As it is, it's possible to get oversized chunks if data is being written to the same chunk from different mongoses, as each mongos would only see a small amount of data going to that chunk and thus won't trigger a chunk split. If the decision on when to split a chunk happened on the shard, then it could know exactly when 64Mb (or whatever the chunk size has been set to) has been written to a chunk, and thus can trigger a split at the right time and prevent chunks from ever going over the chunk size.
- is depended on by
-
SERVER-16715 Distribution of data with hashed shard key suddenly biased toward few shards
- Closed
-
SERVER-11487 Don't make splits/merge bump up collection version
- Closed
- is duplicated by
-
SERVER-25060 Make SplitHeuristics.maxParallelSplits tunable
- Closed
-
SERVER-1327 Move splits and migrates out of connection thread
- Closed
- is related to
-
SERVER-9320 Several connections request to split a same chunk at the same time.
- Closed
-
SERVER-10866 Figure out split triggering mechanism
- Closed
- related to
-
SERVER-44088 Autosplitter seems to ignore some fat chunks
- Closed
-
SERVER-16963 mongos help doesn't list noAutoSplit option
- Closed
-
SERVER-24604 Add doAutoSplit to config.settings
- Closed
-
SERVER-14121 Store size for chunk in config.chunks.
- Closed
-
SERVER-20448 allow sharding.autoSplit to be set at runtime
- Closed