Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-18041Needs Scheduling
Support parallel cloning during initial sync
Components
Replication
SERVER-9ClosedDone
negated $in operator
It would be very nice to be able to use a negated version of the existing $in query operator. If I have an array of values I know I don't want to match against, I'd ideally be able to use a query along the lines of
db.mycoll.find({_id:{'not $in':[some, array]}})
Components
Usability
Affected Versions
0.9.3
Fixed Versions
0.9.3
SERVER-130268ClosedFixed
Windows SSPI GSSAPI server plugin never sets haveCred after AcquireCredentialsHandle, leaking one credential handle per unauthenticated saslStart
h1. Security Context
See SECBUG-1099 for problem description, security impact, preconditions, reproduction steps, and severity rationale.
h1. Root Cause
* src/mongo/db/modules/enterprise/src/sasl/mongo_sspi.cpp:92: SspiConnContext initialises haveCred=false.
* src/mongo/db/modules/enterprise/src/sasl/mongo_sspi.cpp:97-101: the destructor calls FreeCredentialsHandle only when haveCred is true.
* src/mongo/db/modules/enterprise/src/sasl/mongo_sspi.cpp:153-169: sspiServerMechNew calls Acq
Fixed Versions
9.1.0-rc08.3.169.0.78.0.377.0.489.1.0-rc1026
SERVER-132045ClosedFixed
Add an end-to-end Evergreen test for Windows GSSAPI/SSPI authentication
to test SERVER-130268
Fixed Versions
9.1.0-rc0
SERVER-136182ClosedDone
Run applyOps UUID-target authorization check regardless of oplogApplicationMode
SERVER-130139 added an authorization check for applyOps command ops (drop, dropIndexes, collMod, renameCollection) that resolve their target from the op's {{ui}}: the caller must be authorized for the collection the {{ui}} resolves to, not just the one named in {{o}}.
That check only runs when the oplog application mode is {{kApplyOpsCmd}} (repl/apply_ops.cpp). The mode is taken directly from the client-supplied {{oplogApplicationMode}} field of the applyOps command, and {{OplogApplication::par
Fixed Versions
9.1.0-rc08.3.169.0.78.0.377.0.489.1.0-rc1026
SERVER-136174Open
Report the throughput of a currently running index build
Once we have some way of identifying expected/acceptable throughputs of index builds, we should record some metric on actively running index builds so that we can identify index builds that are progressing slower than expected.
SERVER-136047Open
Record normalized index build duration
The ask is to record the normalized duration of an index build - a ratio of the number of the time that an index build took to complete over the number keys/bytes that it wrote. This could be recorded in a histogram metric.
SERVER-128123Needs Scheduling
[Disagg PIT Restore] Report metrics on how fast Magic Restore is running
Just to clarify, these metrics exist in FTDC. It's been hard to get them in Grafana and test them partially because 1) some of the environments that I've tried to run in weren't configured with Prometheus, 2) the current Grafana filters aren't correct and 3) PIT restore should be an offline process in that a mongo shell client can't talk to it, and I'm not sure how the metrics are transported to VictoriaMetrics.
On Grafana, we should report metrics on at least:
* The rate of oplog retrieval
SERVER-139288ClosedFixed
Adjust engine selection logic for 9.1 release
Fixed Versions
9.1.0-rc09.1.0-rc1026
SERVER-134785In Code Review
Decompose pipelined oplog applier latency into per-stage metrics
This ticket decomposes pipelined-apply latency into per-stage counters under {{serverStatus().repl.pipelinedApply}}.
This ticket should include:
* Micros counters: {{batcherWaitMicros}}, {{dispatchMicros}}, {{backpressureMicros}}, {{drainMicros}}, {{inlineApplyMicros}}, {{publishWaitMicros}}, {{workerApplyMicros}}, {{workerIdleMicros}}, {{advancerPublishMicros}}, {{advancerMultikeyMicros}}.
* Backpressure trigger counter: {{backpressure.inflight}}
* A {{workerIdleMicros}} hook in the {{P
SERVER-137201Investigating
Switch from using SDK-provided Spans to using a lightweight implementation
h2. Summary
Recording a span for a traced request costs far more than the 10% latency budget in SPM-3868's scope doc allows: on {{crud_bm}} find/insert, tracing every request added about 28%/27% CPU. About half of that was the work of recording the span itself, both in our own span bookkeeping and in the OpenTelemetry C++ SDK's per-span machinery. The other half came from that work's memory and code footprint slowing the rest of the request. This ticket reduces both:
# *Part 1, span bookkee
NODE-7979Backlog
[Mastra] Support native pre-filtering in text search
h2. TL;DR
In Mastra's MongoDB store, metadata filters on keyword (text) searches are applied only after the search finds its matches. Apply them inside the Atlas Search query instead, for {{textQuery}} and both forms of {{hybridQuery}}, on managed and bring-your-own collections.
h2. Context
Mastra is a TypeScript framework for building AI agents. Its MongoDB store searches by meaning (vector search), by keywords (text search), or both combined (hybrid search). Every search accepts a metadata fi
Components
ABX
SERVER-138209In Code Review
Remove support for persisting chunk samples
We will only permit persisting random samples, since chunk samples are lower quality and with persistence the time it takes to generate a sample isn't as important as quality.
Remove all metadata, logic, validation, etc. that support persisting chunk samples.
Fixed Versions
9.1.0-rc09.1.0-rc1026
SERVER-139094ClosedDone
Update Homebrew Formulae for New Major Patch Release (9.0.7)
Update Homebrew Formulae to support new Major Patch Version 9.0.7
SERVER-139221ClosedFixed
Make batched ixscan+fetch fallback to scalar in classic
Fixed Versions
9.1.0-rc09.1.0-rc1026
NODE-8012Backlog
[Mastra] Define how MongoDBVector refreshes cached index state after external changes
h2. TL;DR
Each {{MongoDBVector}} instance caches where an index lives and which fields its search index can filter on. It doesn't notice when another process or a database admin changes them, so it keeps using stale values until it restarts. Choose and document a refresh strategy.
h2. Context
{{MongoDBVector}} keeps three in-process caches:
* {{indexTargets}}: the logical index name's collection, search index names, and write policy, read from the registry.
* {{declaredFilterPaths}}: the fields
Components
ABX
NODE-8011Backlog
[Mastra] Reject bring-your-own documents whose _id isn't a string or ObjectId
h2. TL;DR
Query results turn every document {{_id}} into a string. For {{_id}} types other than string or ObjectId, that string no longer matches the stored value, so lookups by that {{id}} silently find nothing. Fail with a clear error instead.
h2. Context
With a bring-your-own (BYO) collection, {{query()}} returns each document's {{_id}} as the string {{id}} field of the result, using {{toString()}}. Mastra matches an {{id}} back to a document as a string, and also as an ObjectId when it's a
Components
ABX
NODE-8010Backlog
[Mastra] Verify that concurrent registrations can't retarget a MongoDB vector index
h2. TL;DR
Two processes registering the same index name at the same time must not overwrite each other's mapping. {{createIndex()}} already guards against this. Verify that guard against a real deployment and close the remaining gap in {{createSearchIndex()}}.
h2. Context
{{MongoDBVector}} records each index in a registry collection, {{__mastra_vector_indexes__}}. Each entry maps a logical index name (for example {{precedents}}) to the collection and search indexes behind it. A call reads the e
Components
ABX