Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-134640Needs Scheduling
Add CE support for geo queries
We currently [do not support|https://github.com/mongodb/mongo/blob/d9680f2c54db61b54cf7bc59a1028d1cfc1590ed/src/mongo/db/query/compiler/optimizer/cost_based_ranker/ce_utils.cpp#L85-L86] geo queries for CBR.
For samplingCE: seems we can trivially synthesize & evaluate a [MatchExpression|https://github.com/mongodb/mongo/blob/d9680f2c54db61b54cf7bc59a1028d1cfc1590ed/src/mongo/db/matcher/expression_geo.h]
Also delete the corresponding [comment|https://github.com/mongodb/mongo/blob/9cfd2ac21286
SERVER-134633In Progress
CE of partial index scans
We currently [do not|https://github.com/mongodb/mongo/blob/d9680f2c54db61b54cf7bc59a1028d1cfc1590ed/src/mongo/db/query/compiler/optimizer/cost_based_ranker/cardinality_estimator.cpp#L442] support CEing partial indices.
SERVER-98443ClosedDone
Add pretty_md.js to an OWNERS.yml file
jstests/libs/pretty_md.js should have the same owners as golden testing.
SERVER-114643ClosedFixed
Move jstests/libs/pretty_md.js to jstests/libs/query
We should update the codeowners for this file, as it should be owned by query (its only used for golden tests).
Fixed Versions
8.3.0-rc0
SERVER-137422Open
Change stream error events: add an explicit ThrowWithoutErrorEvent opt-out and harden the classifier
h2. Context
The {{showErrorsAsEvents}} feature converts recoverable, document-scoped failures into {{operationType: "error"}} events. Convertibility is currently determined by the denylist classifier {{{}isConvertibleChangeStreamError{}}}. As a result, a thrown {{tassert}} is still converted.
The required-image carve-out avoids conversion by clearing the recorded event identity before rethrowing.
h2. Proposed change
Add an explicit opt-out mechanism—a {{ThrowWithoutErrorEvent(...)}} ma
SERVER-138673ClosedFixed
Update PGO profiles
Updated PGO profile numbers for performance.
Fixed Versions
9.0.9
SERVER-139240Needs Verification
Exhaust stream continues after a BSONObjectTooLarge reply failure
Exhaust stream continues after a BSONObjectTooLarge reply failure
SERVER-31586ClosedFixed
Recreate PeriodicBalancerSettingsRefresher on the ServiceContext
Originally created by SERVER-30443, then removed by SERVER-31585 for issues that the ticket delineates.
Components
Sharding
Fixed Versions
4.1.2
SERVER-98371ClosedFixed
Reduce allocations when estimating number of index keys of a scan using sampling
In SERVER-98115, we implemented an initial approach to sampling estimation of number of index keys of an IndexScan. It works by invoking the BTreeKeyGenerator on each document to generate the set of index keys expressed as key strings, converting each key string to BSON, and then evaluating how many keys fall in the given IndexBounds.
We can improve the performance by avoiding unnecessary allocations of key string and BSON index keys. We can use [BSONElementIterator|https://github.com/10gen/m
Fixed Versions
8.3.0-rc0
SERVER-139074In Code Review
Add golden test for the full explain output of join-optimized queries
Add a golden test that records the explain output of join-optimized queries as users see it, so any change to the format shows up as a readable diff.
SERVER-133459ClosedDuplicate
$elemMatch of trivially true predicate misinterpreted after serialisation/deserialisation
9.0 and 8.0 disagree on the result set without the inhibit.
