-
Type:
Bug
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: 9.0.0-rc0
-
Component/s: Catalog
-
None
-
Catalog and Routing
-
ALL
-
🟦 Shard Catalog
-
None
-
None
-
None
-
None
-
None
-
None
To list all collections in the cluster, collectionless $listClusterCatalog first gets a list of DBs and then sends individual listCollections commands for each DB.
Â
This means that it doesn't see a consistent state and in particular, like SERVER-118421, $listClusterCatalog may scan the databases in "A" → "B" → "C" order but while it's scanning "B", collection "C.coll" can get renamed to "A.coll". This causes $listClusterCatalog to entirely miss a collection in its output
Â
Run the attached repro FSM in concurrency_replication.
- depends on
-
SERVER-98686 Add support for snapshot read concern in $internalListCollections
-
- Blocked
-
- is related to
-
SERVER-97061 $listClusterCatalog doesn't see a consistent state between the config catalog and the shard/local catalog
-
- Blocked
-
-
SERVER-118421 Catalog changes during setFCV can race with rename across DBs
-
- Closed
-
- related to
-
SERVER-123066 Viewless timeseries being renamed across DBs may be missed by downgrade
-
- In Code Review
-