Known Issues
5279 tracked MongoDB server issues from Jira with status, affected versions, and resolution details.
SERVER-119213ClosedDuplicate
Remove 'importCollection' command
This was an internal command for migrating serverless clusters. There are no serverless clusters anymore and this can be removed.
CSHARP-6275ClosedDone
Remove TMPDIR workaround for DRIVERS-3663 from Evergreen config
DRIVERS-3663 (mongodl's GPG check of "latest" downloads placing its gpg homedir under $TMPDIR, so the gpg-agent socket path exceeds the OS limit) is fixed upstream. Remove the {{unset TMPDIR}} workaround and its TODO from {{evergreen/evergreen.yml}} (added in CSHARP-6225).
Fixed Versions
4.0.03.next
SERVER-137607ClosedFixed
Change ownership of wiredtiger metrics files to SESI
As discussed offline, these are better owned by SESI for a few reasons. Firstly, SESI is better-placed to work with the WiredTiger team on changes to this infrastructure. Secondly, SESI has already completed the majority of SPM-4597, which largely replaced the contents of this file, so in practice they own this area anyway.
Fixed Versions
9.1.0-rc0
SERVER-137532ClosedFixed
Exclude special tests that disable test commands in multiversion suites
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-136796ClosedWon't Do
Provide an option for user to specify the order of configuration parameters to be applied
*Background:*
When one setParameter command carries several parameters, mongod applies them in alphabetical order by name, not in the order you wrote them. The parameters are collected into a
(parameters.cpp:379) and applied by iterating it (line 425). The client can't change this, and
has no option to control it.
*Context:*
Mongotune eviction policy would like to use mongod level configuration for {{wiredTigerEvictionThreadsMin}} & {{wiredTigerEvictionThreadsMax }}instead of {
SERVER-137258ClosedFixed
Register PageServerReader as a MaterializedTracker listener during its own construction
{{PaliRuntime}} currently creates the {{MaterializedTracker}} and registers the {{PageServerReader}} as a per-cell listener by hand after building it. Instead, pass the tracker to the {{PageServerReader}} when it's created and have the reader register itself, so a reader can't be wired up without its materialization listener.
Location: {{src/mongo/db/modules/atlas/src/disagg_storage/pali/pali_runtime.cpp}}
Fixed Versions
9.1.0-rc1025
SERVER-130675ClosedFixed
Remove temporary oplog scan skip when updating fast count metadata
As part of SERVER-130600, we temporarily introduced logic that skips the oplog scan portions of initializeMetadata and replicated fast count manager startup if there is no persisted valid-as-of timestamp and we have more oplog to scan than we reasonably think we can catch up on. This was done to prevent blocking advancing the valid-as-of timestamp and to prevent blocking startup. However, the downside of this is that fast size and count are likely to be incorrect when this happens.
The goal o
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137175ClosedFixed
Change when fetchOplogFromORPAndApply resets attempts
Consider changing when we reset attempts counter in fetchOplogFromORPAndApply. Currently we do so on any progress, even progress that doesn't advance the LSN. Should we change it to a stricter LSN-advancing version?
Also investigate the the difference between queued and applied oplog entries, and how we distinguish reset attempts (if entries got read from ORP and queued, but not applied, does this count as a failed attempt that we retry? That sets us back in our recovery progress as well.)
Fixed Versions
9.1.0-rc1025
SERVER-136846ClosedFixed
Unexclude requires_capped tests from no_passthrough_primary_driven_index_builds
Currently, we [exclude|https://codesearch.core-server.prod.corp.mongodb.com/10gen-mongo-master/rev/09e001b1bd95512337a5a8c7afba778f184fe63d/buildscripts/resmokeconfig/matrix_suites/generated_suites/no_passthrough_primary_driven_index_builds.yml#61] tests marked with requires_capped from the no_passthrough_primary_driven_index_builds suite. This is likely a leftover from when we didn't yet support using replicated record ids with capped collections (required for PDIB). Now that support has been a
Components
Index BuildsTesting Infrastructure
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137849ClosedDuplicate
Complete TODO listed in SLS-1724
There is a TODO in code owned by your team which references a resolved ticket.
