Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-136861ClosedFixed
Complete TODO listed in SERVER-130973
There is a TODO in code owned by your team which references a resolved ticket.
The TODO appears in:
* jstests/noPassthrough/query/cbr/index_seeks_point_prefix.js:22
* jstests/sharding/query/explain/explain_v3_sharded.js:64
You can also search for all references to the resolved ticket here:
https://github.com/10gen/mongo/search?q=SERVER-130973&type=Code
The next steps for this ticket are to either remove the outdated TODO or follow the steps in the TODO if it is correct. If the latter, please
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-125913ClosedFixed
Support building non-contiguous recordid ranges from queries on clustered indexes
Fixed Versions
9.1.0-rc09.1.0-rc1024
SERVER-139193Backlog
Improve samplingCE for wildcard indices
SERVER-134637 introduced support for costing index scans on hashed/wildcard/collated indices.
Currently samplingCE extracts all the keystrings (up to a limit set by a knob) and then matches each keystring to the IndexBounds in the ixscan node. For wildcard indices, this usually means extracting keystrings for all the fields in the wildcard root just to get one of them. This is slow for documents with many fields.
The following have come up as possible improvements:
* Project-away unrelat
SERVER-139245Needs Scheduling
Investigate preparation of complex plans
The subtree for each thread is now "prepared" from within the main thread, so that we can reuse the CollectionAcquisition already obtained by it. This could be not optimal if the plan preparation is costly (e.g. when compiling several projection expressions) so we could use each thread to do their own preparation.
In any case verify if the current solution (running inside the main thread) doesn't expose to deadlock scenarios (e.g. prepare() tries to use the transaction associated to the produ
SERVER-138848Open
Add integration tests for e2e testing of errors
Introduce a set of integration tests for verify that emit errors does work as we expect.
This will have nothing to do with the fsm testing effort. We need a set of ad hoc integration tests.
SERVER-137325In Progress
Implement EnsureResumeTokenPresent error-event positioning and gap materialization
h2. Deliverable
Implement {{{}ChangeStreamEnsureResumeTokenPresentStage{}}}, a subclass of {{{}CheckResumability{}}}. The stage delivers and positions the outcome decided by the detector.
h2. Behavior
* *Materialize the stashed gap:* Release the gap stashed by {{CheckResumability}} only when the exact {{ChangeStreamHistoryLost}} signal is received.
* *Build and position error events:* Build per-event error events from recoverable failures using {{{}tryBuildErrorEvent{}}}. Compare each
SERVER-137324In Progress
Implement CheckResumability error-event detection and history-loss handling
h2. Deliverable
ChangeStreamCheckResumabilityStage becomes a detector that signals rather than emits.
* *Feature flag off:* Translate {{OplogQueryMinTsMissing}} to {{ChangeStreamHistoryLost}} (286).
* *Feature flag on:* Apply the same translation, and also catch a mid-stream {{CappedPositionLost}} and rethrow it as {{ChangeStreamOplogCursorLost}} (533, retriable).
* *History loss:* The oplog scan records the earliest surviving timestamp in {{changeStreamOplogFloorAfterRollover}} ins
SERVER-31417ClosedFixed
Improve tcmalloc when decommitting large amounts of memory
tcmalloc may occasionally release large amounts of pageheap free memory to the kernel by calling madvise. This can take seconds when the amount of memory involved is many GB. A tcmalloc internal lock is held while this happens, so this can potentially stall many threads, causing widespread latency spikes.
There is no direct metric that diagnoses this (SERVER-31380 would provide that), but it can be indirectly inferred to be a likely cause from the following:
* tcmalloc pageheap free memory d
Components
Internal Code
Fixed Versions
8.0.0
SERVER-134529ClosedFixed
Lower the number of collections created in convert_to_capped_unsplittable_collections.js test in a suite with add/remove shard hook
Limit the number of threads in test suites that run the ContinuousAddRemoveShard hook so that they do not exceed the five-minute test timeout.
Each thread creates unsharded collections, and draining a shard takes at least 30 seconds per unsharded collection on that thread. In addition, convertToCapped operations may be delayed because they must serialize with the moveCollection operations issued during shard draining.
To keep the suites within the timeout, configure them to create fewer th
Fixed Versions
9.1.0-rc09.1.0-rc1024
SERVER-139171In Code Review
Reduce accumulator memory usage
Each accumulator embeds its own SimpleMemoryUsageTracker, which adds roughly 100 bytes of overhead per accumulator instance. That overhead adds up in a $group with many keys, because every group key holds one instance per accumulator. The same base class is used by accumulators that back window functions. Accumulators don't need most of what the tracker provides. They only compare their own usage against a local limit. They have no parent tracker, so they never sync with the operation-level tra
SERVER-70201ClosedFixed
Make compact killable
The long-running part of the {{compact}} command is when we actually call into {{WT::compact}}. This means there currently a {{compact}} command essentially can't be killed once it has started. WT-9636 added the ability to interrupt {{WT::compact}}, so we should make use of it here.
