-
Type: Bug
-
Resolution: Duplicate
-
Priority: Major - P3
-
None
-
Affects Version/s: 4.4.0, 5.0.0-rc0
-
Component/s: None
-
None
-
ALL
-
Query Optimization 2021-06-28
As outlined in SERVER-57483, the serialization code isn't appropriate as a mechanism to clone this stage. We should investigate a more robust way to distinguish the two use cases of serializing to the shards and serializing for being copied.
The impact of this problem is quite large. It means that your performance of a non-correlated $lookup might degrade when you move from a replica set to a sharded cluster.
- duplicates
-
SERVER-57168 pipeline->clone() breaks on pipelines with union stage
- Closed
- is caused by
-
SERVER-45538 Implement and test shard version retry logic for $unionWith sub-pipeline
- Closed
- is related to
-
SERVER-57483 Results from $lookup stage are not cached for a resharding operation's sub-pipeline
- Closed
-
SERVER-30399 Add caching for $lookup non-correlated sub-pipeline prefix
- Closed
- related to
-
SERVER-57705 DocumentSourceSequentialCache can wind up in a suboptimal location due to optimizations
- Backlog
-
SERVER-57168 pipeline->clone() breaks on pipelines with union stage
- Closed
-
SERVER-57668 Cache chunk bounds as an array in resharding collection cloning pipeline
- Closed