The TODO appears in:
* src/mongo/db/modules/atlas/src/disagg_storage/replication_coordinator_disagg_storage.h:1491
You can also search for all references to the resolved ticket here:
https://github.com/10gen/mongo/search?q=SLS-1724&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
SERVER-123131ClosedFixed
Mark resharding commands as internal for query stats
We sometimes record internal operations (resharding, etc.) in query stats. These operations should be marked as internal, We should also confirm we record some types of resharding/balancing operations. Notably, user queries where mongos knows its an external client and then mongos send a query to mongod should not be marked as internal for the purposes of query stats. This query should still be considered external (or not marked as internal)
Add isInternal to the key so internal operations ar
Components
Querying
Fixed Versions
9.1.0-rc0
SERVER-137746ClosedFixed
Add isInternalOperation origin flag and propagate it between nodes
_Why:_ $queryStats cannot tell server\-owned work \(resharding, chunk migration\) from user operations, so internal shapes show up in Query Shape Insights. Inferring this from session or client state gives false positives, for example user commands that mongos forwards to shards. An explicit origin bit, set where server\-owned work starts, is precise. Its only failure mode is a false negative, which a one\-line stamp fixes.
_What:_
\* An {{isInternalOperation}} opCtx decoration.
\* Egress forwa
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137747ClosedFixed
Add isInternal to the query stats key for internal operations
_Why:_ Atlas consumers need to separate server\-owned work from customer workload in $queryStats, using {{{"key.isInternal": {$ne: true}}}}.
_What:_
\* Keys get {{isInternal: true}} only when the opCtx carries the origin flag.
\* The field is emitted only when true, so existing keys and {{keyHash}} values do not change.
\* Definition: the work belongs to the server, regardless of who triggered it. Per\-shard requests made for a user command stay unmarked, including the two\-phase write without
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137748ClosedFixed
Mark resharding operations as internal
_Why:_ Resharding runs aggregates on the donor \({{$\_internalReshardingOwnershipMatch}} and the oplog fetch\). Today these appear as unexplained shapes in customers' $queryStats and Query Shape Insights. They are the main motivating example for this ticket.
_What:_ Stamp every opCtx created by {{makeReshardingOperationContext}} \(fetcher, applier, and the donor, recipient, and coordinator services\), and the collection cloner's opCtx. These operations then report {{isInternal: true}}.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-137749ClosedFixed
Mark chunk migration batch inserts as internal
_Why:_ During a chunk migration, the recipient inserts the cloned documents. These inserts are recorded like user inserts and inflate the customer's insert stats.
_What:_ Stamp the migration batch inserter's opCtx so these inserts report {{isInternal: true}}, while concurrent user writes stay unmarked.
Fixed Versions
9.1.0-rc09.1.0-rc1025
SERVER-88023ClosedDuplicate
Create explicit multiversion tests for recordIdsReplicated:true
Make sure that across binary versions & FCV versions things like initial sync, oplog application, etc. don't break.
It should work already (since we've implemented & tested the downgrade path), but we should test it with multi binary versions.
If it looks like our explicit targeted testing is already doing a good enough job, feel free to close this ticket out.
SERVER-136712ClosedWon't Do
Enable explicit resource management in SpiderMonkey
JavaScript explicit resource management is enabled by default in SpiderMonkey ESR 153. The initial 140 -> 153 upgrade will keep it disabled so as to avoid changing existing JS execution behaviour, but we should consider enabling it as a separate change once the upgrade is complete - perhaps only for the 9.1+ releases.
NODE-7783CanceledGone away
Improvements to the Memory Provider for Vercel AI SDK
Ticket to track any improvements to the *MongoDB Atlas Memory Provider for the Vercel AI SDK* for the ABX team
AHA ticket: https://mongodb.aha.io/features/AI-84
Components
ABX
SERVER-137359ClosedGone away
Enable failover suites in POA build variant
Suites that perform failovers have been disabled with {{incompatible_with_pipelined_oplog_applier}} in the Pipelined Oplog Application build variant. We should re-enable them after SERVER-134781 / SERVER-136053 and whatever else is needed to make step up work has been implemented.
SERVER-135193ClosedFixed
Add new fast path to query stats execution
Instead of performing a full store scan, add a fast path that only searches for the query stats store entries with matching key hashes (as requested and stored via the $match pushdown optimization rule).
Fixed Versions
9.1.0-rc09.1.0-rc1025