Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-112104Backlog
Split up sdk/aggregation_stage.h into separate files
Create a new aggregation_stage/ subdirectory in sdk/ and split sdk/aggregation_stage.h into parse_node.h, ast_node.h, stage_descriptor.h, and logical.h.
SERVER-111921Backlog
Move vtable definitions to header files in extensions API
The vtable definitions in files `sdk/extension_status.cpp` and `sdk/byte_buf.cpp` can be moved to the corresponding header files and defined as a constexpr. See [https://github.com/10gen/mongo/pull/41951#] for an example of some of these changes. Confirm that no other vtable definitions are in source files.
SERVER-136053ClosedFixed
Stop POA dispatch after shutdown-related abandonment
There is a bug where we return from {{_dispatchOps()}} on shutdown-related failures, but {{_run()}} continues consuming batches. We should propagate the shutdown result back to {{_run()}} and stop the dispatch loop instead.
In addition to fixing the bug, this ticket should also add regression coverage that fails only the first inline op with a shutdown error and verifies that subsequent work is not attempted and {{lastApplied}} does not advance past the failed op. Also cover the failed-drain
Fixed Versions
9.1.0-rc09.1.0-rc1025
GODRIVER-3144Backlog
Use error “Cause” to deliver better error messaging when supported.
Per a [Slack thread|https://mongodb.slack.com/archives/C7WJZNUTA/p1709133917284529]:
This ticket proposes using Go’s new “Cause” APIs to augment “context canceled” and “context deadline exceeded” messages with more detail. For example, what canceled the context? Or, what was supposed to happen before the deadline passed?
(For reference, Mongosync plans to implement something similar in REP-3632.)
NB: While the Cause APIs are new in Go 1.20, build tags would allow their use without break
Components
Error Handling
SERVER-138999Needs Scheduling
Investigate if getField iteration can be done faster
Since we bound check on construction, can we skip some bound checking on iteration
SERVER-137310ClosedDuplicate
Optimize clients of untrusted to use trusted construction
Optimize existing clients of untrusted construction to use trusted construction in scenarios where the type, fieldname length, and value size are previously evaluated.
This can be conditional on meeting the check-in criteria.
Or can be evaluated post check-in
SERVER-138830Needs Scheduling
Complete TODO listed in SERVER-112325
There is a TODO in code owned by your team which references a resolved ticket.
The TODO appears in:
* docs/change_stream_events.md:242
* jstests/aggregation/expressions/null_chars_string_input.js:410
* jstests/auth/lib/commands_lib.js:9277
* jstests/libs/query/change_stream_rewrite_util.js:251
* jstests/libs/query/change_stream_rewrite_util.js:300
* src/mongo/db/exec/agg/change_stream_check_topology_change_stage.h:20
* src/mongo/db/exec/agg/change_stream_handle_topology_change_stage.cpp:58
* sr
SERVER-138652Needs Scheduling
Test nothing bad happens if streaming group feature flag is ON for some mongod but OFF for another mongod
Test that nothing bad happens when {{featureFlagSbeStreamingGroup}} is on for one mongod and off for another, as during a partial IFR rollout. The flag is not serialized on outgoing requests, so each mongod reads its own value.
* Use a replica set whose primary and secondary disagree (both directions), reading from each, plus a sharded variant with shards that disagree.
* Expect identical results everywhere, {{streaming_group}} only where the flag is on, and no errors, including on {{getMore}}.
SERVER-138585Needs Scheduling
Support non-simple collation for streaming group keys
Streaming group eligibility requires the simple collation ({{query.getCollator()}} in {{eligibleForStreamingGroup}}). Support non-simple collations when the index and query collations match.
* {{StreamingAggStage}} already takes a {{collatorSlot}}; check that {{buildStreamingGroup}} passes it as hash {{$group}} does.
* {{computeSortsForScan}} already drops string sorts when the collators differ, so {{providedSorts()}} should enforce the match; verify it for object and compound keys.
* Add planne
SERVER-138571Needs Scheduling
Test no false positive in plan cache with arrayness for streaming group
Test that a cached plan with a streaming {{$group}} is not reused once the arrayness it relied on no longer holds. Arrayness is not in the plan cache key; the cache is cleared on multikey changes, index builds and drops, and validator changes.
