Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-134533ClosedFixed
Starting shard draining during movePrimary can cause metadata inconsistencies
The configsvrCommitMovePrimary command throws ShardNotFound if the destination shard is found as draining. Because it does this prior to checking if the command had previously succeeded, this can cause the following problematic scenario:
# Start movePrimary from shard0 to shard1
# Commit the movePrimary to the config server via [configsvrCommitMovePrimary|https://github.com/10gen/mongo/blob/2a943eb0e55e5030ef5302b2d45bc70c7d688f6d/src/mongo/db/global_catalog/ddl/move_primary_coordinator.cpp#
Fixed Versions
9.1.0-rc09.0.29.1.0-rc10229.0.48.3.16
SERVER-131779ClosedFixed
Add linux the kernel 7.0.14 exemption for ubuntu
The linux kernel on ubuntu version number are different from upstream. This makes our graceful exit not properly working on ubuntu specific kernels. Our graceful exit should check for those ubuntu kernel and ensure we can start on those compatible kernels.
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.48.3.168.0.37
SERVER-135060ClosedFixed
Track pinned host memory in legacy MozJS engine
The legacy MozJS engine doesn't keep track of pinned host memory of the underlying BSON documents, meaning memory use can grow far above the intended threshold before a GC is triggered, potentially causing an OOM. This does not affect 9.0/master since the legacy MozJS engine is deprecated, but can still be a problem for 8.3 and older.
Affected Versions
7.0.08.0.08.3.0
Fixed Versions
9.1.0-rc09.1.0-rc10248.3.168.0.37
SERVER-127898ClosedFixed
Retry CSRS demotion related InvalidOptions failures in replica_sets_transition_to_csrs suite
In {{replica_sets_transition_to_csrs}} suite during the demotion from CSRS to replicaset phase write operations might fail with
this is a known and accepted issue (by product) as a side effect of SPM-4300
though this introduces flakiness in the evergreen so we should set up a retry hook in the suite if the given error code and error message appears (preferably during write operations only)
Fixed Versions
9.1.0-rc09.1.0-rc10229.0.48.3.16
SERVER-128131ClosedFixed
Update enterprise-ubuntu2204-jepsen to specify itself as a compile variant
Currently it shares a compile_variant name with enterprise-ubuntu2204-64, and this causes bad bep files to be uploaded in BF-43487.
More worrisome than the BF are that this could lead to the wrong binary being plucked in nightly tests in general?
Fixed Versions
9.0.0-rc08.3.168.0.377.0.48
SERVER-132922ClosedDone
WaitForTicketDeadlineBetweenTimeoutWindows unit test timing issue under stress
WaitForTicketDeadlineBetweenTimeoutWindows unit test is used to cover the wall clock timing behaviour, but when physical system running the unit test is under cpu contention, the timing could shift drastically. Removing the timing check would lose the timing scenario coverage.
Fixed Versions
9.1.0-rc09.0.48.3.16
SERVER-120633ClosedFixed
Investigate tassert during subplanning_during_drop_index execution in the CBR suite
Fixed Versions
9.0.0-rc08.3.16
SERVER-123655ClosedFixed
Make the the findTimeseriesConfigCollectionsDocument() JS test helper function resilient to connection errors
Fixed Versions
9.0.0-rc08.3.16
SERVER-132857ClosedFixed
Adapt kernel compatiblity error message to new known fixed versions
Fixed Versions
9.1.0-rc09.0.48.3.168.0.37
SERVER-132359ClosedFixed
Bump db contrib tool version
Fixed Versions
9.1.0-rc09.0.48.3.168.0.37
SERVER-132566ClosedFixed
Fix NetworkInterfaceMock cancellation race that can hang transaction coordinator tests
Unit tests that use `NetworkInterfaceMock` can hang indefinitely when an alarm is canceled.
In CI this surfaced as `db_s_shard_server_part3_test` timing out and being killed
(BF-45092 / BFG-3474784), with the stack stuck in `SharedStateBase::wait` under
`TransactionCoordinator::onCompletion()`.
