Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-110825ClosedFixed
Rename isSerializingForQueryStats() to isShapifying()
We originally built query shape serialization just for query stats, but it's now used by multiple features. This is just outdated and should be updated
Fixed Versions
9.1.0-rc0
SERVER-138726ClosedWorks as Designed
Duplicate reshardCollection change events emitted when a reshard commit is retried
A single reshard operation can emit *two* {{reshardCollection}} change events (i.e., the completion noop oplog entry is written twice to the same shard). This is user-visible: a client consuming change streams receives the event twice for one reshard. It was observed in CI on the change-streams sharding FSM suites (3 failures in ~35 runs on the all-feature-flags variants, tracked in BF-47107) and reproduced deterministically with a diagnostic failpoint.
I spoke with , who is of the opinion th
SERVER-138694ClosedFixed
Add metric for index build multikey side writes
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137192ClosedFixed
CollectionCatalog support for adding / removing foreign key constraints
Limited to:
* CollectionCatalog internal, no integration.
* Only create collection and drop collection
* Inbound / outbound foreign key constraint tracking at Collection level.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-136815ClosedWon't Do
Disambiguate AsyncResultsMerger::exhausted() from cursor-open resource checks
h2. Overview
AsyncResultsMerger::RemoteCursorData::exhausted\(\) returns only "cursorId == 0" \(the remote cursor is closed\), but the name reads as "the remote is truly done". Several call sites use it to decide whether the shard still holds server\-side cursor resources that must be killed or released. For those sites, "cursor is closed" is the correct condition, but the name hides that intent and has already caused confusion about what "exhausted" guarantees.
h2. Proposed change
Split the tw
SERVER-102582ClosedFixed
Disallow overriding readConcern on $listClusterCatalog
Right now, any aggregation command will override its readConcern if the command runs under a session.
However, the $listClusterCatalog command can only run with a 'local' read concern (because the listCollections command running under the hood only accepts 'local' read concern. Therefore, we should [update this code|https://github.com/10gen/mongo/blob/9ca4e977abb03dbf016137dfdbab270ba2554e4b/src/mongo/shell/session.js#L161-L191], and probably [this one|https://github.com/10gen/mongo/blob/3d9399
Affected Versions
8.1.0-rc0
Fixed Versions
8.2.0-rc08.1.1
SERVER-130209ClosedFixed
Switch to using per-operation QuerySettings and QueryKnobConfiguration
Fixed Versions
9.0.0-rc1
SERVER-138499ClosedFixed
Merge main-staging-poc branch 2026.11.03
Includes:
* [SERVER-137535|https://jira.mongodb.org/browse/SERVER-137535] — Move ARM txn bookkeeping into helper function (https://github.com/10gen/mongo/pull/64898)
* [SERVER-137536|https://jira.mongodb.org/browse/SERVER-137536] — Model ARM per-remote async work as an explicit state machine (https://github.com/10gen/mongo/pull/64899)
* [SERVER-137537|https://jira.mongodb.org/browse/SERVER-137537] — Run ARM retry backoff on the TaskExecutor (https://github.com/10gen/mongo/pull/64900)
Batch PR:
Components
Diagnostics
Fixed Versions
9.1.0-rc0
SERVER-137536ClosedFixed
Model ARM per-remote async work as an explicit state machine
h2. Summary
Replace {{AsyncResultsMerger}}'s ad-hoc per-remote retry bookkeeping (an {{outstandingRequest}} flag plus a callback handle) with an explicit per-remote lifecycle state machine, so that retry, detach, and kill interactions are legible and individually assertable. This PR also introduces the {{tassert}}/{{LOGV2}} ids for the new transitions.
h2. Context
Part of the {{SERVER-134598}} ARM retry refactor, delivered as a stack of 8 PRs. This is PR 2 of 8, counting from the bottom of the
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137535ClosedFixed
Extract _keepRecordOfAdditionalTxnParticipants helper in AsyncResultsMerger
h2. Summary
Pull the additional-transaction-participants buffering out of {{AsyncResultsMerger::_handleBatchResponse}} into a named private helper, so the transaction-router plumbing is out of the way for the per-remote async-work refactor that follows. Pure extraction: single caller, behavior preserved exactly.
h2. Context
Part of the {{SERVER-134598}} ARM retry refactor, delivered as a stack of 8 PRs. This is PR 1 of 8, counting from the bottom of the stack.
h2. Details
The implementing PR i
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-134598ClosedGone away
Refactor ARM to clarify retry behavior
During implementation and code review of SERVER-133148, we identified an opportunity to clean up this area of the code which has proven bug-prone in the past.