Fixed Versions
7.1.0-rc0
SERVER-130222Backlog
Throttle oplog application during cache eviction
Evaluate the effectiveness of throttling oplog application on maintaining majority replication lag due to slow secondaries when the ratio of dirty pages exceeds the configured eviction threshold.
SERVER-34938ClosedFixed
Secondary slowdown or hang due to content pinned in cache by single oplog batch
We only advance the oldest timestamp at oplog batch boundaries. This means that all dirty content generated by the application of the operations in a single batch will be pinned in cache. If the batch is large enough and the operations are heavy enough this dirty content can exceed eviction_dirty_trigger (default 20% of cache) and the rate of applying operations will become dramatically slower because it has to wait for the dirty data to be reduced below the threshold.
This can be triggered b
Components
Replication
Affected Versions
3.6.33.6.4
Fixed Versions
5.0.0-rc04.2.164.0.274.4.9
SERVER-139238In Code Review
Add Copybara container build and ECR publish task
Add the manual {{build_copybara_container}} task on the commit-only {{build-copybara-container}} Evergreen variant to publish pinned Copybara images through the existing build-container ECR repository and role.
{{buildscripts/copybara/container/build_image.py}} defaults to compiling and smoke testing with Copybara's supported {{help}} subcommand. The commit task then assumes the ECR write role and runs a separate publication command with fresh credentials, recording the image digest as a task a
Components
Build
SERVER-139237In Code Review
Copybara hardening
Validate configured Copybara branch targets in {{buildscripts_test}} so refs outside source history fail early.
Harden {{sync_repo_with_copybara.py}} with lease-aware container cancellation, a migration watchdog, and safe lock cleanup. Refresh GitHub credentials throughout setup and migration, recover concurrent fast-forward races, and reject divergent history.
Limit Evergreen test syncs to five commits and defer release tags until targets are reached; production syncs remain unlimited.
Valid
Components
buildscripts
SERVER-139234Needs Scheduling
Complete TODO listed in SERVER-136019
There is a TODO in code owned by your team which references a resolved ticket.
The TODO appears in:
* etc/evergreen_nightly.yml:60
You can also search for all references to the resolved ticket here:
https://github.com/10gen/mongo/search?q=SERVER-136019&type=Code
The next steps for this ticket are to either remove the outdated TODO or follow the steps in the TODO if it is correct. If the latter, please update the summary and description of this ticket to represent the work you're actually doin
SERVER-139232Needs Verification
Move required Repl tasks to the merge queue
This is documented in https://docs.google.com/document/d/1L-PNBywZ3vAx3v8XiZuSGJpmm4uzU3VSvnPKnstfFjA/edit?tab=t.0.
SERVER-139231Open
Resurrect the Repl smoke test suite
Meet the Infinite Burndown pre-PR suite goal described in [https://docs.google.com/document/d/1L-PNBywZ3vAx3v8XiZuSGJpmm4uzU3VSvnPKnstfFjA/edit?tab=t.nfvnvea4vt17].
A former Repl engineer built test analysis scripts that selected the highest value tests. We should re-run those scripts to get a new set of high value tests.
SERVER-132390Backlog
Unit test oplog truncation with oplog tailer
We lack tests that integrate size/time-based oplog truncation with the oplog tailer system. Investigate if it's possible to start both systems and test them together.
See also [thread|https://github.com/10gen/mongo/pull/59497#discussion_r3647528025].
SERVER-139145ClosedFixed
Import sls-proto: 321d2a30eca8a32259039f886e93ffb5b67b0332 with log server error code fix
Generated by the `sls-pin-bump` skill, operated by .
Bump the pinned SLS commit from {{6a55df0}} to {{321d2a3}} (94 commits) and remove the {{ERROR_CODE_LOG_SNAPSHOT_INSTALLED}} case from {{classifyDetailCode}} in {{pali/log_server_error.cpp}}. SLS changed that value to {{reserved 46}} in this range, so the automated bump (SERVER-139084) fails to compile. The switch intentionally has no default arm, so the fix compiles only at the new pin and must land together with the bump. This supersedes SE
Components
Replication
Fixed Versions
9.1.0-rc09.1.0-rc1026