Known Issues
5166 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-139500In Code Review
Prepare Windows cross builds for the final switchover
Prepare Windows cross builds for the final switchover in #66113. Align Python, Rust, and WASI tools with their execution platforms; preserve native Windows/WASI Python execution; support Cargo manifest copying and cleanup; share resource-compilation helpers; and propagate PDBs and static libraries. Scope compatibility-warning exceptions to affected cross-build paths.
Depends on merged #66297. This provides the build support needed before #66113 enables Windows cross-build defaults and the Everg
Components
Build
SERVER-138433In Code Review
Effective membership accessor + disagg sentinel
All functional predicates (oplog rid emission, applier rid honoring, rid-inclusive dbhash) and the hello/replSetGetStatus reporting surface need one cheap, unified membership value. Add {{getEffectiveRecordIdsGroupId(opCtx)}} (naming TBD), returning {{boost::optional<\{UUID, OpTime}>\}}:
- Attached storage: the persisted marker and cached it in memory. Loaded once at startup before replication starts. Updated transactionally at every mint, every clear, and initial-sync adoption. The cache must
SERVER-138106Open
Mint and announce the groupId at election (new-primary noop)
A groupless primary (feature flag on, latest FCV, attached storage) mints the {{groupId}} at election and announces it in the new-primary {{noop}}
h2. Details
- Mint: {{{}UUID::gen(){}}}; {{foundingOpTime}} is the new-primary noop's opTime. Both written in the same WUOW as the ([noop|https://github.com/10gen/mongo/blob/5fce3b99dea39b4f9453477819cedd935fba9a2c/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp#L646]).
- Extend the noop payload with {{groupId}} and {{{}founding
SERVER-135570Open
Add observability for collections that have an absent collection hash
* Have counters with cumulative invalidations and record why we invalidated with the time, including prepared transactions and missing contributions (tbh this should be a very rare event after we support prepared txn, so there aren't many use cases where we would have an absent hash).
* Distinguish absent hashes from unsupported collections vs. collections that should be supported but aren't.
* Matching size/count with absent hashes must not count as a successful content-hash comparison (mak
SERVER-137736Open
Parse replicated metadata checkpoint only once
SERVER-137224 can be improved by
# Parsing the first op exactly once
# Parsing the checkpoint into a usable data structure within {{isCheckpointApplyOpsEntry()}}
Combine {{isCheckpointApplyOpsEntry()}} and {{parseCheckpoint()}} into a single function called {{{}tryParseCheckpoint(){}}}:
Record the checkpoint if it was parseable. Otherwise don't. This interacts with SERVER-136501, so do that first.
SERVER-136501Open
Handle checkpoint applyOps chain in parseCheckpoint()
See [comment|https://github.com/10gen/mongo/pull/63958#discussion_r4200143626]. The checkpoint parsing code does not currently parse chained applyOps for the replicated metadata checkpoint. The write is unlikely to be split because it requires 80k collection changes in a single checkpoint, but it falls within the upper bound of allowable collections.
See also [https://github.com/10gen/mongo/pull/65018#discussion_r4351536820.] This _should_ go away if we properly handle checkpoints split acros
SERVER-139004In Progress
Measure Phase 2 checkpoint validation duration
* Measure elapsed time for merge/expected-checkpoint computation and comparison, with the measured scope clearly documented.
* Expose invocation count and total duration; we may want to expose this as part of the Grafana dashboard.
GODRIVER-4182Backlog
Gate releases on a sensible list of pre-checks
h3. *Context*
Commit [14be9fb86|https://www.google.com/url?q=https://github.com/mongodb-labs/drivers-github-tools/commit/14be9fb86%23diff-c19011a32d4cf4f04a0089ef86511e3ce1c2493d4c25697e1625f27c7abf8a13R43&sa=D&source=docs&ust=1791321208473977&usg=AOvVaw3cGGFaaTzTDaXMgoPkH5O0] in github.com/mongodb-labs/drivers-github-tools fixed a typo in the secure-checkout/action.yml that was accidentally making it possible to checkout master in the golang/pre-publish action. While it was there, actions/ch
Components
Build
SERVER-125570ClosedFixed
Pushdown $match on keyHash into $queryStats stage to avoid full store scan
h2. Overview
The {{$queryStats}} aggregation stage currently performs a full linear scan of the in-memory query stats store on every invocation, regardless of any downstream {{$match}} predicates. Pipelines of the form
re-serialize the representative shape for every entry in the store and then discard most of them via the filter — wasted CPU proportional to store size, not result size.