The core idea is to refactor {{AsyncResultsMerger}}'s per-remote retry bookkeeping from ad-hoc flags ({{outstandingRequest}} + callback handle + retry-strategy internals) into an explicit per-remote lifecycle state machine (e.g. {{Undispatched -> Dispatched -> ResultsBuffered -> Exhausted}}), with a {{HitRetryableError}} state carrying
SERVER-137537ClosedFixed
Run ARM retry backoff on the TaskExecutor
h2. Summary
Run the ARM retry backoff wait on the {{TaskExecutor}} instead of the opCtx sub-baton while attached, so a pending backoff is no longer cut short by detaching the ARM and a backoff cancelled by {{kill()}} contributes no backoff time.
h2. Context
Part of the {{SERVER-134598}} ARM retry refactor, delivered as a stack of 8 PRs. This is PR 3 of 8, counting from the bottom of the stack.
h2. Details
The implementing PR is the source of truth for design rationale, testing, and review disc
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137999ClosedFixed
Join optimization NLJ with a sub-pipeline-projected foreign node as the outer side hits tassert
Plan HJ(NLJ(PROJECTION_DEFAULT(COLLSCAN b), COLLSCAN a), COLLSCAN c) where b's sub-pipeline $project renames the field the second join reads (b.y). The NLJ stage builder in gen_lookup.cpp does not export the renamed slot from the outer side, and the plan fails at prepare time.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-44539ClosedFixed
Re-enable "missing RWC" logging
After both SERVER-43712 and SERVER-43126 have been completed, re-enable the "missing RWC" logging (on shard/config servers) once we're sure that they won't result in logspam (or in a way which avoids that). This may also require adjusting any jstests that issue commands directly against shard/config servers.
Components
DiagnosticsSharding
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137837ClosedGone away
Complete TODO listed in SERVER-121712
There is a TODO in code owned by your team which references a resolved ticket.
The TODO appears in:
* buildscripts/modules/atlas/suites/no_passthrough_disagg_override_sharded.yml:92
You can also search for all references to the resolved ticket here:
https://github.com/10gen/mongo/search?q=SERVER-121712&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 thi
GODRIVER-4201Needs Triage
Remove legacy unified spec test runner
h3. *Context*
The last legacy unified spec tests in the [specifications|https://github.com/mongodb/specifications] repo have been converted to the new unified spec test format, leaving no more legacy unified spec tests. As a result, we no longer need the [TestUnifiedSpecs|https://github.com/mongodb/mongo-go-driver/blob/8437267056d47e3f5bbd67b9e4bc06d80bcddb0b/internal/integration/unified_spec_test.go#L175] test or any of the supporting code.
h3. *Definition of done*
* Remove the [TestUnifie
Components
Unified Test Runner
GODRIVER-4122Ready for Work
Make the GridFS API goroutine-safe
{panel:borderStyle=solid|borderColor=#21313c|titleBGColor=#21313c|titleColor=#fafbfc|bgColor=#fafbfc}
h3. *Summary*
Currently some Go Driver GridFS API methods are not goroutine-safe. While we don't explicitly document that the GridFS API is goroutine safe, it's surprising to users because the Go Driver CRUD API is goroutine-safe and there's no technical reason that the GridFS API shouldn't be goroutine-safe.
In the best case, like what happened in GODRIVER-3841, users detect that the GridF
GODRIVER-3823Ready for Work
Deprecate mongo-orchestration in favor of mongodb-runner
This ticket was split from {{DRIVERS-3335}}, please see that ticket for a detailed description.
Components
Evergreen ToolsMongo Orchestration
GODRIVER-3861Backlog
Improve Go release automation
{panel:bgColor=#fafbfc}
h3. *Summary*
Make the Go Driver release automation more reliable, easier to use, and easier to debug.
Currently the Go Driver release automation uses a series of Github Actions to perform the Go release. In the past year, some of those of those Github Actions have failed after the Git tag was created. When that happens, the release is publicly consumable by Go users but none of the other artifacts are created (e.g. the Github draft release, the SBOM files, the com
Components
Build
GODRIVER-3274Backlog
SCRAM-SHA-256 FIPS Compliance
This ticket was split from {{{}DRIVERS-2931{}}}, please see that ticket for a detailed description.
h2. Background
Go now ships with a FIPS 140-3 certified "cryptographic module" (called the Go Cryptographic Module in documentation) that can be optionally linked at build time. We want to confirm that customers building applications with the Go Driver can run their application in "FIPS 140-3 mode" and reasonably expect to pass a FIPS 140-3 compliance audit when using SCRAM-SHA-256 auth.
Go
Components
Authentication