Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-138023ClosedFixed
Construct BSONElementStorage directly to avoid invalid null-buffer span in binary reopen
Fixed Versions
9.1.0-rc09.1.0-rc1025
PYTHON-6182ClosedDone
TLS session reuse (4.18+) leaks ~5 KB per new TLS connection on CPython < 3.12.7 and 3.13.0 (python/cpython#116810)
## Description
4.18.0 added TLS session reuse ("Improved TLS connection performance by reusing TLS sessions across connections to the same server"). After each handshake the pool reads the session back (`_get_ssl_session()` on the sync path, `ssl_obj.session` in `_configured_protocol_interface` on the async path) and stores it in `_ssl_session_cache`.
On CPython 3.11 and earlier, 3.12.0-3.12.6 and 3.13.0, the `SSLSocket.session` / `SSLObject.session` getter leaks a copy of the session on e
Components
Connection MgmtTLS/SSL
Affected Versions
4.19.04.18.2
Fixed Versions
4.20.0
SERVER-130968ClosedWon't Do
Evaluate OTel metrics for maxEstimatedScanBytes rejection counters
h2. Overview
{{maxEstimatedScanRejected}} and {{maxEstimatedScanRejectedAndOverridden}} in {{max_estimated_scan_bytes_metrics.cpp}} are implemented as {{Counter64}}/{{MetricBuilder}} serverStatus metrics (under {{serverStatus.metrics.query.maxEstimatedScanBytes.*}}).
During review of SERVER-127688, denis631 asked whether these should instead be OTel metrics, contingent on whether the feature is planned for backport to older release branches.
h2. Findings
* No jstest anywhere in the codebase re
SERVER-138079ClosedWon't Do
Investigate linear growth of ConnectionPool::SpecificPool in Atlas cluster
In [this report|https://pages.monglify.aix.prod.corp.mongodb.com/p/rampart-wizard-napkin] derived from the linked HELP ticket, we see linear growth of {{ConnectionPool::SpecificPool}} over time. We'd like to know why this is happening.
The report was generated from shard 00-02, on the mongos.
In order of priority:
# Is this a leak related to SERVER-135020? My initial judgement says no, but it's possible.
# Do we need to change any log verbosity components on the server to find out more?
SERVER-137378ClosedFixed
Increase metadataRefreshInTransactionMaxWaitMS in retry_on_transient_error_basic.js to avoid timeouts on slow hosts
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-103659ClosedFixed
Prevent commitQuorum from being disabled
When running the {{createIndexes}} command, [setting {{commitQuorum}} to 0 disables commit quorum|https://www.mongodb.com/docs/manual/reference/command/createIndexes/#command-fields]. We should consider just implicitly upgrading this to 1 so that commit quorum is never disabled.
Fixed Versions
8.3.0-rc0
SERVER-137412ClosedFixed
Wait for cluster parameter replication in read_command_max_time_ms_repl_set.js
h2. Problem
{{jstests/auth/read\_command\_max\_time\_ms\_repl\_set.js}} checks default read timeout enforcement on secondaries without waiting for the {{defaultMaxTimeMS}} cluster parameter update to be applied on those nodes. BF\-46791 records an unexpected successful read at line 96 on the macOS nightly auth task at revision {{ca116a97f067e8e5fd3732f247167ea049381b40}}.
h2. Evidence
The failing run on 2026\-10\-12 shows the following UTC timeline:
\* 23:11:08.414: the primary finishes setting
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137976ClosedDone
[spike] Verify 10x moveCollection works with existing data
As a sanity check, verify that the proposed 10x moveCollection change will be able to import onto a new shard that has existing data.
The existing data should still be intact at the end of the moveCollection workflow.
SERVER-136231ClosedFixed
Core QO structures for parallel queries
Foundational structures for the 3-pass parallel planner: *DistributionDesc, ExchangeRouting, RequiredDistribution, ExchangeNode, ParallelismPlannerContext*
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138081ClosedCannot Reproduce
Investigate linear growth of ChunkManager::_splitAndCommitUpdatedChunkVector in Atlas cluster
In [this report|https://pages.monglify.aix.prod.corp.mongodb.com/p/rampart-wizard-napkin] derived from the linked HELP ticket, we see linear growth of {{ChunkManager::_splitAndCommitUpdatedChunkVector}} over time. We'd like to know why this is happening.
The report was generated from shard 00-02, specifically on the mongos.
In order of priority:
# Is this a leak related to SERVER-135020? My initial judgement says no, but it's possible.
# Do we need to change any log verbosity components
GODRIVER-2909Blocked
Indicate the last operation in a transaction in a $changeStream event
This ticket was split from {{DRIVERS-2660}}, please see that ticket for a detailed description.
NODE-5457Blocked
Indicate the last operation in a transaction in a $changeStream event
This ticket was split from {{DRIVERS-2660}}, please see that ticket for a detailed description.
JAVA-5075Blocked
Indicate the last operation in a transaction in a $changeStream event
This ticket was split from {{DRIVERS-2660}}, please see that ticket for a detailed description.
PYTHON-3840Blocked
Indicate the last operation in a transaction in a $changeStream event
This ticket was split from {{DRIVERS-2660}}, please see that ticket for a detailed description.
SERVER-137375ClosedFixed
Add "allowEndOfTransaction" flag to applyOps oplog entry
Adds the "allowEndOfTransaction" boolean flag to the applyOps oplog entry. The field is optional and not set by default. Currently no code sets this flag. It will later be set conditionally on the final applyOps entry of replica set and single-shard transactions if the allowEndOfTransaction parameter is set at the time of the transaction commit. An equivalent will be added to the oplog entry of sharded transactions.
Change streams can then use this to unambiguously determine the final event in a
Components
Change streamsReplication
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137407ClosedFixed
Exclude internal/external transaction classification tests from disagg override suites
SERVER-135549 turned featureFlagServerInitiatedTransactionClassification (FCV 9.1 gated) on by default and removed the flag's tag from txn_subrouter_participant_internal_external_classification.js and server_transaction_metrics_internal_external_sharded.js, so both tests started running in the disagg override suites. Those suites start DSC fixtures at the last stable FCV (currently 9.0, per SERVER-132289), which disable the 9.1-gated features both tests rely on.
Exclude the first test from no
Fixed Versions
9.1.0-rc09.1.0-rc1025
NODE-7787CanceledDuplicate
Official MongoDB Integration for OpenAI Agent SDK
{panel:bgColor=#fafbfc}
*What it is:*
Provide MongoDB-backed session persistence for the OpenAI Agents SDK in Python TypeScript. Persist conversation history in MongoDB so applications can reuse session state across runs and resume conversations without building their own MongoDB adapter.
The Python {{MongoDBSession}} implementation has already been merged into the OpenAI Agents Python SDK ([PR #2902|https://github.com/openai/openai-agents-python/pull/2902], merged April 17, 2026). The
Components
ABX
SERVER-98935ClosedFixed
$elemMatch with nested $all is not reparseable
parses into
which is not a reparseable query ("unknown top level operator: $eq. If you have a field name that starts with a '$' symbol, consider using $getField or $setField.").
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137459ClosedFixed
modify read_committed.js to not make assumptions that may be broken in disagg
In ASC, in the following block:
we expect {{initialOplogTs}} to equal the majority committed timestamp, and the test is written based on this assumption. However, in disagg, other operations such as checkpoints, can set the cluster time, meaning that the local cluster time at this point is not guaranteed to be the majority committed time. This ticket is to rewrite the test to test the same functionality without making this assumption.
Fixed Versions
9.1.0-rc09.1.0-rc1025