Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-135633ClosedFixed
Fix flaky updateLookup test
Fix flaky test `jstests/sharding/query/change_streams/change_stream_read_preference.js`.
The test was previously flaky and could non-deterministically fail with an error message such as follows ([example failed patch run](https://parsley.corp.mongodb.com/test/mongodb_mongo_master_enterprise_rhel_8_64_bit_dynamic_classic_engine_burn_in:sharding_gen_enterprise_rhel_8_64_bit_dynamic_classic_engine_0_patch_94c253e337508c29d836607582fdac497aada7ec_6ab5014f17a7e50007bab825_26_09_24_10_54_24/0/b77c7
Components
Change streams
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.7
SERVER-131339ClosedFixed
Use custom py_binary for resmoke instead of py_test to anticipate coverage impl
Fixed Versions
9.1.0-rc09.0.7
SERVER-136502ClosedFixed
Abandon IndexBuildsCoordinator::onStepUp if the stepup is abandoned
IndexBuildsCoordinator::onStepUp acquires a lock [in a loop|https://github.com/10gen/mongo/blob/022869cbd9f2c67ee6cf0a275ad2b92a49f64926/src/mongo/db/index_builds/index_builds_coordinator.cpp#L2271-L2284] to make sure we are primary. This loop starts before we are guaranteed to be a writable primary which necessitates the loop. If the stepUp is cancelled then this loop is not able to complete
Subsequent stepups need to [wait for the previous stepUp to end|https://github.com/10gen/mongo/blob/d
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.7
SERVER-136195ClosedFixed
Add test tags
Add IFR flag and FCV to test tags. This is not required for master, because the flag is default on here and the last reachable FCV is already 9.0, but the changes are going to be backported to 9.0 where this is still required.
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.7
SERVER-129849ClosedFixed
Fix mongod shutdown hang waiting for search task executor
Fixes a `mongod` shutdown hang where idle `$search`/`$vectorSearch` cursors keep the mongot task executor alive, blocking executor teardown. Disposes idle cursors during shutdown (before the executor wait) and bounds the wait as a backstop.
Fixed Versions
9.1.0-rc09.1.0-rc10248.3.169.0.7
SERVER-133744ClosedFixed
Tolerate file shrinkage in SimulateCrash hook's copy_file
Currently, {{SimulateCrash.copy_file}} ({{{}buildscripts/resmokelib/testing/hooks/simulate_crash.py{}}}) stats a file once, then copies it via a {{sendfile}} loop up to that stat'd size. If the file shrinks between the stat and the end of the copy - e.g. a background WT thread hadn't fully parked when the node was paused - {{sendfile}} returns 0 early and the hook raised unconditionally, aborting the whole test run over what is only a best-effort diagnostic dbpath snapshot. Update SimulateCrash.
Fixed Versions
9.1.0-rc09.1.0-rc10229.0.7
SERVER-137293ClosedFixed
Remove maxTimeMS restrictions on PBRT tests
This change removes arbitrary `maxTimeMS` limits from the PBRT regression tests that have previously led to timeouts in slow build environments.
This is a test-only change.
Components
Change streams
Fixed Versions
9.1.0-rc09.1.0-rc10259.0.7
SERVER-109479ClosedFixed
Core analysis task is not aware of the fallback download links for multiversion binaries/debug symbols
It assumes the multiversion downloads file contains "evg_urls_info" with download links.
See [https://github.com/mongodb/mongo/blob/master/buildscripts/resmokelib/hang_analyzer/extractor.py#L597]
Instead, if "evg_urls_info" is empty it should fallback to
binaries = version_downloads["release_urls_info"]["urls"]["binary"]
debug_symbols = version_downloads["release_urls_info"]["urls"]["debug_symbols"]
It's also worth re-visiting whether this logic has to be re-implemented
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.7
SERVER-136499ClosedFixed
Rename task selection tags
Rename these Evergreen tags:
release_critical -> all_variants
non_deterministic -> non_required
Components
Testing Infrastructure
Fixed Versions
9.1.0-rc09.1.0-rc10248.3.169.0.78.0.37
SERVER-134625ClosedFixed
Enforce correct query shape output from extensions to support reparseability
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.7
SERVER-134495ClosedFixed
make keys_collection_non_deprioritizable.js less timing sensitive
https://jira.mongodb.org/browse/BF-45960 hit timing issue under stress in window variant. Need to modify the test implementation to make it more deterministic and less timing dependent
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.7
SERVER-131338ClosedFixed
Refactor resmoke.bzl to anticipate custom py_binary
Fixed Versions
9.1.0-rc09.0.7
SERVER-134356ClosedFixed
Reduce SlotBasedStageBuilder::buildGroup stack frame and max number of DocumentSource stages pushed to SBE
BF-46094 exposed a stack-overflow risk in recursive SBE aggregation lowering on ARM64 after the SERVER-110427 PGO profile refresh.
