I'm using MongoDB 3.0.5 (with WiredTiger) to store about 20M docs like this:

{
"_id": ObjectId("55e7018e413a306517a980bd"),
"s": 200
"r": ["related1","related2","related3","related4","related5","related6","related7","related8","related9"],
"m": [

{ "v": ["value1"], "c": 0.9329141439596679 }

,

{ "v": ["value1", "value2"], "c": 0.85163020143722 }

,

{ "v": ["value1"], "c": 0.905295321980623 }

,

{ "v": ["value1"], "c": 0.893197290636285 }

],
"p": "ee3954d86e6228ee5e76eb3c20ffc234"
}

I have indexes on both "s" and "r" fields, but this query:

db.test.find(

{"r":"something"}

).limit(30).sort(

{"s":-1}

)

sometimes run forever and eventually timeout when using the node driver (about 50% success, 50& timeout).

I've test the same query with the mongo shell, and it always return within several ms and never got a timeout.

I can reproduce this issue on both Linux and Max, this is so strange...