Known Issues
5166 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-138231Open
Logical initial sync joins the RecordIds group
Make logical initial sync the join procedure for set-wide Replicated RecordIds : a node initial-syncing from an in-group source ends the sync holding the source's groupId, with its data in the group's rid space.
h2. Details
- New cloner stage fetching the sync source's groupId, modeled on [getInitialSyncIdStage|https://github.com/10gen/mongo/blob/5fce3b99dea39b4f9453477819cedd935fba9a2c/src/mongo/db/repl/initial_sync/all_database_cloner.cpp#L57]. The fetch must run after {{beginApplyingTs
SERVER-122753Open
Remove recordIdsReplicated from listCollections output
The per-collection recordIdsReplicated option is retired and rid replication becomes node state, so the field no longer describes anything. This ticket replaces its original motivation (the field is shard-local and misleading in sharded clusters) with removal from listCollections output.
h2. Details
- listCollections stops emitting recordIdsReplicated at active FCV (9.0+). At downgraded FCV the field keeps appearing for version-appropriate output.
- Delete the mongos-side stripping speci
PYTHON-6278Blocked
Enable all ruff D rules
h3. *Context*
{color:#505f79}Describe the background behind the problem.{color}
PYTHON-6264 tracks the enablement of the most-violated ruff D rules (D205, D400, and D401). Enabling the full D suite once those are enables will protect us against future linting violations.
h3. *Definition of done*
{color:#505f79}What must be done to consider the task complete?{color}
All ruff D rules are enabled.
h3. *Pitfalls*
{color:#505f79}What should the implementer watch out for? What are the risk
Components
Infrastructure
SERVER-138224Open
Clear the groupId on flag-off startup (C3)
A node starting with the feature flag off while holding a {{groupId}} on disk drops it before joining replication. This is the per-node disable path.
h2. Details
- Load-bearing restriction to enforce in the parameter registration: the flag must remain startup-only. A runtime setParameter flip on a live primary would stop rid emission before secondaries stop trusting rids, a silent divergence window; startup-only means a flag-off node can only emit rid-less ops after winning an election, w
SERVER-139337Backlog
Add maxBatch to SortedDataInterface
Add a `maxBatch` parameter to SortedDataInterface indicating that the cursor should internally buffer KV for subsequent calls.
SERVER-138230Needs Scheduling
Derive SLSLogEncryptorManager mirrored keystore state from snapshots
h3. Overview
The manager keeps its own copies of keystore state on the hot path (\{{KEKCache}}, \{{_activeKEK}}, \{{_activeKEKID}}), because encrypt/decrypt can't read the store. Keep them, but set them only from a snapshot, through one place, so they can't drift from the store. Part of SERVER-136037.
h3. Scope of Work
* Inventory every manager member that mirrors keystore state in \{{sls_log_encryptor_manager_impl.h}} and every place it is set.
* Add one sync function that updates the mir
SERVER-138229Needs Scheduling
Remove legacy KEK store accessors, mutators and standalone pushToWT
h3. Overview
Delete the old API so the compiler enforces snapshot + commitAndPush everywhere. Part of SERVER-136037.
h3. Scope of Work
* Remove from \{{SLSKEKStorageAPIInterface}} and both implementations: get, put, getAll, size, updateAll, getActiveKEKId / setActiveKEKId, pop, standalone pushToWT, getCMKList / setCMKList, getActiveCMKListId / setActiveCMKListId, setKeystore / getKeystore.
* Delete the unused \{{_getActiveKEKFromStorage}}.
* Update \{{HookedWTStorage}} (ticket 1 test seam
SERVER-138228Needs Scheduling
Move page_decryptor and encryption test fixtures off KEK store mutators
h3. Overview
Remove the remaining non-manager users of the old mutators so SERVER-138229 can delete them. Part of SERVER-136037.
h3. Scope of Work
* \{{page_decryptor.cpp}}: seed a prebuilt \{{SLSKeystore}} instead of calling put.
* \{{sls_encryptor_manager_test_fixture.h}} and the encryption unit tests (\{{sls_kek_storage_api_wt_test.cpp}}, \{{sls_log_encryptor_manager_*_test.cpp}}, \{{sls_phylog_decryptor_test.cpp}}, \{{sls_oplog_decryptor_test.cpp}}, \{{sls_key_management_commands_test.