The failing test constructs a legal 1,000-stage aggregation pipeline. Each $group is lowered through SlotBasedStageBuilder::buildGroup(), which recursively calls build() for the child. The PGO/BOLT layout change appears to move the optimized binary across the stack-use threshold; the failure surfaces in computeChildReqsForGroup()/PlanStageReqs::setResultObj() as a
Fixed Versions
9.1.0-rc09.1.0-rc10248.3.169.0.7
SERVER-134475ClosedFixed
Avoid activate_version_tasks race in activate_task.py
Concurrent activation requests for the same version can race: another request may materialize and activate the task first, and Evergreen returns 400 for activation requests with nothing left to activate.
Rather than pre-checking activation state (which narrows the TOCTOU window instead of closing it), call `activate_version_tasks` unconditionally and handle the 400 after the fact: re-look up the task and only swallow the error once a fresh lookup confirms it is activated. Non-400 errors and gen
Fixed Versions
9.1.0-rc09.1.0-rc10229.0.78.0.37
SERVER-135517ClosedFixed
Skip metadata consistency check if snapshot predates fcv change
On secondaries if the CMC timestamp is the secondary's latest timestamp it can happen that we take a metadata snapshot before fcv change which causes metadata inconsistency on auth shard enabling
Fixed Versions
9.1.0-rc09.1.0-rc10249.0.7
SERVER-134954ClosedFixed
Fix race in NetworkInterfaceTest.TimeoutDuringConnectionHandshake
Please see the comments in BF-46335 for details.
Affected Versions
8.3 Required
Fixed Versions
9.1.0-rc09.1.0-rc10248.3.169.0.7
SERVER-125754ClosedFixed
Reduce SessionRuntimeInfoMap lock contention
To reduce lock SessionRuntimeInfoMap need to improve the read performance for session scanning. This can be done in a few ways. Partitioning, using a LockFreeReadList, and ensuring more operations are read only rather than read/write.
Fixed Versions
9.1.0-rc09.0.48.3.16
SERVER-134714ClosedFixed
Fix potential std::terminate in MovePrimaryCoordinator
(copilot)
[_runImpl|https://github.com/10gen/mongo/blob/4c929333f113e6d05cc00a3620846fecea4d0344/src/mongo/db/global_catalog/ddl/move_primary_coordinator.cpp#L158] is noexcept, but this newly hoisted operation-context setup now runs before an ExecutorFuture can capture failures. makeOperationContext() applies forwardable metadata (including token parsing) and is not noexcept; if it throws, the process calls std::terminate instead of returning an error to the coordinator retry machinery. Keep th
Fixed Versions
9.1.0-rc09.0.29.1.0-rc10249.0.48.3.16
SERVER-132904ClosedFixed
Make isShardedTimeseries test helper resilient to step downs
SERVER-119572 removed the overrides from the find in the viewless timeseries helper. This allowed it to run with snapshot read concern without being excluded from suites that modify the read concern, but it also removed the network error retries in stepdown suites. The helper already retries on HostUnreachable, but would be more resilient if it retried on all retriable errors.
Fixed Versions
9.1.0-rc09.0.0-rc48.3.16
SERVER-114507ClosedFixed
$out targeting moved unsharded collection should create the new collection on the same shard
Scenario:
# Shard A is the primay shard for {{db}}
# Unsharded {{db.coll}} has been placed on shard B (via {{{}moveCollection{}}})
# {{$out}} writing to the same collection db.coll creates it on shard A (that is the primary shard) rather than shard B (that is the shard it used to live on).
This ticket is to ensure that {{$out}} recreates the unsharded collection on the same shard as it existed before.
Affected Versions
8.0.08.2.0
Fixed Versions
9.1.0-rc09.0.48.3.16