The sessions collection is sharded in order to spread sessions info across different nodes to avoid bottlenecks.
Sessions load distribution has always happened by balancing the number of chunks across shards; with the introduction of data-size aware balancing, we should find a way to spread documents across shards considering that they are very small (~150/200 bytes in average) and their cumulative size is often much less than the default max chunk size (meaning that if balancing would happen on data size, nothing would be considered eligible to move).
- causes
-
SERVER-72065 The logical sessions cache cannot be refreshed in secondary nodes of the CSRS
-
- Closed
-
-
-
- Closed
-
-
-
- Closed
-
- is related to
-
SERVER-45624 Pre-split and distribute chunks of sessions collection
-
- Closed
-