-
Type:
Task
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
Replication
-
Repl 2026-10-12
-
None
-
None
-
None
-
None
-
None
-
None
-
None
The thread which takes fetched entries and writes them to the oplog collection before sending them to the apply buffer is single threaded, and only grabs oplog entries up to 16MB before writing them and sending to the apply buffer.
If we have a scenario (resharding is a common case) where we have a large amount of big oplog entries (~5MB each) then we are only able to write a couple of entries into the oplog at a time, which turns the writer thread into the bottleneck, which can cause Replication lag.
The apply buffer can hold 100MB of entries at max, so there might be room to improve throughput here by writing larger batches into the oplog if the oplog entry size is large
- duplicates
-
-
- Closed
-
- is related to
-
SERVER-135037 Ensure sufficient metrics exist to diagnose replication lag
-
- Open
-
- related to
-
-
- Closed
-
-
-
- Backlog
-