SERVER-138227Needs Scheduling
Migrate CMK write paths to commitAndPush
h3. Overview
Make CMK rotation and CMK oplog apply atomic. This is the path with the lost-update, rollback-clobber and push-not-tied-to-change bugs. Part of SERVER-136037.
h3. Scope of Work
*CMK rotation (primary):* \{{updateESECMKIdentifierList}}, \{{_rotateCMKInBackgroundImpl}}, \{{_writeCMKEntry}}
* Take the snapshot in \{{updateESECMKIdentifierList()}}, where the CMK list ID is allocated, and pass it through as the \{{commitAndPush}} base.
* \{{_reEncryptAllKEKsWithNewCMKList}} takes
SERVER-138226Needs Scheduling
Migrate KEK write paths to commitAndPush
h3. Overview
Replace put / pushToWT / pop in the KEK paths with snapshot -> build next -> commitAndPush, and cache a new KEK only after its push succeeds. Part of SERVER-136037.
h3. Scope of Work
*KEK rotation (primary):* \{{rotateKEK}}, \{{_rotateKEKInBackgroundImpl}}, \{{_writeKEKEntry}}
* Take the snapshot in \{{rotateKEK()}} and derive the pending KEK ID as \{{base.size() + 1}}, not from the mirrored \{{_activeKEKID}}. Pass the snapshot through as the \{{commitAndPush}} base.
* Build
SERVER-138225Needs Scheduling
Migrate KEK store readers to KeystoreSnapshot
h3. Overview
Make every reader take one \{{snapshot()}} per operation instead of calling several getters. Part of SERVER-136037.
h3. Scope of Work
Readers in \{{sls_log_encryptor_manager_impl.cpp}} and the status commands. Each takes one \{{snapshot()}} per operation and reads everything from it:
* \{{_refreshFromKeyStorage}} (checkpoint install via \{{load_key}}, and step-up). Today it reads \{{getActiveCMKListId}}, \{{getCMKList}}, \{{get(id)}} per KEK and \{{getActiveKEKId}} separately,
CSHARP-6252Backlog
Add test coverage for top-level error handling in multi-batch bulkWrite
This ticket was split from {{DRIVERS-3670}}, please see that ticket for a detailed description.
Components
CRUD & Command-Level Operations
SERVER-139482Needs Scheduling
Add optimal DoP to parallel plan
As of the writing of this ticket, the parallel planner produces a plan to hand off to lowering/SBE and the runtime greedily uses up to maxDoP threads assuming the system is not overloaded. This ticket tracks to improve the interaction such that QO utilizes the estimated scanned data as an indicator of _how_ parallel a plan should be.
CSHARP-6297Backlog
Add prose tests that a top-level error halts client bulkWrite execution (ordered and unordered)
This ticket was split from {{DRIVERS-3699}}, please see that ticket for a detailed description.
Components
CRUD & Command-Level Operations
SERVER-136110Open
LogCaptureGuard::_isSubset early-returns incorrectly for nested objects
Some of {{LogCaptureGuard}}'s BSON subset matching functions ignore any elements following a nested object field if the object matches. For example, when checking the following (in a code block for jira formatting reasons):
It will erroneously return true because b: 1 matched the key a in both sides.
The following test added in log_capture_test.cpp would illustrate the issue:
CSHARP-6305Investigating
TLS handshake has no timeout: a frozen server node hangs new connections until TCP keepalive (~2 h on Linux)
h4. Summary
When a server node freezes (the TCP connection is established, but the server stops responding), a new connection opened by the driver can hang in the TLS handshake indefinitely. ConnectTimeout does not cover the handshake, so the application does not recover after the node is back; it recovers only after a process restart or when the OS TCP keepalive drops the socket (~2 hours with Linux defaults).
Driver: MongoDB.Driver 3.13.0 (also 3.5+ and 2.x).
Server: MongoDB 8.0, re
SERVER-136532ClosedFixed
Create benchmark to measure lookup unwind regression from SERVER-135622
Fixed Versions
9.1.0-rc0
SERVER-139344Backlog
Update record_store_bm_cursor to use maxBatch API
Update the microbenchmark testing cursor overhead to vary the maxBatch parameter for the RecordStoreAdvance variant.