Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-132398ClosedFixed
Store only one sample per collection
Currently JOO and CBR use different sample sizes, meaning that in order for both to run `analyze` must be run twice, once for each sample size. Investigate whether we can consolidate samples by only persisting the larger sample, and when the smaller sample is attempted to be loaded only reading that smaller number of docs from the larger persisted sample.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-117454ClosedFixed
Enable replicated collection size and count in all feature flag builds
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137473ClosedFixed
Delete buffered result documents from rejected plans after picking winning plan
When we pick the winning plan in the multiplanner, we don't free the buffered result documents from the rejected plans (see [here|https://github.com/mongodb/mongo/blob/4393af13a33ac4dd9ab9770ef1278b2774dc5cf0/src/mongo/db/exec/classic/multi_plan.cpp#L618-L650], we just move the rejected plans into a different vector). We should free them since we are no longer using them.
We shouldn't remove the results from a backup plan if one exists since we might have to use it.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-106407ClosedFixed
listIndexes with includeIndexBuildInfo:true over a timeseries collection returns OK with no indexes
If the {{listIndexes}} command is run over a timeseries collection with the {{includeIndexBuildInfo: true}} option, it will return a {{{}ok:1{}}}, but the response cursor will be empty, rather than containing all indexes as expected:
Technically the cause is that {{listIndexes}} with {{includeIndexBuildInfo}} generates a different response format:
Yet the [timeseries index translation logic|https://github.com/mongodb/mongo/blob/fb658dab8f4e485b082f1685f0c1914ab9d0dfd5/src/mongo/db/com
Affected Versions
6.0.07.0.08.0.08.2.0-rc08.1.0
Fixed Versions
9.0.0-rc0
SERVER-138330In Code Review
Productionize metric primitives for the unified metrics framework
h2. Overview
Turn the POC {{observability::}} metric primitives into production C\+\+ types for the unified metrics framework \(epic SPM\-4962\).
h2. Scope of Work
\* Productionize {{Counter}}, {{UpDownCounter}}, {{Gauge}}, {{HighWaterMark}}, {{AtomicHistogram}}, and {{Histogram}}.
\* {{AtomicHistogram}}: lock\-free {{fetch\_add}} accumulator that snapshots {{{count, sum, buckets}}}; concurrent counts are approximate; values above the last bucket boundary saturate into the top bucket. Prefer th
Fixed Versions
9.1.0-rc09.1.0-rc1025
PYTHON-6223ClosedDone
Improve Copilot automated reviews
h3. *Context*
{color:#505f79}Describe the background behind the problem.{color}
Our current Copilot reviews only use copilot-instructions.md, with no reference to CODING-STANDARDS.md. This weakens automated review quality.
h3. *Definition of done*
{color:#505f79}What must be done to consider the task complete?{color}
* Have Copilot read CODING-STANDARDS.md.
* Remove duplication between copilot-instructions.md and CODING-STANDARDS.md.
* Improve the quality of Copilot reviews.
h3. *Pit
SERVER-137676ClosedGone away
Rollback can hang indefinitely in KVDropPendingIdentReaper::immediatelyCompletePendingDrop when rolling back a timeseries collection drop
NB This is an AI-generated ticket. I read the description, and it sounds reasonable, but do send it back to Repl if it's a Repl issue.
h2. Summary
A replica set node performing rollback of a {{drop}} of a *timeseries* collection can hang indefinitely inside {{{}KVDropPendingIdentReaper::immediatelyCompletePendingDrop(){}}}'s unbounded retry loop while re-creating the collection during the rollback reconstruct phase. No timeout value in the driving test hook can recover; the node only escapes
SERVER-136115ClosedFixed
Analyze winning query plans to use non-default index scan + fetch batch size
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-134048ClosedFixed
Avoid leaking sessions on RTS
Leaking sessions due to RTS is a problem because, unlike for a real shutdown, RTS doesn't close the storage engine connection. The suspicion is that such leaked sessions are [pinning tables and preventing them from being dropped](https://jira.mongodb.org/browse/BF-45290).
Components
Storage
Fixed Versions
9.1.0-rc09.1.0-rc1022
SERVER-138568ClosedDuplicate
Handle OBJ_obj2txt failure and fix buffer sizing in x509ShortNameToOid; stop parsing unverified MONGODB-X509 principal names for cluster membership
h1. Overview
See SECBUG-6434 for the finding, the pre-auth call path, reproduction, CVSS scoring and the Minios fix contract (code anchors, invariants and three named tests). This ticket covers the engineering fix and the gaps found while triaging it.
h1. Additional Findings Not In the SECBUG Ticket
Off-by-one in the same function, so the contract's "preservation" test is red on master. x509ShortNameToOid (src/mongo/util/net/ssl_manager.cpp) resizes the buffer to the length OBJ_obj2txt r
SERVER-137388ClosedWon't Do
Update signing key to 9.2 in master
After branching the new version, master should use the next version's signing key.
