Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-138631ClosedFixed
ClientThreadNiceValueTest is brittle to environments that make the default nice value nonzero
The test as written assumes a default nice of 0; {{bazel}} sometimes sets it to -15, which results in this sort of failure:
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-136925ClosedWon't Do
Bound the join plan cache trial window by the query's limit
The join plan cache measures its baseline (and, on a cache hit, runs its trial) over a fixed window of {{internalQueryJoinPlanCacheTrialWindow}} documents (default 101), independent of how many documents the query can actually return. When the query has a limit smaller than the window, the trial runs and buffers more than it needs. This increases the latency to first result in the case of limits.
We should investigate using {{min(limit, internalQueryJoinPlanCacheTrialWindow)}} (mirroring the
SERVER-137670ClosedDuplicate
Wire router synthesis stage
Wire the router synthesis stage from SERVER-137669 into a sharded cluster pipeline if the change stream is opened with the {{\{showEndOfTransaction: "whenAvailable"\}}} or {{\{showEndOfTransaction: "required"\}}} option.
Add the stage to the pipeline at the correct pipeline position (directly after the handle-topology-change-v2 stage), and add a copy of the user pipeline stages as the stage's sub-pipeline.
SERVER-138572ClosedFixed
Tassert streaming group key values are never arrays
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-136248ClosedFixed
Create basic batched index scan + fetch stages in SBE
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-135200ClosedFixed
Add namespace to progress meter during validation index traversal
Similar to SERVER-135136. Log lines can be interleaved, so have the progress meter identifies the namespace.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138280ClosedFixed
Fix hashing of SBE decimal values
This issue was found by co-pilot: some equal decimal values can get a different hash.
Proposed unit test in value_test.cpp to demonstrate it:
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138715ClosedFixed
Improve test coverage for $add in SBE
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-134237ClosedFixed
Change resharding temporary collection name to include source namespace
This may become more helpful as we start allowing multiple reshardings at the same time.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138573ClosedFixed
Add disableContinuousInternodeValidationPerCollection server parameter
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-122379ClosedFixed
Extend countNDV method to support multikey, filtered cases
Currently countNDV is only required in non-multikey cases.
To avoid duplication going forward, generalise this method for use in index seek estimation.
Fixed Versions
9.0.0-rc0
SERVER-136033ClosedFixed
Run resumable_drain_phase_multikey_two_failovers.js when doesNotSupportGracefulUnplannedStepdown
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-138293ClosedFixed
Make endOfTransaction metadata multiversion-safe
The `commitTransaction` command is defined in the IDL as strict, so older versions may not be able to understand the newly added `endOfTransactionScopes` field that the txn coordinator can send since SERVER-137904. In a mixed-version cluster (running 9.1/master for the transaction coordinator and 9.0 for a participant shard), it is possible that the receiving shard cannot parse the `commitTransaction` command request with the extra metadata sent by the coordinator. This situation is now auto-det
Components
Change streams
Fixed Versions
9.1.0-rc09.1.0-rc1025
PYTHON-6028ClosedDone
Consolidate non-IO code in _ClientConnectionRetryable into shared base class
All of the non-IO methods in {{_ClientConnectionRetryable}} can be lifted into a shared base class with some modification of type hints.
SERVER-138718ClosedFixed
Refactor persistent samples local coll tests into new file so they can be excluded from disagg suites
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-91841ClosedFixed
Repaired nodes rejoining replica set can hit invariant
If a node that was repaired via [{{--repair}}|https://www.mongodb.com/docs/manual/tutorial/recover-data-following-unexpected-shutdown/] rejoins a replica set that it was a part of (like in [this test|https://github.com/mongodb/mongo/blob/94bf4508f622ca9a986226f0a54cb529beda90cb/jstests/disk/repair_invalidates_replica_set_config.js#L121-L128]), it will go through initial sync. This is because we run repair in cases of potential data loss or corruption, and we don't want the repaired node becoming
Fixed Versions
8.1.0-rc08.0.0-rc11
SERVER-14868Open
secondary should not switch to syncTarget that has electionTime lower than its own
If secondary "knows" electionTime and it loses its syncTarget when it finds a new one it should reject it if the potential syncTarget's electionTime is lower than the one its tracking.
Components
Replication
Fixed Versions
Needs Further Definition
SERVER-20277Open
Better handling of failing long running index builds in initial sync
When a document is failed to index in initial sync, the build fails and is reattempted 9 times. For example something like the following might be seen:
Because this index takes a long time to fail, user intervention could be required (or the node will fall off the oplog) before 9 attempts expire.
Components
Replication
Fixed Versions
features we're not sure of
SERVER-14983Open
Ability to immediately mark the node as unable to service user queries
In production systems there is a common problem scenario when a particular underprovisioned mongod node becomes overloaded by a sudden load peak, causing service timeouts on the application level. In this case a plain retry strategy will add more load to the already overloaded system, frequently causing massive service degradation and prolonged periods of outage.
The ideal, yet very complicated way to gracefully handle the situation is for drivers or mongoS to detect that the node is overloaded
Components
ReplicationUsability
Affected Versions
2.6.4
Fixed Versions
Needs Further Definition
SERVER-2195Open
Change names/ips of members
Add the ability to change the names/ips for members. In the case that you have a total failure in ec2 when things come back up they will have new ips (and names).
Components
ReplicationSharding
Fixed Versions
features we're not sure of