SplitVector fails to find the mid-point of a chunk if it is forced and there are more than 250000 records in the chunk.

To test this run the following from the mongo shell:

> use test
switched to db test
> for (i=0; i < 500000; i++) {db.col.insert(

{"i":i}

);}
> db.col.ensureIndex(

{"i":1}

)

> db.adminCommand({"splitVector": "test.col", "keyPattern":

{i:1}

, "min":

{i:0}

, "max":

{i: 250001}

, force: true})

it will output something like this:

{"timeMillis" : 81, "splitKeys": [

{"i": 250000}

], "ok": 1}

so the split point it suggests is the record just before the last one, which is pretty far from the middle.