-
Type:
Improvement
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
Query Execution
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Currently we don't have a test for using $changeStreamSplitLargeEvents in the context of large events coming simultaneously from multiple shards.
Reconstructing such events brings an additional challenge since it is a bit hard to understand which fragments belong together. One way to do that is to decode resume token. Another way to do that is to merge the fragment based on the resume token 'prefix' --> the resume tokens of a fragment is the resume token of the original large event suffixed with a 'fragmentNum' field, and for the first fragment it is always '904'.
- related to
-
SERVER-116400 Ensure split event fragments are always consecutive on sharded clusters
-
- Open
-