SERVER-136262Open
Anchored $regex with a {0,n} quantifier on a literal (e.g. /^a{0,2}$/) gets index bounds that exclude zero-repetition matches -- indexed find() silently returns fewer documents
h3. Summary
For a left-anchored \{{$regex}}, the planner derives a string prefix and scans only \{{\[prefix, prefix+1)}} in an ascending index. \{{analyze_regex::getRegexPrefixMatch()}} handles \{{*}} and \{{?}} correctly by dropping the preceding character from the prefix. It treats \{{\{}} like \{{+}} instead, ending the prefix but *keeping* the preceding character. \{{\{0,n\}}}, \{{\{0\}}} and \{{\{0,\}}} all make that character optional, so documents that match with zero repetitions fall
SERVER-136629Open
$expr comparison with null pushed into a sort-providing full index scan drops documents where the field is missing
_Found by an AI bug hunt and confirmed by a differential oracle: results differ with {{internalQueryPlannerPushdownFilterToIxscanForSort}} on and off, while the classic and SBE engines agree with each other._
h2. Problem
A {{$expr}} filter that compares a field with {{null}} using {{$lt}}, on a field that is not the leading field of an index that provides the sort, is pushed into a full index scan. In aggregation expressions a missing field sorts below null, so the comparison is true for docume
SERVER-136939Backlog
Multikeyness change during planning but before plan cache insertion causes incorrect query results
If an index becomes multikey during the planning of a query which has a stale snapshot of the non-multikey index, that query can activate its plan in the plan cache. At least in the repro attached, this results in missing/incorrect query results since we applied optimizations based on the non-multikeyness.
One idea for the fix is to encode multikeyness of the index(es) when creating and inserting the plan cache entry. If multikeyness changes, then future queries of the same shape will not hit
SERVER-139248Needs Scheduling
Move isDisagg() to a general-purpose test utility
{{isDisagg()}} currently lives in [{{oplog_truncation_util.js}}|https://github.com/10gen/mongo/blob/1fd970f73a712d3a8bb25506996e5872ec2c0097/jstests/libs/oplog_truncation_util.js#L21-L27], although detecting disaggregated storage mode is useful beyond oplog truncation tests. This makes the helper difficult to discover and requires unrelated tests to import an oplog-specific utility.
As part of this ticket, we should move the helper to an appropriately scoped shared utility and update existing
SERVER-117620Backlog
Determine pageSizeBytes from WT for join opt
We should obtain a better value for pageSizeBytes in executor.cpp when we initialize catalog stats for the cost model.
SERVER-129082ClosedFixed
Use WiredTiger's approximate leaf page count in cost calculations
Once WT-17175 lands, we can use their new API to get an estimate for number of leaf pages. We should integrate with this API and run all our (performance) tests to see how this impacts plan quality.
Fixed Versions
9.1.0-rc0
SERVER-139083ClosedFixed
Enable multikey_timestamp_transaction_large_apply_ops.js in no_passthrough_disagg_override_replsets
The test pauses secondary replication using {{stopReplProducer}} before executing a transaction large enough to produce multiple {{applyOps}} oplog entries. DSC bypasses the conventional oplog fetcher that checks this failpoint, so the test waits 10 minutes and fails with {{MaxTimeMSExpired}} before exercising its multikey timestamp assertions.
As part of this ticket, use {{pauseOplogApplication}} on the secondary when running under DSC and enable the test in {{no_passthrough_disagg_override_
Fixed Versions
9.1.0-rc09.1.0-rc1026
SERVER-137963In Code Review
Use field metadata from the dependency graph when producing type
getType should incorporate information from the field metadata:
* canFieldBeArray (removes array)
* knownToBeMissing (intersects missing)
* hasConstant (gives the type)
{$set: \{a: 1}} -> type int (hasConstant)
{$project: \{a: 0}} -> type missing (knownToBeMissing)
SERVER-138842ClosedFixed
Intern the shardIds inside the commitTransaction command/oplog entry
When building the metadata for a {{commitTransaction}} command, we currently repeat the shardId values verbatim in all sets, e.g.
ShardIds are typically 20 bytes long on Atlas.
Interning the shardId strings and only storing a positional index can reduce the storage and wire-transfer size, e.g.
Investigate this interning as a potential performance improvement.
As part of this effort, we should also investigate shortening the verbose field names inside the structure so that they become s
Fixed Versions
9.1.0-rc09.1.0-rc1026