Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-136425Needs Scheduling
Make QueryLifespan::get() threadsafe
In https://jira.mongodb.org/browse/SERVER-116447 we duplicate query settings ([https://app.graphite.com/github/pr/10gen/mongo/63905]). It would be much better to just read them directly from decoration which [https://app.graphite.com/github/pr/10gen/mongo/63864] attempted to do, but as currently the lifespan is not thread-safe we ended up with too many locks.
SERVER-139052Backlog
Audit POA test feature-flag tags and experimental variant coverage
Audit POA-specific jstests for consistent use of the {{featureFlagPipelinedOplogApplier}} feature-flag tag, distinguishing tags that control test selection from fixture settings that explicitly enable POA.
Ensure we are getting the test coverage we want from the four-hour POA experimental variant, including standalone POA tests and transaction passthrough coverage. Verify that tests actually execute rather than being filtered out or left unscheduled.
We expect to remove the experimental varian
SERVER-131934ClosedFixed
delinquent_ops.js sleep hit internalQueryJavaScriptFnTimeoutMillis under CPU pressure
in https://jira.mongodb.org/browse/BF-44896. in testDelinquencyOnShard, the sleep for 10s could extend to over 60s and hit internalQueryJavaScriptFnTimeoutMillis when cpu utilization > 90% due to 4 parallel jobs in windows-debug-suggested variant.
Fixed Versions
9.1.0-rc0
SERVER-138910ClosedFixed
Generate one failure per test for test_api_version_compatibility
In order to isolate distinct failures and prevent new ones in patch builds, evergreen must report distinct tests for every failure generated by this script.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-135518Open
Prototype MC/DC reporting + create coverage report guide
Create a POC and write up how to get MC/DC code coverage.
Also include how to interface with `bazel coverage`.
SERVER-131502In Code Review
Create a unittest utility to expect tassert failures
tassert is process fatal in testing during shutdown, so many tests currently use DEATH_TEST when expecting tasserts are triggered.
Fixed Versions
9.1.0-rc09.1.0-rc1022
SERVER-137356ClosedFixed
Allow passing fassertOnLockTimeoutForStepUpDown to replSetStepDown to override RSTL timeout
Apply Stripe patch #2: [https://github.com/10gen/mongo/tree/brad.cater/stripe-patch-2-timeout-stepdown-command]
* Specify a timeout value, which is generally configured via server param fassertOnLockTimeoutForStepUpDown, so that the timeout period can be tuned per call/stepDown rather than having to change the server params.
* This ticket is limited to configuring this additional parameter on ASC.
Acceptance Criteria: pass the fassertOnLockTimeoutForStepUpDownOverride parameter through
Fixed Versions
9.1.0-rc09.1.0-rc1025
GODRIVER-4250ClosedFixed
mongo-go-driver - PR #2715: Fix typo in comment about ErrNoDocuments
[SKalt|https://github.com/SKalt] has created [PR #2715: Fix typo in comment about ErrNoDocuments|https://github.com/mongodb/mongo-go-driver/pull/2715] in [mongo-go-driver|https://github.com/mongodb/mongo-go-driver]
Fixed Versions
2.10.1
SERVER-138365Open
Primary degrades over weeks: WiredTiger checkpoints grow to ~3 min and reads slow ~60x until restart (8.2.11, Windows)
*Environment*
* MongoDB 8.2.11 Community, Windows Server 2022, PSA replica set (primary, secondary, arbiter)
* Primary: 96 GB RAM, 12 cores, default WiredTiger cache (51 GB), journal and data on the same volume
* Workload: about 1,700 inserts, 2,000 updates and 3,900 finds per second, arriving in 16 s bursts every minute. Documents get frequent small updates ({{{}$v:2{}}} diffs) and are upserted by {{{}_id{}}}. A TTL index removes about 16M documents per day.
*Problem*
After 4 to 6 wee
Affected Versions
8.2.11
SERVER-139031ClosedFixed
Improve MakeObj code to avoid using constant literals
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-139006ClosedFixed
Don't register a write intent for an operation that has been killed
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-135295In Code Review
Log and emit a metric when a node detects that another node of the replset has gone back in time
Log and emit a metric if a member's last_xxx timestamp goes back in time through a heartbeat or replSetUpdatePosition directly from that node.
Each node keeps track of the last heartbeat response it had from another node. We use this info for sync source selection.
But we separately keep track of every node known lastWritten / lastApplied / lastDurable updated through heartbeat or the replSetUpdatePosition command, and this is not overrode if the new optime is not greater than the previous
SERVER-136941ClosedFixed
Change SBE blueprint tree to include multi range scan
Fixed Versions
9.1.0-rc0
SERVER-138513ClosedFixed
Track all min/max accumulator slots in SBE blueprint
Fixed Versions
9.1.0-rc0
SERVER-138714ClosedFixed
Make SBE loop join remap slots that that are carried over from inner side
SBE loop join can "redefine" slots that are taken from the inner side. This makes analyzing the plan hard because it's effectively a mutation. Instead, we should define new slot IDs above the inner join and remap them.
Fixed Versions
9.1.0-rc0
SERVER-138938ClosedFixed
Teach Coverity about synchronized_value locking
Coverity does not follow the lock released by synchronize()'s UpdateGuard destructor. It treats the guarded mutex as held for the rest of the enclosing function, which produces false positives.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-139046In Code Review
Add QueryPlannerAnalysis::canProvePathIsNotMissing
Adds {{QueryPlannerAnalysis::canProvePathIsNotMissing(node, path)}}, which returns true only when {{path}} in the output of {{node}} can never be missing (SBE {{Nothing}}).
Index keys store a missing value as null, so a field that comes unchanged from a covered {{IXSCAN}} is never missing. A streaming {{$group}} with a multi-field object {{_id}} needs a {{fillEmptyNull}} cluster key for every field that may be missing, because index order and sort order interleave null and missing rows. This he
SERVER-133475Needs Scheduling
Ensure priority takeover does not lead to rollback
In a recent help ticket we saw an occurrence of SERVER-118902 triggered by a priority takeover.
SERVER-118902 is a known bug where nodes can accept a write for term T after voting in term T+1 but before it selects a sync source that knows about T+1. If these writes need to be rolled back, then it can result in a scenario where a majority of nodes are in the ROLLBACK state, leading to a loss of write availability.
Step-down and failover make this unlikely or impossible by ensuring we do not