h2. Background
Production clusters with many distinct query shapes have been observed running this pi
Fixed Versions
9.1.0-rc0
SERVER-135371ClosedFixed
Enable optimization query knob
Fixed Versions
9.1.0-rc09.1.0-rc1026
GODRIVER-3351Backlog
Create PR for interesting fuzz tests
h3. *Context*
The fuzz test is designed to fail whenever an interesting case is found, see [here|https://github.com/mongodb/mongo-go-driver/blob/32ff39b0f816d49e2a09b1059a8f741e59c75459/.evergreen/run-fuzz.sh#L72]. This causes confusion, since it is not actually a testing error but rather just an interesting case that we should add to the test corpus.
h3. *Definition of done*
Create a script similar to [backport-pr|https://github.com/mongodb-labs/drivers-evergreen-tools/blob/master/.eve
Components
Testing
SERVER-136630Blocked
Core server migration to virtual task push-completion
This will require a migration of EngFlow workflows to use [virtual tasks|https://docs.devprod.prod.corp.mongodb.com/evergreen/Project-Configuration/Virtual-Tasks] in generate.tasks and call [virtual_tasks.complete|https://docs.devprod.prod.corp.mongodb.com/evergreen/Project-Configuration/Project-Commands#virtual_taskscomplete] from the runner instead of the current approach of generating result-fetching tasks that each allocate a host.
Once migration is complete, the old generate.tasks result
Components
Testing Infrastructure
GODRIVER-2123Backlog
Use an administrative Client in test runners
Both unified test runners currently use a single client for administrative actions (creating initial colllections, tearing down failpoints, etc.) and actual test operations. Ideally, these administrative actions would be done through a separate client that was unaffected by client options set in the test file (such as {{socketTimeoutMS}} or {{serverSelectionTimeoutMS}}). At the moment, we have to skip two SDAM tests that setup a failpoint and use a socket or server selection timeout since the fa
Components
Testing
SERVER-135194ClosedFixed
Add benchmark and jstest for query stats $match pushdown
Fixed Versions
9.1.0-rc09.1.0-rc1026
PYTHON-6284Needs Triage
Investigate automated greener build rotation slack handle + github reviewer assignment
h3. *Context*
{color:#505f79}Describe the background behind the problem.{color}
{color:#505f79}other teams on slack have @<team>-on-call or something of the like, is it possible to have this sync up with our calendar? and better yet, github PR assignments? (spec-resyncs, dependabot, uv.lock update, etc etc) {color}
h3. *Definition of done*
{color:#505f79}What must be done to consider the task complete?{color}
h3. *Pitfalls*
{color:#505f79}What should the implementer watch out for?
SERVER-137396Backlog
Add server-release as a codeowner on the v9.1 branch
server-release should be the codeowner on the v9.1 branch. See the changes made in SERVER-101869. Please also add server-release as a codeowner to src/third_party/wiredtiger/ to allow us to promote commits that were WT imports. See SERVER-117202 for an example of that change.
This depends on the DevProd "Build and Verification Branching Task", which should be linked to this ticket.
SERVER-96344Open
mongos doesn't honor a failpoint's full blockTimeMS
The behavior of {{MaxTimeMSExpired}} differs between mongod and mongos when using a blocking failpoint:
- mongod: Returns {{MaxTimeMSExpired}} after {{max(blockTimeMS, maxTimeMS)}}.
- mongos: Returns {{MaxTimeMSExpired}} after maxTimeMS only.
Repro: https://gist.github.com/prestonvasquez/d4b8193b4f67750994051ca61d8f035a
results for mongos:
results for mongod:
Failpoints should be uniform across topologies, behaving as mongod does. In particular, (for sharded clusters) drivers
Components
Testing Infrastructure
GODRIVER-3328Blocked
Make failpoints reliable in sharded clusters
h3. *Context*
Failpoints currently aren't reliable on sharded clusters for multiple reasons, including:
* {{mtest.SetFailPoint}} only sets a failpoint on one mongoS node by default. If a subsequent operation doesn't select the same mongoS, that can lead to non-deterministically failing tests because the failpoint isn't applied.
* mongoS doesn't block for the full duration specified by {{blockTimeMS}}. See SERVER-96344.
h3. *Definition of done*
* Set failpoint on every mongoS in sharded cl
Components
Testing
PYTHON-6283Blocked
Adopt shrub.py allowed_requesters support in the Evergreen config generator
h3. *Context*
PYTHON-6282 tracks the shrub.py fix for {{EvgTask}} dropping unknown keywords such as {{allowed_requesters}}. Two places in {{.evergreen/scripts/generate_config.py}} wait on it:
- {{backport-pr}} passes {{allowed_requesters=["commit"]}}; shrub.py drops it, so the generated task runs for any requester, patch builds included.
- The PYTHON-6178 review (PR #3215) split fuzzing into two variants because a single task cannot be kept off PR builds: PR builds run every task a pr-tag
Components
Infrastructure