* Add an aggregate version of {{plan_cache_path_arrayness_multikey_invalidation.js}} with a {{streaming_group}} plan: cache it, make the key an array, and expect an uncached plan without {{streaming_group}} and correct results.
* Known gap, to decide separ
SERVER-138570Needs Scheduling
Test path arrayness invalidation with streaming group
Test that a streaming {{$group}} never returns wrong results when a key path that path arrayness says cannot be an array becomes one mid-query (e.g. an index becomes multikey). Model it on {{jstests/noPassthrough/query/yield/traversef_elision_invalidation.js}}.
* Pause a {{streaming_group}} plan with {{setYieldAllLocksHang}}, make the index multikey, resume; expect {{QueryPlanKilled}} from the {{PathArraynessChecker}} and a bump of {{query.pathArrayness.queriesFailedDueToInvalidation}}.
* Cover
SERVER-138743ClosedFixed
Allow initial sync when node has existing data
Allow an escape hatch to pass through the check added in SERVER-91841.
This will be for custom builds only to temporarily fix unsupported workflows and not exposed to general customers
Fixed Versions
8.0.39
PYTHON-2943Backlog
Socks5 Proxy Support
This ticket was split from DRIVERS-1357, please see that ticket for a detailed description.
Components
Spec Comp
SERVER-137255Open
Report on test infrastructure timeouts / failure occurrences for test hardening via retries
One semi-common occurrences of server security BFs is infrastructure failures (spawned hosts time out or are not accessible, OS packages are updated and cause failures due to OpenSSL mismatches, various python libraries being out of date, etc.).
This ticket represents a sprint (2 weeks) worth of work to investigate this class of failure and generate a team proposal on possibilities to reduce or eliminate these kinds of no-signal BFs from the team by any means possible.
SERVER-139019Backlog
Add boolean on StreamingAggNode QSN that marks whether or not the underlying QuerySolution is a covered ixscan or a fetch plan
SERVER-95560ClosedFixed
Majority secondaries can enter rollback state during a failover where the old primary is frozen momentarily
We have observed two cases of failover on our mongo setup running v4.4.9 where majority secondaries enter rollback state. Chaining is disabled on our setup. We then attempted to reproduce this scenario on v7.0 using JS tests and believe the bug still exists.
Below is a rough sequence of events that can lead to rollback and the associated JS test is attached as a patch file . Note that we have sleeps added in the source code to help better simulate what we saw on our setup.
* Old primary is
Affected Versions
4.4.97.0.11
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137357ClosedFixed
DSC/ASC discrepancy: DSC does not use fassertOnLockTimeoutForStepUpDown during RSTL acquisition
ASC makes use of global server parameter fassertOnLockTimeoutForStepUpDown to trigger timeouts during RSTL acquisition. DSC does not access this value; investigate why and if there is any benefit to adding this timeout mechanism to DSC.
Acceptance Criteria: Make a decision on whether to utilize the timeout server parameter in DSC.
- If yes, consider threading the additional parameter fassertOnLockTimeoutForStepUpDownOverride through the DSC code to override the existing server parameter whe
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-129800Open
Add tests for SERVER-114846 _doReplSetReconfig documentation
[{{_doReplSetReconfig}}|https://github.com/10gen/mongo/blame/5ce698e22b69702aa9cff9f0fcca904ea2a21476/src/mongo/db/repl/replication_coordinator_impl.h?plain=1#L1793-L1796] takes in two boolean arguments: {{force}} and {{skipSafetyChecks}}. SERVER-114846 added documentation that describes which behaviors will and will not be activated depending on which of the four possible permutations were given ({{!force && !skipSafetyChecks}}, {{!force && skipSafetyChecks}}, {{force && !skipSafetyChecks}}, {{
SERVER-135612Open
Make the steady-state OplogWriter's batch limits runtime-tunable
{panel:title=About this ticket|borderColor=#ccc|titleBGColor=#f4f5f7}
Drafted by Claude, working under my direction, from the investigation in SERVER-134234. I reviewed the findings and the numbers.
{panel}
h2. Problem
The steady-state {{OplogWriter}}'s batch limits are compile-time constants ({{src/mongo/db/repl/oplog_writer_batcher.cpp:19}}):
{{replBatchLimitBytes}} and {{replBatchLimitOperations}} look like the corresponding knobs but are *not* wired to this path -- they feed {