The cause was a missed wakeup in `NetworkInterfaceMock::setAlarm()`. The `onCancel` callback
did not resolve the alarm's promise; it only recorded the alarm id in `_canceledAlarms`, leaving
resolu
Fixed Versions
9.1.0-rc09.0.48.3.16
SERVER-131396ClosedFixed
Propagate stepdown data to CheckMetadataConsistencyInBackground for MalformedTimeseriesBucketsCollection skip
With viewful timeseries, it's a known issue that a stepdown/kill can leave a system.buckets collection without the timeseries view, so [in the CheckMetadataConsistency helper we skip the inconsistency in stepdown suites (if runningWithShardStepdowns is set)|https://github.com/mongodb/mongo/blob/19a7dfe09834f83d72c1228cb3d8cf6b10510032/jstests/libs/server_catalog/check_metadata_consistency_helpers.js#L64-L73].
However the CheckMetadataConsistencyInBackground hook runs in a separate shell
Affected Versions
8.0.08.2.09.0.0-rc08.3.09.1.0-rc0
Fixed Versions
9.1.0-rc09.0.48.3.16
SERVER-135526ClosedFixed
[v8.3] Remove references to initialSyncWaitForSyncSourceLastStableRecoveryTs
SERVER-113184 was originally backported to v8.3 but then partially reverted. There are still references to some code related to SERVER-113184 on v8.3, so we should remove them as we are no longer implementing the solution that was originally proposed and implementing.
Components
Replication
Fixed Versions
8.3.16
SERVER-131103ClosedFixed
Add RHEL10 s390x/ppc release variants
Fixed Versions
9.1.0-rc09.0.0-rc18.3.168.0.37
SERVER-134438ClosedFixed
MovePrimary stucks if the target is a draining node
If the MovePrimary restarts because of a stepdown (so the phase is not unset) then in the next run [it checks again if the target node is draining|https://github.com/10gen/mongo/blob/332ea27ae1086387e51926edd7dfc192c34dd12f/src/mongo/db/global_catalog/ddl/move_primary_coordinator.cpp#L196]. If during the stepdown the target started to drain, the check fails, but since the [mustAlwaysMakeProgress|https://github.com/10gen/mongo/blob/332ea27ae1086387e51926edd7dfc192c34dd12f/src/mongo/db/global_cata
Fixed Versions
9.1.0-rc09.0.29.1.0-rc10229.0.48.3.16
SERVER-134077ClosedFixed
Query stats supplemental metrics are only recorded when the initial batch exhausts the cursor
Supplemental query stats metrics (e.g. the JoinOptimization section) only appear in the recorded query stats entry when the query's initial batch exhausts the cursor. With the default 101-doc batch plus getMores, the entry has no {{supplementalMetrics}} at all, even though the query ran join optimization.
Repro: in {{{}jstests/noPassthrough/query/joins/query_stats_join_optimization_metrics.js{}}}, remove the {{{batchSize: 100000}}} option from any aggregate, {{stats[0].metrics.supplementalMe
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.48.3.16
SERVER-124529ClosedFixed
find command with $elemMatch projection on a DBRef field fails to reparse for query stats
See linked ticket for details
Fixed Versions
9.0.0-rc08.3.16
SERVER-134025ClosedFixed
Preserve primary build command for package provenance
## Summary
Keep the normal .bazel_build_invocation for ordinary build artifacts while writing task-specific provenance invocation files from the primary Bazel command. Archive and release package provenance now use the dedicated invocation, and crypt provenance uses its own file; stale dedicated files are removed before each compile.
## Testing
- bazel test //evergreen:bazel_evergreen_shutils_test //buildscripts/tests:test_package_test_provenance
- bazel run format (attempted; the formatter w
Components
Build
Fixed Versions
9.1.0-rc09.1.0-rc10229.0.48.3.168.0.37
SERVER-135173ClosedFixed
update fuzzer has duplicate --useEsModules command-line option
SERVER-116433 was merged without one commit , resulting in the update fuzzer on 8.3 failing because the --useEsModules command-line option was passed twice.
This new ticket is being created so that the fix PR, https://github.com/10gen/mongo/pull/60782 can be merged. A previous attempt to merge this fix was using a SERVER ticket that was already closed, which caused the bot to reject it.
Fixed Versions
8.3.16
SERVER-126463ClosedFixed
Collection cloner with function_ref has memory bug
We have this [piece of code in collection_cloner.cpp|https://github.com/mongodb/mongo/blob/7f59a42e61116840b38a7df540404dad21298ebd/src/mongo/db/repl/initial_sync/collection_cloner.cpp#L565-L569]:
This is problematic as described in SERVER-121669, because the temporary that the function_ref is bound to is destroyed right after that line.
Fixed Versions
9.0.0-rc08.3.168.0.37