@@ -58,6 +58,16 @@ You may want to split chunks manually if:
5858 values between ``300`` and ``400``, *but* all values of your shard
5959 keys are between ``250`` and ``500`` are in a single chunk.
6060
61+ .. warning::
62+
63+ Be careful when splitting chunks. When you shard a collection that
64+ has existing data, MongoDB automatically creates chunks to evenly
65+ spread the collection. Performing additional splits requires
66+ knowledge of the resulting chunk sizes by numbers of documents and by
67+ size. You do not want splits that cause some chunks to be much larger
68+ than others. This leads to balancing based on count of chunks, not on
69+ their size, which may cause extreme load/data-distribution problems.
70+
6171Use :method:`sh.status()` to determine the current chunks ranges across
6272the cluster.
6373
@@ -101,11 +111,13 @@ chunk splits.
101111Create Chunks (Pre-Splitting)
102112-----------------------------
103113
114+ Pre-splitting lets you preemptively split chunks in an empty collection
115+ and is used *only* in certain situations.
104116In most situations a :term:`sharded cluster` will create and distribute
105117chunks automatically without user intervention. However, in a limited
106118number of use profiles, MongoDB cannot create enough chunks or
107- distribute data fast enough to support required throughput. Consider
108- the following scenarios :
119+ distribute data fast enough to support required throughput.
120+ For example, if :
109121
110122- you must partition an existing data collection that resides on a
111123 single shard.
@@ -123,7 +135,7 @@ the following scenarios:
123135Preemptively splitting chunks increases cluster throughput for these
124136operations, by reducing the overhead of migrating chunks that hold
125137data during the write operation. MongoDB only creates splits after an
126- insert operation, and can only migrate a single chunk at a time. Chunk
138+ insert operation and can migrate only a single chunk at a time. Chunk
127139migrations are resource intensive and further complicated by large
128140write volume to the migrating chunk.
129141
0 commit comments