.
See the [Feature Flag Github page|https://github.com/mongodb/mongo/blob/master/src/mongo/db/repl/FCV_AND_FEATURE_FLAG_README.md#feature-flags] for additional guidance.
SERVER-138621Backlog
Smart test selector for pre-merge patch testing
The smart test selector picks a small set of JS tests for a PR patch and it maps those tests to resmoke suite tasks (similar to burn-in style), and also picks C++ unit-test targets. It uses below signals:
1. Git co-change history. A co-change occurs when two files change in the same commit.
2. Manual C++-to-JS test associations that the owning team maintains in YAML.
3. Golden-output test mappings.
4. A static reverse-import graph for JS libraries.
5. A bounded Bazel {{rdeps}} query for C
Components
Query Execution
SERVER-136427Backlog
Clustered collscan: Avoid overshooting inclusive range boundaries
Clustered indices are always unique, so once we hit the end of a range with inclusive boundary, we can immediately seek to the beginning of the next range/flag the stage as EOF instead of calling another cursor->next() & only noticing the overshoot afterwards.
Reference: [SBE|https://github.com/10gen/mongo/blob/81b916c3ca9757cd95f68b6658e01db2f421f52e/src/mongo/db/exec/sbe/stages/multi_range_clustered_scan_stage.cpp#L88] and [classic|https://github.com/10gen/mongo/blob/81b916c3ca9757cd95f68b6
SERVER-135753Backlog
Investigate whether the execution ticket can be released independent of a nested non-global lock wait
* Problem: when an operation already holds a ticket (from an earlier, broader lock acquisition, e.g. the global lock) and then blocks waiting on a second, narrower lock (e.g. a $merge target collection lock, or $collStats with waitForLock: true), it keeps holding the ticket for the entire time it's blocked. Confirmed via repro (a 4.2-second lock wait produced 4,224ms of ticket delinquency with no interrupt-side signal at all).
* This is a ticket delinquency-only issue; the interrupt side exe
SERVER-139047In Code Review
Propagate Stitch Support static test API declarations
Propagate {{STITCH_SUPPORT_STATIC}} from {{stitch_support_duplicate}} to its static test consumers in {{src/mongo/embedded/stitch_support/BUILD.bazel}}. This makes callers use direct symbols rather than Windows DLL-import references. The mismatch can produce LNK4217, and fails when linker warnings are fatal. {{STITCH_SUPPORT_COMPILING}} stays private.
The exported {{stitch_support}} target has {{linkshared=True}}; its definition matches master exactly. Neither target propagates {{STITCH_SUPPORT
Components
Bazel
Fixed Versions
9.1.0-rc09.1.0-rc1026
PYTHON-6266In Progress
Make the synchro pre-commit hook atomic and write only real changes
h3. *Context*
The {{synchro}} pre-commit hook regenerates the synchronous driver and test
files in place, and the {{synchro-stage-files}} hook stages what it wrote. When
a run fails partway, or a later hook fails, dozens of modified generated files
are left in the working tree. [CONTRIBUTING.md|https://github.com/mongodb/mongo-python-driver/blob/master/CONTRIBUTING.md]
documents this failure mode and a manual cleanup workaround. Developers must
then tell their own changes apart from gene
Components
Infrastructure
SERVER-139118ClosedFixed
Have SBE plan trialing produce Documents instead of BSON when document sources follow
When a single solution plan uses SBE, we lower the QSN to SBE and execute through {{PlanExecutorSBE}} which is aware of whether it needs to [produce BSON or Documents|https://github.com/10gen/mongo/blob/5d9e8a85d4eadd4ba300cc1e0db8b6a6e69abea2/src/mongo/db/query/plan_executor_sbe.cpp#L371-L379]. If a document source follows the SBE plan, we produce Documents since this is the representation they use. Otherwise if the SBE plan is the end of execution, it can produce BSON to return to the user.
W
Components
Query Execution
Fixed Versions
9.1.0-rc09.1.0-rc1026
SERVER-138551Needs Scheduling
StaleConfig error when $unionWith pipeline begins with $collStats in random shard passthough
An aggregation whose pipeline contains a {{$unionWith}} whose sub-pipeline begins with a cursorless stage ({{{}$collStats{}}}) fails with a *non-retryable* StaleConfig ("timestamp mismatch detected") whenever the node that executes the {{$unionWith}} is not the shard that owns the foreign collection. This is hit deterministically on clusters where tracked unsplittable collections live off the merging node (e.g. after {{{}moveCollection{}}}, or when unsharded collections are tracked upon creation
SERVER-138288In Code Review
Add ability to disable resharding auto commit
This will add a new parameter to prevent resharding from automatically deciding to proceed to commit procedure and enter critical section automatically. Users will have to manually send the commit command to force it to proceed to commit or turn off the parameter to allow resharding to do it automatically again.
Fixed Versions
9.1.0-rc09.1.0-rc10259.0.98.3.18
SERVER-139458Needs Scheduling
Pass monguard option through ReplicaSetFixture and ShardedClusterFixture
After MongoDFixture and MongoSFixture supports a monguard option, pass the monguard option through ReplicaSetFixture and ShardedClusterFixture
SERVER-139462Needs Scheduling
Add monguard option to ReplSetTest JS test shell helper
Use the Monguard option in MongoRunner, add monguard option for ReplSetTest JS Test Helper
SERVER-139459Needs Scheduling
Add a --monguard resmoke flag and set up matrix suite overrides
SERVER-139457Needs Scheduling
Add a monguard option to MongoSFixture
Similar to https://github.com/10gen/mongo/pull/66521, add the monguard option for sharded resmoke clusters