NODE-7889ClosedFixed
Update libmongocrypt to 1.20.5
Update [mongodb-client-encryption|https://www.npmjs.com/package/mongodb-client-encryption] to bundle libmongocrypt 1.20.5
Components
Client Side Encryption
Fixed Versions
mongodb-client-encryption-7.2.2
SERVER-133871ClosedWon't Fix
Windows NamedPipeOutput leaks handle if writer thread creation fails
h2. Problem
On Windows, {{NamedPipeOutput}} creates a valid pipe handle in its constructor but its destructor only calls CloseHandle() when the pipe is marked open. If the object is destroyed before open() succeeds, the handle is leaked.
There are Windows and POSIX variants of {{NamedPipeOutput}}. Currently this code is only used in JavaScript and C++ tests.
h2. Impact
The handle remains open until the shell process exits.
h2. Reproduction Context
The issue is exposed when cal
SERVER-135755ClosedFixed
Update $where, idle change-stream/tailable getMore, timeseries create, and $_internalReshardingOwnershipMatch to prevent inflated delinquency metrics on query interrupt.
Currently, if a wait exits as expected, the time waiting is exempted from our delinquency metric tracking. However, when an interrupt does occur during the wait, the flag exempting the time spent in the wait is cleared and the wait time is counted toward delinquency metrics. To fix this, we should reset the delinquency stopwatch at the same location where we unset the exemption flag.
Fixed Versions
9.1.0-rc09.1.0-rc1025
PYTHON-5683ClosedDone
Spike: Investigate using Rust for Extension Modules
h3. *Context*
Our Python C Extension modules are difficult to maintain and require vigilance to avoid memory leaks. We should investigate using Rust for the existing C extensions. We could wrap the Rust [BSON crate|https://docs.rs/bson/latest/bson] for the heavy lifting in our BSON code, and do a straight port of our cmessagemodule code.
In the spike we would need to wrap a subset of the BSON types and compare the encode/decode performance of the Rust version to the CPython version.
h
Components
Infrastructure
SERVER-138598ClosedFixed
Make internalQueryFrameworkControl mod private
Removes the public IDL access modifier so the knob can't be accessed outside QueryKnobConfiguration.
Components
IDL
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-124336ClosedFixed
FCV downgrade path
Drop system.stats collections on FCV downgrade. FCVStep subclass.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137291ClosedFixed
Mark PersistentSampleDoc IDL struct as strict: false
Mark {{PersistentSampleDoc}} in {{src/mongo/db/query/compiler/ce/sampling/persistent\_sample.idl}} as {{strict: false}}, as advised by the \[IDL docs|[https://github.com/10gen/mongo/blob/master/docs/idl.md#structs|https://github.com/10gen/mongo/blob/master/docs/idl.md#structs]\], for future backwards compatibility needs.
Fixed Versions
9.1.0-rc09.1.0-rc1025
NODE-7974ClosedDone
Backport (code changes) into release branch
h3. Use Case
As a driver maintainer
I want to backport timeoutMS deprecation into v6.22.0
So that the feature is consistently unsupported in v6 and v7
h3. User Experience
* timeoutMS is a deprecated no-op in v6
h3. Dependencies
* None
h3. Risks/Unknowns
* None
h3. Acceptance Criteria
h4. Implementation Requirements
* Backport NODE-7958 into 6.22.0
h4. Testing Requirements
* CI on backport branch needs to be green (tests should either be intentionally skipped as non-a
Components
node-driver
Fixed Versions
6.22.0
NODE-7973ClosedDone
Update release tooling & CI (backport into release branch)
h3. Use Case
As a driver maintainer
I want to backport release tooling into the 6.22.0 release branch
So that a backport of timeoutMS deprecation can land
h3. User Experience
* Release tooling should be backported into the backport release branch so a release fires in the same way on main
h3. Dependencies
* None
h3. Risks/Unknowns
* A partial sync could break the release flow
h3. Acceptance Criteria
h4. Implementation Requirements
* merge the backport release tooling bran
Components
node-driver
Fixed Versions
6.22.0